Data compression transmission method and device, equipment and storage medium

Abstract

The invention provides a data compression transmission method and device, equipment and a storage medium. A first device obtains M pieces of first data, one piece of sub-data in the first data corresponds to one first sparse matrix, the first sparse matrix expresses one piece of sub-data in the corresponding first data based on a first dictionary matrix, and the first dictionary matrix comprises features of M pieces of sub-data corresponding to the M pieces of first data, the first device outputs compressed data of M first sparse matrices, M being an integer greater than 1. Effective and reliable data compression transmission can be realized in a transmission scene with a relatively large data volume.

Description

Data compression transmission method, device, equipment and storage medium Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a data compression transmission method, apparatus, device, and storage medium. Background Art

[0002] At present, in some communication scenarios, such as the transmission scenario of point cloud data or the transmission scenario of artificial intelligence (AI) model data (hereinafter referred to as AI model data), when the data transmitted between communication devices is large, it will occupy more transmission resources and increase the transmission delay. Based on this, before data transmission, the data to be transmitted can be compressed in a scalar quantization or vector quantization manner, and then the compressed data can be transmitted to save transmission resources and reduce transmission delay. However, when the data to be transmitted is compressed by the existing method, the compression rate is low and the compressed data has a large data loss. How to achieve effective and reliable data compression transmission is a problem that needs to be solved urgently.

[0003] Summary of the Invention

[0004] The embodiments of the present application provide a data compression transmission method, apparatus, device, and storage medium that can achieve effective and reliable data compression transmission.

[0005] In a first aspect, an embodiment of the present application provides a data compression and transmission method, comprising: a first device obtains M first data, a sub-data in the first data corresponds to a first sparse matrix, the first sparse matrix expresses a sub-data in the corresponding first data based on a first dictionary matrix, and the first dictionary matrix includes features of M sub-data corresponding to the M first data respectively; the first device outputs compressed data of the M first sparse matrices; wherein M is an integer greater than 1.

[0006] Through the data compression and transmission method provided by the first aspect, the M sub-data corresponding to the M first data are subjected to dictionary learning based on a first dictionary matrix to obtain a sparse expression of each sub-data, thereby realizing effective and reliable data compression transmission in transmission scenarios with large data volumes.

[0007] In one possible embodiment, the first device outputs the compressed data of the M first sparse matrices, including: the first device determines the first matrix based on the M first sparse matrices; the first device performs low-rank approximation on the first matrix to obtain the first compressed data of the M first sparse matrices; and the first device outputs the first compressed data.

[0008] The data compression and transmission method provided by this embodiment jointly compresses the M first sparse matrices by means of low-rank approximation, thereby further improving the compression rate.

[0009] In one possible embodiment, the first device determines the first matrix based on the M first sparse matrices, including: the first device combines the M first sparse matrices to obtain the first matrix; or, the first device compresses data of at least one first sparse matrix among the M first sparse matrices, and combines the M first sparse matrices after data compression to obtain the first matrix.

[0010] Through the data compression and transmission method provided by this embodiment, the first device combines M first sparse matrices to obtain a first matrix with high processing efficiency, and the first device compresses data on at least one first sparse matrix and then combines M first sparse matrices to obtain a first matrix, further improving the compression rate.

[0011] In a possible embodiment, the first device performs data compression on at least one first sparse matrix among the M first sparse matrices, including: the first device sets the value of the first element in one of the M first sparse matrices to a first numerical value according to first position indication information; wherein the first position indication information is used to indicate the position of the element in the first sparse matrix that has the ability to express a corresponding sub-data, and the first element does not have the ability to express a sub-data in the first data.

[0012] Through the data compression and transmission method provided in this embodiment, elements capable of expressing a corresponding sub-data are screened from the first sparse matrix, thereby achieving data compression of the first sparse matrix.

[0013] In one possible embodiment, the first device performs low-rank approximation on the first matrix to obtain first compressed data of the M first sparse matrices, including: the first device performs singular value decomposition on the first matrix to obtain K eigenvalues ​​and eigenvectors corresponding to the K eigenvalues, and the K eigenvalues ​​and the eigenvectors corresponding to the K eigenvalues ​​are used to express the first matrix; the first device uses K0 eigenvalues ​​of the K eigenvalues ​​and the eigenvectors corresponding to the K0 eigenvalues ​​as the first compressed data of the M first sparse matrices.

[0014] Through the data compression and transmission method provided by this embodiment, K0 eigenvalues ​​and eigenvectors corresponding to the distribution of K0 eigenvalues ​​are selected from the eigenvalues ​​after singular value decomposition to achieve low-rank approximation of M first sparse matrices, thereby improving the compression rate of M first data.

[0015] In a possible implementation, it further includes: the first device outputting second compressed data of the M first sparse matrices, the second compressed data including: K1 eigenvalues ​​among the K eigenvalues ​​except the K0 eigenvalues ​​and the eigenvectors corresponding to the K1 eigenvalues ​​respectively.

[0016] The data compression transmission method provided in this embodiment transmits K1 eigenvalues ​​and the eigenvectors corresponding to the K1 eigenvalues ​​to supplement the first compressed data so that the second device can accurately construct M first data.

[0017] In one possible embodiment, the first device outputs compressed data of the M first sparse matrices, including: the first device outputs first residual information, which is determined based on information of the j-th first sparse matrix and the i-th first sparse matrix in the M first sparse matrices; wherein, i is less than j, and i and j are both positive integers.

[0018] Through the data compression and transmission method provided by this embodiment, the M first sparse matrices are jointly compressed based on the residual method, thereby further improving the compression rate of the M first data.

[0019] In one possible embodiment, the first device outputs the first residual information, including: the first device determines the similarity between the jth first sparse matrix and the i-th first sparse matrix among the M first sparse matrices; when the similarity is less than or equal to a second similarity threshold, the first device outputs the first residual information.

[0020] Through the data compression and transmission method provided by this embodiment, when the similarity between the j-th first sparse matrix and the i-th first sparse matrix is ​​less than or equal to the second similarity threshold, the first residual information is output, thereby avoiding outputting the first residual information when the j-th first sparse matrix is ​​similar to the i-th first sparse matrix, thereby improving the compression rate of the M first data.

[0021] In a possible implementation, the information of the i-th first sparse matrix is ​​obtained by decompressing compressed data of the i-th first sparse matrix.

[0022] Through the data compression transmission method provided by this embodiment, when the j-th first sparse matrix needs to be transmitted, the first residual information is determined based on the recovered i-th first sparse matrix and the j-th first sparse matrix, so that the second device can accurately construct the j-th first sparse matrix based on the first residual information.

[0023] Optionally, the first residual information includes: a first residual element sequence, which is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and both i and j are positive integers.

[0024] Optionally, the first residual information also includes: second position indication information, which is used to indicate the position of the residual element in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, whose absolute value is greater than or equal to the first residual threshold.

[0025] In a possible embodiment, it also includes: the first device outputs second residual information, and the second residual information is determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix; wherein the second residual information includes: third position indication information and a second residual element sequence, the third position indication information is used to indicate the residual elements in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold, and the second residual element sequence includes the residual elements in the residual matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold.

[0026] Through the data compression transmission method provided by this embodiment, the second residual information transmitted during the incremental process can supplement the first residual information to enrich the residual information, so that the second device can construct the jth first sparse matrix based on the first residual information and the second residual information with higher accuracy.

[0027] In a possible implementation manner, the method further includes: the first device performs a first operation on the kth first data Y among the M first data. k , the kth first data Y k A sub-data is decomposed into the first dictionary matrix and the kth first sparse matrix, where k is a positive integer less than or equal to M; the first device determines M-1 first sparse matrices other than the kth first sparse matrix according to the first dictionary matrix.

[0028] The data compression and transmission method provided by this embodiment has higher accuracy when used for data compression and data decompression, based on the first dictionary matrix obtained by decomposing one sub-data among the M sub-data.

[0029] In one possible implementation, for the p-th first data and the q-th first data among the M first data, the similarity between a sub-data in the p-th first data and a sub-data in the q-th first data is greater than or equal to a first similarity threshold; wherein p and q are both positive integers, and p is not equal to q.

[0030] Through the data compression and transmission method provided by this embodiment, the sub-data in different first data have similarities, so that the data compression of the M first sparse matrices has higher reliability.

[0031] In one possible embodiment, the M first data are data within M time units respectively, and one first data includes N sub-data divided according to spatial position relationships, where N is a positive integer; or, the M first data are data within h time units, and the data within the h time units are sorted according to spatial position relationships to obtain the M first data, and one first data includes N sub-data, where h is a positive integer; wherein the spatial position similarity between a sub-data in the p-th first data and a sub-data in the q-th first data among the M first data is greater than or equal to the first similarity threshold.

[0032] Through the data compression and transmission method provided by this embodiment, the data to be transmitted is divided to obtain M first data. On the one hand, the sparsity of the sparse matrix after dictionary learning (such as M first sparse matrices) can be improved. On the other hand, the different first data are correlated, which facilitates subsequent data compression processing.

[0033] In a possible implementation, the method further includes: the first device sending the first dictionary matrix to the second device; or the first device receiving the first dictionary matrix sent by the second device.

[0034] Through the data compression transmission method provided by this embodiment, the first dictionary matrix is ​​synchronized between the first device and the second device, so that the M first sparse matrices output by the first device can be used by the second device to accurately construct M first data.

[0035] In a possible implementation, the first device outputs the first compressed data, including: the first device performs compression processing on the first compressed data, the compression processing including quantization and / or entropy coding; and the first device outputs the first compressed data after compression processing.

[0036] The data compression transmission method provided by this embodiment further improves the compression rate of the M first sparse matrices.

[0037] In a possible implementation, the first device outputs the first residual information, including: the first device performs compression processing on the first residual information, the compression processing includes quantization and / or entropy coding; and the first device outputs the compressed first residual information.

[0038] The data compression transmission method provided by this embodiment further improves the compression rate of the M first sparse matrices.

[0039] In a possible embodiment, it also includes: the first device sends a first indication information to the second device; or the first device receives the first indication information sent by the second device; wherein the first indication information is used to indicate at least one of the following: the number of features K0 of the low-rank approximation; the number M of the first data; a capability threshold, the capability threshold is used to determine whether the elements in the first sparse matrix have the ability to express a sub-data in the corresponding first data; a first residual threshold, the first residual threshold is used to determine a first residual element sequence, the first residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix , i is less than j, and i and j are both positive integers; the proportion of elements in a first sparse matrix among the M first sparse matrices that have the ability to express a sub-data in the corresponding first data; the data loss of the first compressed data relative to the M first sparse matrices, the first compressed data is determined based on the M first sparse matrices; parameters of the compression processing, the compression processing includes quantization and / or entropy coding, the parameters of the compression processing include at least one of quantization accuracy, quantization codebook, and encoding method; whether to send the first residual information, the first residual information is determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix among the M first sparse matrices.

[0040] The data compression transmission method provided by this embodiment realizes flexible indication of data compression transmission.

[0041] In a possible implementation, it also includes: the first device determines at least one of the following based on the first time-frequency resources of the M first data: a capability threshold, the capability threshold is used to determine whether the elements in the first sparse matrix have the ability to express a sub-data in the corresponding first data; a first residual threshold, the first residual threshold is used to determine a first residual element sequence, the first residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers; parameters of compression processing, the compression processing includes quantization and / or entropy coding, and the parameters of the compression processing include at least one of quantization accuracy, quantization codebook, and encoding method.

[0042] The data compression transmission method provided by this embodiment implicitly configures one or more thresholds and / or compression processing parameters through the first time-frequency resources of M first data, thereby reducing the overhead of configuration signaling.

[0043] In a possible implementation, the method further includes: the first device receiving first configuration information sent by the second device, where the first configuration information is used to configure the first time-frequency resource.

[0044] The data compression transmission method provided by this embodiment enables flexible configuration of transmission resources.

[0045] In a possible implementation, it also includes: the first device sends a compressed transmission request to the second device, and the compressed transmission request carries the data type of the M first data, and the data type includes point cloud data and / or artificial intelligence AI data.

[0046] This implementation facilitates the second device data type to determine the time-frequency resources for transmitting compressed data of the first data, so that the first time-frequency resources configured by the second device meet the data type, facilitating compressed transmission.

[0047] In a possible embodiment, it also includes: the first device sends second indication information to the second device; or, the first device receives second indication information sent by the second device; wherein the second indication information is used to indicate at least one of the following: the number of features K1 of the low-rank approximation; a second residual threshold, the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold, and the second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers.

[0048] The data compression transmission method provided by this embodiment realizes flexible indication of data compression transmission in the incremental transmission process.

[0049] In a possible embodiment, it also includes: the first device determines a second residual threshold based on the second time-frequency resource of the M first data, and the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold, and the second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers.

[0050] The data compression transmission method provided by this embodiment reduces the overhead of configuration signaling.

[0051] In a possible implementation, the method further includes: the first device receiving second configuration information sent by the second device, where the second configuration information is used to configure the second time-frequency resource.

[0052] The data compression transmission method provided by this embodiment enables flexible configuration of transmission resources during incremental transmission.

[0053] In a second aspect, an embodiment of the present application provides a data compression and transmission method, comprising: a second device receives compressed data of M first sparse matrices, wherein the first sparse matrix expresses corresponding sub-data in the first data based on a first dictionary matrix, and the first dictionary matrix includes features of M sub-data corresponding to the M first data respectively, and one sub-data in the first data corresponds to a first sparse matrix; the second device outputs decompression information based on the compressed data.

[0054] In one possible embodiment, the compressed data includes first compressed data, and the second device outputs decompression information based on the compressed data, including: the second device performs low-rank matrix recovery based on the first compressed data to obtain a first matrix; the second device determines the M first sparse matrices based on the first matrix; the second device constructs the M first data based on the M first sparse matrices and the first dictionary matrix; and the second device outputs the M first data.

[0055] In one possible embodiment, the second device determines the M first sparse matrices based on the first matrix, including: the second device splits the first matrix to obtain the M first sparse matrices; or, the second device splits the first matrix to obtain decompression information of the M first sparse matrices, and decompresses data of at least one first sparse matrix among the M first sparse matrices based on the information of the M first sparse matrices.

[0056] In one possible embodiment, the second device performs data decompression on at least one first sparse matrix among the M first sparse matrices, including: the second device performs data decompression on a first sparse matrix among the M first sparse matrices according to first position indication information; wherein the first position indication information is used to indicate the position of an element in the first sparse matrix that has the ability to express a corresponding sub-data, and the first element does not have the ability to express a sub-data in the first data.

[0057] In one possible embodiment, the first compressed data includes K0 eigenvalues ​​and eigenvectors corresponding to the K0 eigenvalues ​​respectively, and the second device performs low-rank matrix recovery based on the first compressed data to obtain the first matrix, including: the second device performs low-rank matrix recovery based on the K0 eigenvalues ​​and the eigenvectors corresponding to the K0 eigenvalues ​​respectively to obtain the first matrix.

[0058] In a possible implementation, the compressed data further includes second compressed data, the second compressed data including K1 eigenvalues ​​and eigenvectors corresponding to the K1 eigenvalues ​​respectively, and the K1 eigenvalues ​​are different from the K0 eigenvalues.

[0059] In one possible embodiment, the compressed data includes first residual information, which is determined based on information of the j-th first sparse matrix and the i-th first sparse matrix in the M first sparse matrices. The second device outputs decompression information based on the compressed data, including: the second device constructs the j-th first sparse matrix based on the first residual information and the i-th first sparse matrix; the second device constructs the M first data based on the j-th first sparse matrix; the second device outputs the M first data; wherein, i is less than j, and i and j are both positive integers.

[0060] In a possible implementation, the information of the i-th first sparse matrix is ​​obtained by decompressing compressed data of the i-th first sparse matrix.

[0061] In one possible embodiment, the first residual information includes: a first residual element sequence, which is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and i and j are both positive integers.

[0062] In one possible embodiment, the first residual information also includes: second position indication information, which is used to indicate the position of the residual element whose absolute value is greater than or equal to the first residual threshold in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix.

[0063] In one possible embodiment, the compressed data also includes second residual information, which is determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix; wherein the second residual information includes: third position indication information and a second residual element sequence, the third position indication information is used to indicate the residual elements in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold, and the second residual element sequence includes the residual elements in the residual matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold.

[0064] In one possible implementation, for the p-th first data and the q-th first data among the M first data, the similarity between a sub-data in the p-th first data and a sub-data in the q-th first data is greater than or equal to a first similarity threshold; wherein p and q are both positive integers, and p is not equal to q.

[0065] In a possible embodiment, the M first data are data within M time units respectively; or, the M first data are data within h time units, and the data within the h time units are sorted according to the spatial position relationship to obtain the M first data, one first data includes N sub-data, and h is a positive integer; wherein the spatial position similarity between a sub-data in the p-th first data and a sub-data in the q-th first data among the M first data is greater than or equal to the first similarity threshold.

[0066] In a possible implementation, the method further includes: the second device receiving the first dictionary matrix sent by the first device; or the second device sending the first dictionary matrix to the first device.

[0067] In a possible embodiment, it also includes: the second device receives the first indication information sent by the first device; or the second device sends the first indication information to the first device; wherein the first indication information is used to indicate at least one of the following: the number of features K0 of the low-rank approximation; the number M of the first data; a capability threshold, the capability threshold is used to determine whether the elements in the first sparse matrix have the ability to express a sub-data in the corresponding first data; a first residual threshold, the first residual threshold is used to determine a first residual element sequence, the first residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix , i is less than j, and i and j are both positive integers; the proportion of elements in a first sparse matrix among the M first sparse matrices that have the ability to express a sub-data in the corresponding first data; the data loss of the first compressed data relative to the M first sparse matrices, the first compressed data is determined based on the M first sparse matrices; parameters of the compression processing, the compression processing includes quantization and / or entropy coding, the parameters of the compression processing include at least one of quantization accuracy, quantization codebook, and encoding method; whether to send the first residual information, the first residual information is determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix among the M first sparse matrices.

[0068] In a possible implementation, the further step includes: the second device sending first configuration information to the first device, where the first configuration information is used to configure a first time-frequency resource for the first data of the M.

[0069] In a possible implementation, it also includes: the second device receiving a compressed transmission request sent by the first device, the compressed transmission request carrying a data type of the M first data, and the data type includes point cloud data and / or AI data.

[0070] In a possible embodiment, it also includes: the second device receives second indication information sent by the first device; or the second device sends second indication information to the first device; wherein the second indication information is used to indicate at least one of the following: the number of features K1 of the low-rank approximation; a second residual threshold, the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold, and the second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers.

[0071] In a possible implementation, the method further includes: the first device receiving second configuration information sent by the second device, where the second configuration information is used to configure second time-frequency resources for the M first data.

[0072] The beneficial effects of the data compression transmission method provided by the second aspect and its possible implementation methods can be found in the beneficial effects brought about by the first aspect and its possible implementation methods, and will not be repeated here.

[0073] In a third aspect, an embodiment of the present application provides a communication device, comprising: a processing module for obtaining M first data, where a sub-data in the first data corresponds to a first sparse matrix, and the first sparse matrix expresses a sub-data in the corresponding first data based on a first dictionary matrix, and the first dictionary matrix includes features of the M sub-data corresponding to the M first data respectively; a transceiver module for outputting compressed data of the M first sparse matrices; wherein M is an integer greater than 1.

[0074] In one possible embodiment, the processing module is also used to: determine a first matrix based on the M first sparse matrices; perform low-rank approximation on the first matrix to obtain first compressed data of the M first sparse matrices; and the transceiver module is used to output the first compressed data.

[0075] In one possible embodiment, the processing module is specifically used to: combine the M first sparse matrices to obtain the first matrix; or, the first device performs data compression on at least one first sparse matrix among the M first sparse matrices, and combines the M first sparse matrices after data compression to obtain the first matrix.

[0076] In one possible embodiment, the processing module is specifically used to: for a first sparse matrix among the M first sparse matrices, set the value of the first element in the first sparse matrix to a first numerical value according to the first position indication information; wherein the first position indication information is used to indicate the position of the element in the first sparse matrix that has the ability to express a corresponding sub-data, and the first element does not have the ability to express a sub-data in the first data.

[0077] In one possible embodiment, the processing module is specifically used to: perform singular value decomposition on the first matrix to obtain K eigenvalues ​​and eigenvectors corresponding to the K eigenvalues, and the K eigenvalues ​​and the eigenvectors corresponding to the K eigenvalues ​​are used to express the first matrix; K0 eigenvalues ​​of the K eigenvalues ​​and the eigenvectors corresponding to the K0 eigenvalues ​​are used as the first compressed data of the M first sparse matrices.

[0078] In a possible implementation, the transceiver module is further configured to output second compressed data of the M first sparse matrices, the second compressed data including: K1 eigenvalues ​​among the K eigenvalues ​​except the K0 eigenvalues ​​and eigenvectors corresponding to the K1 eigenvalues ​​respectively.

[0079] In one possible embodiment, the transceiver module is specifically used to: output first residual information, which is determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix in the M first sparse matrices; wherein i is less than j, and i and j are both positive integers.

[0080] In one possible embodiment, the processing module is also used to determine the similarity between the jth first sparse matrix and the i-th first sparse matrix in the M first sparse matrices; when the similarity is less than or equal to a second similarity threshold, the transceiver module outputs the first residual information.

[0081] In a possible implementation, the information of the i-th first sparse matrix is ​​obtained by decompressing compressed data of the i-th first sparse matrix.

[0082] In one possible embodiment, the first residual information includes: a first residual element sequence, which is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and i and j are both positive integers.

[0083] In a possible implementation, the first residual information further includes: second position indication information, where the second position indication information is used to indicate that the absolute value in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix is ​​greater than or equal to.

[0084] In one possible embodiment, the transceiver module is also used to output second residual information, which is determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix; wherein the second residual information includes: third position indication information and a second residual element sequence, the third position indication information is used to indicate the residual elements in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold, and the second residual element sequence includes the residual elements in the residual matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold.

[0085] In a possible implementation manner, the processing module is further configured to: for the kth first data Y among the M first data k , the kth first data Y k A sub-data is decomposed into the first dictionary matrix and the kth first sparse matrix, where k is a positive integer less than or equal to M; the first device determines M-1 first sparse matrices other than the kth first sparse matrix according to the first dictionary matrix.

[0086] In one possible implementation, for the p-th first data and the q-th first data among the M first data, the similarity between a sub-data in the p-th first data and a sub-data in the q-th first data is greater than or equal to a first similarity threshold; wherein p and q are both positive integers, and p is not equal to q.

[0087] In a possible embodiment, the M first data are data within M time units respectively; or, the M first data are data within h time units, and the data within the h time units are sorted according to the spatial position relationship to obtain the M first data, one first data includes N sub-data, and h is a positive integer; wherein the spatial position similarity between a sub-data in the p-th first data and a sub-data in the q-th first data among the M first data is greater than or equal to the first similarity threshold.

[0088] In a possible implementation, the transceiver module is further configured to: send the first dictionary matrix to the second device; or receive the first dictionary matrix sent by the second device.

[0089] In a possible implementation, the transceiver module is specifically configured to: perform compression processing on the first compressed data, where the compression processing includes quantization and / or entropy coding; and output the compressed first compressed data.

[0090] In a possible implementation, the transceiver module is specifically configured to: perform compression processing on the first residual information, where the compression processing includes quantization and / or entropy coding; and output the compressed first residual information.

[0091] In a possible embodiment, it also includes: the transceiver module is also used to: send a first indication message to the second device; or, receive the first indication message sent by the second device; wherein the first indication message is used to indicate at least one of the following: the number of features K0 of the low-rank approximation; the number M of the first data; a capability threshold, the capability threshold is used to determine whether the elements in the first sparse matrix have the ability to express a sub-data in the corresponding first data; a first residual threshold, the first residual threshold is used to determine a first residual element sequence, the first residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix , i is less than j, and i and j are both positive integers; the proportion of elements in a first sparse matrix among the M first sparse matrices that have the ability to express a sub-data in the corresponding first data; the data loss of the first compressed data relative to the M first sparse matrices, the first compressed data is determined based on the M first sparse matrices; parameters of the compression processing, the compression processing includes quantization and / or entropy coding, the parameters of the compression processing include at least one of quantization accuracy, quantization codebook, and encoding method; whether to send the first residual information, the first residual information is determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix among the M first sparse matrices.

[0092] In one possible embodiment, the processing module is also used to determine at least one of the following based on the first time-frequency resources of the M first data: a capability threshold, which is used to determine whether the elements in the first sparse matrix have the ability to express a sub-data in the corresponding first data; a first residual threshold, which is used to determine a first residual element sequence, which is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers; parameters of compression processing, which include quantization and / or entropy coding, and the parameters of compression processing include at least one of quantization accuracy, quantization codebook, and encoding method.

[0093] In a possible implementation, the transceiver module is further configured to receive first configuration information sent by a second device, where the first configuration information is used to configure the first time-frequency resource.

[0094] In a possible implementation, the transceiver module is further used to send a compressed transmission request to the second device, where the compressed transmission request carries the data type of the M first data, and the data type includes point cloud data and / or AI data.

[0095] In one possible embodiment, the transceiver module is also used to: send second indication information to the second device; or, receive second indication information sent by the second device; wherein the second indication information is used to indicate at least one of the following: the number of features K1 of the low-rank approximation; a second residual threshold, the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold, and the second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers.

[0096] In one possible embodiment, the processing module is also used to determine a second residual threshold based on the second time-frequency resource of the M first data, and the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold, and the second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and i and j are both positive integers.

[0097] In a possible implementation, the transceiver module is further configured to receive second configuration information sent by a second device, where the second configuration information is used to configure the second time-frequency resource.

[0098] The beneficial effects of the communication device provided by the third aspect and each possible implementation of the third aspect can be referred to the beneficial effects brought about by the first aspect and each possible implementation of the first aspect, and will not be repeated here.

[0099] In a fourth aspect, an embodiment of the present application provides a communication device, comprising: a transceiver module for receiving compressed data of M first sparse matrices, wherein the first sparse matrix expresses sub-data in the corresponding first data based on a first dictionary matrix, and the first dictionary matrix includes features of M sub-data corresponding to the M first data respectively, and one sub-data in the first data corresponds to a first sparse matrix; a processing module for outputting decompression information based on the compressed data.

[0100] In a possible embodiment, the compressed data includes first compressed data, and the processing module is specifically used to: perform low-rank matrix recovery based on the first compressed data to obtain a first matrix; determine the M first sparse matrices based on the first matrix; construct the M first data based on the M first sparse matrices and the first dictionary matrix; and the second device outputs the M first data.

[0101] In one possible embodiment, the processing module is specifically used to: split the first matrix to obtain the M first sparse matrices; or, split the first matrix to obtain decompression information of the M first sparse matrices, and decompress data of at least one first sparse matrix among the M first sparse matrices based on the information of the M first sparse matrices.

[0102] In one possible embodiment, the processing module is specifically used to: for a first sparse matrix among the M first sparse matrices, decompress the data of the first sparse matrix according to the first position indication information; wherein the first position indication information is used to indicate the position of an element in the first sparse matrix that has the ability to express a corresponding sub-data, and the first element does not have the ability to express a sub-data in the first data.

[0103] In one possible embodiment, the first compressed data includes K0 eigenvalues ​​and eigenvectors corresponding to the K0 eigenvalues ​​respectively, and the second device performs low-rank matrix recovery based on the first compressed data to obtain the first matrix, including: the second device performs low-rank matrix recovery based on the K0 eigenvalues ​​and the eigenvectors corresponding to the K0 eigenvalues ​​respectively to obtain the first matrix.

[0104] In a possible implementation, the compressed data further includes second compressed data, the second compressed data including K1 eigenvalues ​​and eigenvectors corresponding to the K1 eigenvalues ​​respectively, and the K1 eigenvalues ​​are different from the K0 eigenvalues.

[0105] In one possible embodiment, the compressed data includes first residual information, which is determined based on information of the j-th first sparse matrix and the i-th first sparse matrix in the M first sparse matrices. The second device outputs decompression information based on the compressed data, including: the second device constructs the j-th first sparse matrix based on the first residual information and the i-th first sparse matrix; the second device constructs the M first data based on the j-th first sparse matrix; the second device outputs the M first data; wherein, i is less than j, and i and j are both positive integers.

[0106] In a possible implementation, the information of the i-th first sparse matrix is ​​obtained by decompressing compressed data of the i-th first sparse matrix.

[0107] In one possible embodiment, the first residual information includes: a first residual element sequence, which is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and i and j are both positive integers.

[0108] In one possible embodiment, the first residual information also includes: second position indication information, which is used to indicate the position of the residual element whose absolute value is greater than or equal to the first residual threshold in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix.

[0109] In one possible embodiment, the compressed data also includes second residual information, which is determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix; wherein the second residual information includes: third position indication information and a second residual element sequence, the third position indication information is used to indicate the residual elements in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold, and the second residual element sequence includes the residual elements in the residual matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold.

[0110] In one possible implementation, for the p-th first data and the q-th first data among the M first data, the similarity between a sub-data in the p-th first data and a sub-data in the q-th first data is greater than or equal to a first similarity threshold; wherein p and q are both positive integers, and p is not equal to q.

[0111] In a possible embodiment, the M first data are data within M time units respectively; or, the M first data are data within h time units, and the data within the h time units are sorted according to the spatial position relationship to obtain the M first data, one first data includes N sub-data, and h is a positive integer; wherein the spatial position similarity between a sub-data in the p-th first data and a sub-data in the q-th first data among the M first data is greater than or equal to the first similarity threshold.

[0112] In a possible implementation, the transceiver module is further configured to: receive the first dictionary matrix sent by the first device; or send the first dictionary matrix to the first device.

[0113] In a possible embodiment, the transceiver module is also used to: receive a first indication message sent by the first device; or send a first indication message to the first device; wherein the first indication message is used to indicate at least one of the following: the number of features K0 of the low-rank approximation; the number M of the first data; a capability threshold, the capability threshold is used to determine whether the elements in the first sparse matrix have the ability to express a sub-data in the corresponding first data; a first residual threshold, the first residual threshold is used to determine a first residual element sequence, the first residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is small in j, and i and j are both positive integers; the proportion of elements in a first sparse matrix among the M first sparse matrices that have the ability to express a sub-data in the corresponding first data; the data loss of the first compressed data relative to the M first sparse matrices, the first compressed data is determined based on the M first sparse matrices; parameters of the compression processing, the compression processing includes quantization and / or entropy coding, the parameters of the compression processing include at least one of quantization accuracy, quantization codebook, and encoding method; whether to send first residual information, the first residual information is determined based on information of the j-th first sparse matrix and the i-th first sparse matrix among the M first sparse matrices.

[0114] In a possible implementation, the transceiver module is further used to: send first configuration information to the first device, where the first configuration information is used to configure a first time-frequency resource for the first data of the M.

[0115] In a possible embodiment, the transceiver module is also used to receive a compressed transmission request sent by the first device, and the compressed transmission request carries the data type of the M first data, and the data type includes point cloud data and / or artificial intelligence AI data.

[0116] In one possible embodiment, the transceiver module is also used to: receive second indication information sent by the first device; or, send second indication information to the first device; wherein the second indication information is used to indicate at least one of the following: the number of features K1 of the low-rank approximation; a second residual threshold, the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold, and the second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers.

[0117] In a possible implementation, the transceiver module is further configured to receive second configuration information sent by a second device, where the second configuration information is used to configure second time-frequency resources for the M first data.

[0118] The beneficial effects of the communication device provided by the fourth aspect and each possible implementation of the fourth aspect can be referred to the beneficial effects brought about by the first aspect and each possible implementation of the first aspect, and will not be repeated here.

[0119] In a fifth aspect, an embodiment of the present application provides a communication device, comprising: a processor, wherein the processor is configured to execute the method in the first aspect, the second aspect, or each possible embodiment by running a computer program or through a logic circuit.

[0120] In a possible implementation, the device further includes a memory configured to store the computer program.

[0121] In a possible implementation, the device further includes a communication interface for inputting and / or outputting signals.

[0122] In a sixth aspect, an embodiment of the present application provides a communication device, comprising: a processor and a memory, the memory being used to store a computer program, the processor being used to call and run the computer program stored in the memory, and executing the method as in the first aspect, the second aspect, or each possible implementation.

[0123] In the seventh aspect, an embodiment of the present application provides a chip, including: a processor, for calling and running computer instructions from a memory, so that a device equipped with the chip executes the method in the first aspect, the second aspect or each possible implementation method.

[0124] In an eighth aspect, an embodiment of the present application provides a computer-readable storage medium for storing computer program instructions, wherein the computer program enables a computer to execute the method in the first aspect, the second aspect, or each possible implementation manner.

[0125] In a ninth aspect, an embodiment of the present application provides a computer program product, comprising computer program instructions, which enable a computer to execute the method in the first aspect, the second aspect, or each possible implementation manner.

[0126] In a tenth aspect, an embodiment of the present application provides a computer program that enables a computer to execute the method in the first aspect, the second aspect, or each possible implementation manner described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0127] FIG1 is a schematic diagram of the architecture of a mobile communication system used in an embodiment of the present application.

[0128] FIG2 is a schematic diagram of a dictionary learning framework provided in an embodiment of the present application.

[0129] FIG3 is a schematic diagram of an interactive process of a data compression transmission method provided in an embodiment of the present application.

[0130] FIG4a is a schematic diagram of data partitioning provided in an embodiment of the present application.

[0131] FIG4 b is another data partitioning schematic diagram provided in an embodiment of the present application.

[0132] FIG5 is another data partitioning diagram provided in an embodiment of the present application.

[0133] FIG6 is a schematic diagram of data compression of a sparse matrix provided in an embodiment of the present application.

[0134] FIG7 is a schematic diagram of an interactive process of another data compression transmission method provided in an embodiment of the present application.

[0135] FIG8 is a schematic block diagram of a communication device provided in an embodiment of the present application.

[0136] FIG9 is another schematic block diagram of a communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0137] The technical solution in this application will be described below with reference to the accompanying drawings.

[0138] Figure 1 is a schematic diagram of the architecture of a mobile communication system used in an embodiment of the present application. As shown in Figure 1, the mobile communication system includes a core network device 110, a network device 120, and at least one terminal device (such as terminal device 130 and terminal device 140 in Figure 1). The terminal device is connected to the network device wirelessly, and the network device is connected to the core network device wirelessly or by wire. The core network device and the network device can be independent and different physical devices, or the functions of the core network device and the logical functions of the network device can be integrated into the same physical device, or a physical device can integrate some of the functions of the core network device and some of the functions of the network device. The terminal device can be fixed or mobile. Figure 1 is only a schematic diagram, and the communication system can also include other network devices, such as wireless relay devices and wireless backhaul devices, which are not shown in Figure 1. The embodiments of the present application do not limit the number of core network devices, network devices, and terminal devices included in the mobile communication system.

[0139] In the embodiments of the present application, the network device may be any device with wireless transceiver functions. The network device includes, but is not limited to, an evolved Node B (eNB), a home evolved Node B (HNB), a baseband unit (BBU), an access point (AP) in a wireless fidelity (WiFi) system, a wireless relay node, a wireless backhaul node, a transmission point (TP), or a transmission and reception point (TRP). The network device may also be a mobile switching center, a device that performs base station functions in device-to-device (D2D), vehicle-to-everything (V2X), machine-to-machine (M2M), and drone communications, and a network device in a non-terrestrial network (NTN) communication system (i.e., a network device that can be deployed on a high-altitude platform, satellite, or high-altitude aircraft). It can also be a gNB in ​​a 5G system, one or a group of antenna panels (including multiple antenna panels) of a base station in a 5G system, or it can also be a network node constituting a gNB or a transmission point, such as a BBU, or a distributed unit (DU), etc. The embodiments of the present application do not specifically limit this.

[0140] In some deployments, the gNB may include a centralized unit (CU) and a DU. The CU and DU each implement portions of the gNB's functionality, and the CU and DU can communicate over the F1 interface. The gNB may also include an active antenna unit (AAU). The AAU implements some physical layer processing, RF processing, and active antenna-related functions.

[0141] It is understood that the network device may include one or more of a CU node, a DU node, and an AAU node. In addition, the CU may be classified as a network device in an access network (RAN) or a network device in a core network (CN), which is not limited in this application.

[0142] In an embodiment of the present application, the terminal device may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device.

[0143] The terminal device may be a device that provides voice / data connectivity to users, such as a handheld device or vehicle-mounted device with wireless connection function. At present, some examples of terminals include: mobile phones, tablet computers, computers with wireless transceiver functions (such as laptops, PDAs, etc.), drones, customer-premises equipment (CPE), smart point of sale (POS) machines, mobile internet devices (MID), virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in remote medical, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, cellular phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), and so on. assistant, PDA), handheld devices with wireless communication capabilities, computing devices or other processing devices connected to a wireless modem, vehicle-mounted devices, wearable devices, terminal devices in a 5G network or terminal devices in a system evolved after 5G, etc.

[0144] The network device and the terminal device can communicate through the licensed spectrum, or through the unlicensed spectrum, or through both the licensed spectrum and the unlicensed spectrum. The network device and the terminal device can communicate through the spectrum below 6G, or through the spectrum of 6G and above, or through both the spectrum below 6G and the spectrum of 6G and above. The embodiments of the present application do not limit the spectrum resources used between the network device and the terminal device.

[0145] This application does not limit the specific forms of network devices and terminal devices.

[0146] The communication method provided in this application can be applied to various communication systems, such as Long Term Evolution (LTE) systems, 5G mobile communication systems, and 6G mobile communication systems that evolve after 5G. The 5G mobile communication system or the 6G mobile communication system may include non-standalone (NSA) and / or standalone (SA) networking.

[0147] The communication method provided in this application can also be applied to machine type communication (MTC), long term evolution technology for machine-to-machine communication (LTE-M), device-to-device (D2D) networks, machine-to-machine (M2M) networks, Internet of Things (IoT) networks, or other networks.

[0148] Point cloud data refers to a collection of spatial points in a three-dimensional coordinate system. For example, data is collected by machine vision sensors and recorded as points. Each spatial point contains three-dimensional coordinates and may include color information (RGB), position information, reflection intensity information, etc. In a point cloud scenario, terminal devices (such as 130 and / or 140 in Figure 1) can collect data through sensors and transmit the collected data to a network device (such as 120 in Figure 1). The network device (such as 120 in Figure 1) can reconstruct the point cloud data.

[0149] Federated learning is a distributed machine learning technology that performs distributed model training among multiple data sources with local data. Without exchanging local individual or sample data, it only exchanges model parameters or intermediate results to build a global model based on virtual fusion data, thereby achieving a balance between data privacy protection and data sharing computing. The model parameters, AI gradients or intermediate results transmitted during the federated learning process can be called AI model data. Of course, AI model data is only a possible naming method, and this application does not limit its name. For example, it can also be called model data, AI data, etc. In a federated learning scenario, a terminal device (such as 130 and / or 140 in Figure 1) can transmit locally updated AI model data to a network device (such as 120 in Figure 1) to update a global AI model data.

[0150] Point cloud data and AI model data are large in size. Generally speaking, point cloud data can contain hundreds of thousands of spatial points, and AI model data can reach tens of millions of dimensions.

[0151] Currently, there is a lack of effective and reliable compression transmission solutions for communication scenarios with large amounts of transmitted data, such as point cloud scenarios and federated learning scenarios. In response to this, in an embodiment of the present application, for multiple data to be transmitted (such as the M first data below, where M is an integer greater than 1), dictionary learning is performed on a sub-data in each data based on the same dictionary matrix to obtain a sparse representation of each sub-data, thereby achieving effective and reliable data compression for the large amount of data to be transmitted.

[0152] The above examples are only used to illustrate the uplink transmission of point cloud data and AI model data, but should not be understood as any limitation of this application. For example, in a point cloud scenario, terminal device A (such as 130 in Figure 1) can send point cloud data to terminal device B (such as 140 in Figure 1) through forwarding by a network device (such as 120 in Figure 1); or, terminal device A can send point cloud data to terminal device B through a side link; for another example, in a federated learning scenario, a network device (such as 120 in Figure 1) can transmit locally updated AI model data to a terminal device (such as 130 and / or 140 in Figure 1).

[0153] The data compression and transmission method provided in the embodiments of the present application is mainly described by taking its application in a communication system as an example. However, the present application is not limited thereto. For example, the data compression and transmission method provided in the present application can also be applied to any electronic device with processing capabilities, which can be any of the above-mentioned terminal devices or servers. The electronic device can implement data compression based on the method provided in the embodiments of the present application and output the compressed data, such as sending the compressed data to another electronic device.

[0154] To facilitate understanding of this application, dictionary learning is first exemplified.

[0155] Figure 2 is a schematic diagram of a dictionary learning framework provided by an embodiment of the present application. As shown in Figure 2, source data Y can be represented by a matrix with L rows and Q columns, also referred to as a source matrix. Through the numerical iteration process of dictionary learning, source data Y is decomposed into a dictionary matrix D and a sparse matrix X, i.e., Y = DX. Each element in the matrix representation of source data Y can be a floating-point number.

[0156] The dictionary matrix D, or base, may be an L-row by L-column matrix, and each column in the dictionary matrix D may be referred to as a base vector. The dictionary matrix D includes features of the source data, for example, features of the source data expressed by each base vector.

[0157] The sparse matrix X can represent the source data based on the dictionary matrix D. The sparse matrix X uses a combination of basis vectors in the dictionary matrix D, for example, by weighting and combining the basis vectors in the dictionary matrix D to express the characteristics of the source data. The sparse matrix X is a matrix with L rows and Q columns. Each column vector in the sparse matrix X can represent each column vector of the source matrix Y. For example, the first column vector l1 in the sparse matrix X can represent y1 in the source matrix.

[0158] Each row vector in the sparse matrix X corresponds to each basis vector in the dictionary matrix D, and the values ​​of each row in a column of the sparse matrix X represent the weights of each basis vector of the dictionary matrix D. For example, the first column vector x1 in the sparse matrix X is used to express y1 in the source data Y, and the elements of each row in the first column vector of the sparse matrix X are x, 11 ,x 12, ……,x 1L , then x 11 The weight of the first basis vector d1 in the dictionary matrix used when the sparse matrix X expresses the source matrix Y.

[0159] The elements in the sparse matrix X can be represented or saved in a coordinate format. Each element can include the row information, column information, and value of the element in the sparse matrix X. Therefore, the elements in the sparse matrix X can also be called coordinates.

[0160] The more zero elements in the sparse matrix X, the less resources are occupied by information with low similarity to the learning task, so as to achieve better expression of the source data Y while reducing the overhead of storage resources and transmission resources. In this case, the sparse matrix X is considered to have better sparse performance.

[0161] The communication method provided in the embodiments of the present application will be described below with reference to the accompanying drawings.

[0162] It should be understood that the following description is for ease of understanding and explanation only, and the method provided in the embodiments of the present application is described with the first device as the execution subject. The first device can be any terminal device in the communication system shown in Figure 1, such as terminal device 130 or terminal device 140, or the first device can be network device 120 in the communication system shown in Figure 1. In some embodiments, the method provided in the embodiments of the present application will be described using the interaction between the first device and the second device as an example. When the data compression transmission method provided by the embodiment of the present application is applied to uplink transmission, the first device may be the terminal device 130 or the terminal device 140 in the communication system shown in Figure 1, and the second device may be the network device 120 in the communication system shown in Figure 1; when the data compression transmission method provided by the embodiment of the present application is applied to downlink transmission, the first device may be the network device 120 in the communication system shown in Figure 1, and the second device may be the terminal device 130 or the terminal device 140 in the communication system shown in Figure 1; when the data compression transmission method provided by the embodiment of the present application is applied to sidelink transmission, the first communication device may be any terminal device in the communication system shown in Figure 1, such as the terminal device 130, and the second communication device may be any terminal device in the communication system shown in Figure 1 except the first communication device, such as the terminal device 140.

[0163] It should also be understood that this should not constitute any limitation on the execution subject of the method provided in the present application. As long as it is possible to execute the method provided in the embodiment of the present application by running a program having the code of the method provided in the embodiment of the present application, it can serve as the execution subject of the method provided in the embodiment of the present application. For example, any of the above-mentioned communication devices can be implemented as a terminal device or as a component in a terminal device, such as a chip, a chip system or other functional module that can call a program and execute a program; any of the above-mentioned communication devices can be implemented as a network device or as a component in a network device, such as a chip, a chip system or other functional module that can call a program and execute a program.

[0164] FIG3 is a schematic diagram of an interactive process of a data compression transmission method provided by an embodiment of the present application. In conjunction with FIG3 , the method 200 includes some or all of the following processes:

[0165] S210, the first device obtains M first data, one sub-data in the first data corresponds to a first sparse matrix, the first sparse matrix expresses the corresponding sub-data in the first data based on a first dictionary matrix, and the first dictionary matrix includes features of the M sub-data corresponding to the M first data respectively.

[0166] S220, the first device sends the compressed data of the M first sparse matrices to the second device, and correspondingly, the second device receives the compressed data of the M first sparse matrices sent by the first device.

[0167] S230: The second device outputs decompressed data based on the compressed data.

[0168] The M first data may be M sub-data of the data to be transmitted. In other words, the M first data may be obtained by dividing the data to be transmitted. Therefore, the M first data may have similarities in the time dimension and / or the spatial dimension. Of course, this application is not limited to this. For example, the M first data may be any M data in the data to be transmitted.

[0169] It should be noted that if the data to be transmitted is used as the above-mentioned source data Y, data compression is performed based on dictionary learning. When the amount of the first data is large, the spatial distribution of the source data Y is relatively wide, and it is difficult to determine the dictionary matrix to obtain a sparse matrix with better sparsity, resulting in a more complex data processing process for decomposing the dictionary matrix D and the sparse matrix X, and a longer delay. Based on this, in an embodiment of the present application, the first device can split the data to be transmitted into M first data, and then perform dictionary learning on each first data. Furthermore, if the first data still has a large amount of data, further splitting the first data will facilitate further improving the sparsity of the sparse matrix. The embodiment of the present application may include but is not limited to the following two data division methods. The following is an exemplary description of the two data division methods in conjunction with Figures 4a and 4b.

[0170] Method 1: divide data according to the time dimension.

[0171] As shown in FIG4a , in the above-mentioned method 1, the data to be transmitted may include data within M time units, the data within each time unit being treated as a first data, and each time unit may include N sub-data, where N is a positive integer. A time unit may be one or more data frames, one or more time slots, etc., and should not be understood as a minimum time unit. Adjacent time units in the M time units may be two consecutive time units, or may be two time units with a time interval. To ensure similarity between the M time units, the time interval between the two time units should be less than or equal to a preset time interval.

[0172] It should be noted that in the above-mentioned method 1, there is a sub-data related to the spatial dimension in each of the M first data, that is, in the M sub-data corresponding to each of the M first data, the similarity of the spatial position between any two sub-data is greater than or equal to the first similarity threshold. For example, the similarity of the spatial position between a sub-data in the p-th first data and a sub-data in the q-th first data is greater than or equal to the first similarity threshold, where p and q are positive integers and p is not equal to q. Referring to FIG4a, the first sub-data Y in the first first data Y1 is (1,1) , and the first sub-data Y in the second first data Y2 (2,1) The similarity between them is greater than or equal to the first similarity threshold; the first sub-data Y in the second first data Y2 (2,1) , and the first sub-data Y in the third first data Y3 (3,1) The similarity between them is greater than or equal to the first similarity threshold...the M-1th first data Y M-1 The first sub-data Y in (M-1,1) , and the Mth first data Y M The first sub-data Y in (M,1) The similarity between them is greater than or equal to the first similarity threshold. Similarly, in FIG4a, the second sub-data Y in the first data Y1 is (1,2) , and the second sub-data Y in the second first data Y2 (2,2) The similarity between them is greater than or equal to the first similarity threshold; the second sub-data Y in the second first data Y2 (2,2) , and the second sub-data Y in the third first data Y3 (3,2) The similarity between them is greater than or equal to the first similarity threshold...the M-1th first data Y M-1 The second sub-data Y in (M-1,2) , and the Mth first data Y M The second sub-data Y in (M,2) The similarity between them is greater than or equal to the first similarity threshold. Other sub-data in the first data may also have the above similarity, which will not be described in detail for the sake of brevity. However, this application is not limited to this. For example, any first data may include at least one sub-data that does not have the above similarity with the sub-data in other first data. And the number of sub-data included in the M first data may be different. For the convenience of description below, the correlation between each n-th sub-data in the M first data is used as an example for explanation, where n is a positive integer less than or equal to N.

[0173] In the above-mentioned method 1, for any one of the M first data (such as the m-th first data, where m is a positive integer less than or equal to M), the first data may be a matrix, or in other words, the first data may be expressed in the form of a matrix. The first device may arrange the first data into a matrix Y having L rows and Q columns. m For example, L is equal to the dimension of the spatial point, that is, L=3, and Q is equal to the total number of spatial points in the point cloud data; for another example, the first data is AI model data, and L can be a preset value, such as a value obtained through experiments to make the sparse performance of the sparse matrix better, then in, Indicates rounding up, Dim is the gradient dimension. In this case, LQ-Dim zeros need to be added to the first data Y.

[0174] The first device may split the first data into N sub-data, for example, according to the spatial relationship between the data in the first data, split the first data into N sub-data. Each sub-data may be a matrix.

[0175] As shown in FIG5 , the mth first data Y m The Q columns in (such as y1 to y Q ) can be divided into N sub-data, such as y1 to y K Divide into the first sub-data Y(m,1), y K+1 to y 2K Divide into the second sub-data Y(m,2)...y Q-K+1 to y Q Divide into the Nth sub-data Y(m,N=Q / K). Figure 5 illustrates the example of a matrix formed by each of the N sub-data having the same number of columns, but this application is not limited to this. For example, if at least two of the N sub-data have different numbers of matrices, the number of columns of each matrix can be reduced. When the number of columns of the matrices formed by each of the sub-data is the same, the data processing complexity of the first communication device and the second communication device can be reduced.

[0176] Taking the first data as point cloud data as an example, the first device can divide the first data into N sub-data according to the distance between each spatial point in the first data in the Euclidean space. For example, P spatial points can be determined in the order of the distance between the spatial points from small to large, P is a preset value, and the P spatial points are divided into one sub-data; or the spatial points within a preset range are divided into a sub-value, for example, the spatial points in a circle with a spatial point as the center and R as the radius are divided into one sub-data, R is a preset value, and so on. In different first data, the sub-data related in the spatial dimension can be the data of the same or similar spatial points in the Euclidean space. For example, the nth sub-data Y in the first first data Y1 (1,n) , the nth sub-data Y in the second first data Y2(2,n) ...Nth sub-data Y in the Mth first data (M,N) , which are all data of the same (same) spatial point in different time units.

[0177] Taking the first data as AI model data as an example, the first device can divide the first data into M sub-data according to the neural network layer where each data in the first data is located, for example, dividing the data of the same neural network layer or multiple similar neural network layers into one sub-data. If the number of data in the same neural network layer exceeds the data volume limit of one sub-data, the data of the same neural network layer can be divided into two or more adjacent sub-data. Neural network layers include but are not limited to fully connected layers, convolutional layers, pooling layers, etc. In different first data, sub-data that are related in the spatial dimension may be data of the same or similar neural network layers. For example, the nth sub-data Y in the first first data Y1 (1,n) , the nth sub-data Y in the second first data Y2 (2,n) ...Nth sub-data Y in the Mth first data (M,N) , which are all data of the same neural network layer (or similar neural network layers) in different time units.

[0178] It should also be understood that similarity between sub-data can be determined not only by spatial similarity but also by temporal similarity. For example, if the difference in acquisition time between a sub-data in the pth first data and a sub-data in the qth first data is less than a preset time difference, then the similarity between the two sub-data is greater than or equal to a first similarity threshold. The similarity between sub-data can also be expressed as a difference less than or equal to a preset difference threshold. The difference between sub-data can be determined based on the residual difference between the sub-data.

[0179] Method 2: divide data according to spatial dimensions.

[0180] As shown in FIG4b , in the second approach, the data to be transmitted may be data within h time units, where h is a positive integer. The M first data may be obtained by sorting and dividing the data to be transmitted within the h time units according to a spatial position relationship. The time units have been described in the first approach and are not further described for the sake of brevity.

[0181] The data to be transmitted may be a matrix, or in other words, the data to be transmitted may be expressed in the form of a matrix. The first device may arrange the data to be transmitted into a matrix Y having L rows and Q columns. Furthermore, the first device may split the data to be transmitted within h time units into N' sub-data, for example, based on the spatial relationship between the data in the first data, splitting the first data into N' sub-data. Each sub-data may be a matrix. For details, please refer to the description of the first data in the above-mentioned method 1, which will not be repeated for the sake of brevity.

[0182] The first device may divide N' sub-data into M first data in a round-robin manner. For example, y1 in the data to be transmitted is placed in the first first data Y1 as the first sub-data Y in the first first data Y1. (1,1) , place y2 in the second first data Y2 as the first sub-data Y in the second first data Y2 (2,1) ...will y M Placed at the Mth first data Y M As the Mth first data Y M The first sub-data Y in (M,1) , y M+1 Placed in the first first data Y1 as the second sub-data Y in the first first data Y1 (1,2) ...and so on, until all N' sub-data in the data to be transmitted are placed in the M first data.

[0183] According to this, the first sub-data Y in the first data Y1 in FIG4b (1,1) , the first sub-data Y in the second first data Y2 (2,1) ...the Mth first data Y M The first sub-data Y in (M,1) There is a correlation in the spatial dimension between them. That is, the first sub-data Y in the first data Y1 (1,1) , and the first sub-data Y in the second first data Y2 (2,1) The similarity between them is greater than or equal to the first similarity threshold; the first sub-data Y in the second first data Y2 (2,1) , and the first sub-data Y in the third first data Y3 (3,1) The similarity between them is greater than or equal to the first similarity threshold...the M-1th first data Y M-1 The first sub-data Y in (M-1,1) , and the Mth first data Y M The first sub-data Y in (M,1) The similarity between them is greater than or equal to the first similarity threshold. Similarly, in FIG4b, the second sub-data Y in the first data Y1 is(1,2) , and the second sub-data Y in the second first data Y2 (2,2) The similarity between them is greater than or equal to the first similarity threshold; the second sub-data Y in the second first data Y2 (2,2) , and the second sub-data Y in the third first data Y3 (3,2) The similarity between them is greater than or equal to the first similarity threshold...the M-1th first data Y M-1 The second sub-data Y in (M-1,2) , and the Mth first data Y M The second sub-data Y in (M,2) The similarity between the first data and the sub-data is greater than or equal to the first similarity threshold. Other sub-data in the first data may also have the above similarity, which will not be described in detail for the sake of brevity. However, this application is not limited to this. For example, any first data may include at least one sub-data that does not have the above similarity with sub-data in other first data.

[0184] The data Y in FIG4 b is data within h time units, such as data within one time unit, data within two time units, or data within more time units.

[0185] Regardless of the above-mentioned method 1 or method 2, the similarity between the sub-data at the same position in each first data (such as the first sub-data in the first data) is greater than or equal to the first similarity threshold as an example for explanation, but this application is not limited to this. For example, the similarity between the second sub-data in the first first data and the fourth sub-data in the second first data is greater than or equal to the first similarity threshold. Moreover, the similarity between sub-data may not be limited to adjacent first data, but may be any two first data in M ​​first data, such as the similarity between a sub-data in the p-th first data and a sub-data in the q-th first data is greater than or equal to the first similarity threshold, where p and q are positive integers and p is not equal to q. For example, the similarity between the sub-data in the first first data and the sub-data in the third first data may be greater than or equal to the first similarity threshold.

[0186] The above-mentioned method 1 and method 2 are both described by taking the similarity between one sub-data in each different first data as an example. The present application does not limit the number of similar sub-data in each first data. For example, the similarity between the 1st sub-data and the 2nd sub-data in the pth first data and the 1st sub-data in the qth first data is greater than or equal to the first similarity threshold, that is, the 1st sub-data and the 2nd sub-data in the pth first data and the 1st sub-data in the qth first data share the first dictionary matrix; for another example, the similarity between the 1st sub-data and the 3rd sub-data in the pth first data and the 1st sub-data and the 2nd sub-data in the qth first data is greater than or equal to the first similarity threshold, that is, the 1st sub-data and the 3rd sub-data in the pth first data and the 1st sub-data and the 2nd sub-data in the qth first data are greater than or equal to the first similarity threshold, that is, the 1st sub-data and the 3rd sub-data in the pth first data, and the 1st sub-data and the 2nd sub-data in the qth first data all share the first dictionary matrix.

[0187] In order to improve the compression rate, in the embodiment of the present application, one sub-data in each of the M first data shares the first dictionary matrix. In other words, the first dictionary matrix includes the features of the M sub-data corresponding to the M first data. The M sub-data can be the sub-data related in the spatial dimension in the above example. For example, the M sub-data can include the nth sub-data Y in the first first data Y1. (1,n) , the nth sub-data Y in the second first data Y2 (2,n) ...the Mth first data Y M The nth sub-data Y in (M,n) .

[0188] When M sub-data share the first dictionary matrix, for each sub-data in the M sub-data, the first device performs dictionary learning on the sub-data based on the first dictionary matrix to obtain a first sparse matrix corresponding to the sub-data, and the first sparse matrix performs sparse expression on the sub-data based on the first dictionary matrix.

[0189] It should be understood that each first data may include at least one sub-data that is correlated with sub-data in other first data. When each first data includes multiple sub-data that are respectively correlated with sub-data in other first data, M first data may correspond to multiple groups of sub-data, each group of sub-data includes M sub-data, and the M sub-data in each group of sub-data are correlated, that is, the M sub-data in each group of sub-data share a first dictionary matrix, such as the nth sub-data Y in the first first data Y1 mentioned above. (1,n) , the nth sub-data Y in the second first data Y2 (2,n) ...the Mth first data Y M The nth sub-data Y in (M,n)It can be a group of sub-data. For the sake of convenience, the embodiment of the present application is only exemplified by a group of sub-data corresponding to M first data. For some sub-data in the first data, the dictionary matrix may not be shared with the sub-data in other first data. Such sub-data can be decomposed into a second dictionary matrix and a second sparse matrix based on dictionary learning, and compressed and transmitted. It should be understood that this application only expresses M sub-data with correlation as a group of sub-data for the sake of convenience, rather than an actual data set, and does not perform actual group division on the sub-data of the first data.

[0190] The first dictionary matrix may be a preconfigured fixed dictionary matrix, for example, configured by the second device to the first device, or configured by the first device to the second device; or, the first dictionary matrix may be generated or updated by the first device, in which case the first device may send the first dictionary matrix to the second device.

[0191] Exemplarily, the first dictionary matrix can be obtained by decomposing the kth sub-data among the M sub-data after the first device performs dictionary learning. For example, based on dictionary learning, the first device decomposes the kth sub-data to obtain the first dictionary matrix and the first sparse matrix corresponding to the kth sub-data. In order to improve the convenience of data processing, the first device can decompose the first sub-data among the M sub-data to obtain the first dictionary matrix and the first sparse matrix corresponding to the first sub-data. The kth sub-data can be the kth first data Y k A sub-data in .

[0192] Continuing with the above example, for the M-1 sub-data except the k-th sub-data in the M sub-data, the first device can determine the first sparse matrix corresponding to each sub-data in the M-1 sub-data based on the first dictionary matrix, thereby obtaining M first sparse matrices.

[0193] It should be understood that a sparse matrix with good sparse performance has many elements with zero values, which achieves compression of the first data to a certain extent. In order to further improve the compression performance, the first device in the embodiment of the present application can compress some or all of the M first sparse matrices. For example, as shown in Figure 6, the first device can compress the first sparse matrix X mThe first position indication information is used to indicate the position of an element in the sparse matrix that has the ability to express the corresponding sub-data. The first device sets the value of the first element in the first sparse matrix that does not have the ability to express the corresponding sub-data to a first numerical value according to the indication of the first position indication information. The element sequence includes elements that have the ability to express the corresponding sub-data, that is, elements that have not been set to the first numerical value. The first numerical value can be zero or any other numerical value. As shown in FIG6 , the elements in the first row of the x1 column, the elements in the second row of the x2 column, and the elements in the second row of the x3 column in the first sparse matrix have the ability to express the corresponding sub-data. The element sequence includes the elements in the first row of the x1 column, the elements in the second row of the x2 column, and the elements in the second row of the x3 column in the sparse matrix that have the ability to express the sub-data. For example, the value of the element in the first row of the x1 column in the sparse matrix is ​​0.1, the value of the element in the second row of the x2 column is -0.5, the value of the element in the second row of the x3 column is 0.8… These elements have the ability to express the sub-data, and the element sequence can be expressed as [0.1, -0.5, 0.8…], and the sparse matrix X i Excluding the elements at the above positions, the absolute values ​​of the elements at other positions are small (including elements with a value of 0 or a value whose difference from 0 is less than a preset capability threshold), and the elements at these positions do not have the ability to express the sub-data. For example, the value of the element in the second row of the x1 column in the first sparse matrix is ​​-0.005, and the value of the element in the first row of the x2 column is 0.0002, etc., and the first position indication information indicates the positions of these elements with the ability to express the sub-data in the first sparse matrix, or in other words, the first position indication information indicates which elements at which positions in the first sparse matrix have the ability to express the corresponding sub-data. The first position indication information and the element sequence can be used as compressed data of the first sparse matrix.

[0194] The first position indication information may include a tree structure, a bitmap, a position index, etc., which is not limited in this application. Taking the first position indication information as a bitmap as an example, in combination with Figure 6, the bits in the bitmap correspond one-to-one to the elements in the first sparse matrix, and the bits in the bitmap are used to indicate whether the corresponding elements in the first sparse matrix have the ability to express the corresponding sub-data. For example, when the bit in the bitmap is 1, it indicates that the corresponding element in the first sparse matrix has the ability to express the corresponding sub-data; when the bit in the bitmap is 0, it indicates that the corresponding element in the first sparse matrix does not have the ability to express the corresponding sub-data. For another example, when the bit in the bitmap is 0, it indicates that the corresponding element in the first sparse matrix has the ability to express the corresponding sub-data; when the bit in the bitmap is 1, it indicates that the corresponding element in the first sparse matrix does not have the ability to express the corresponding sub-data.

[0195] The first position indication information may include at least one position index, each position index may indicate the position of an element in the first sparse matrix that has the ability to express corresponding sub-data in the first sparse matrix, for example, the position index indicates that an element that has the ability to express corresponding sub-data is in the 2nd row and 5th column in the first sparse matrix.

[0196] When the first position indication information includes a tree structure, a basic tree structure corresponding to the first sparse matrix can be constructed, and some or all of the subnodes in the basic tree structure correspond to the first sparse matrix one-to-one. Exemplarily, in the process of constructing the basic tree structure, the basic two-dimensional plane can be divided into multi-level intervals, and each interval is used as a subnode of the basic tree structure, and the number of the minimum intervals divided (i.e., the subnodes of the lowest level in the basic tree structure) is greater than or equal to the number of elements in the first sparse matrix, wherein the present application does not limit the size of the basic two-dimensional plane. For example, in the process of constructing the basic tree structure, two mutually perpendicular lines are used to divide it into four intervals, and the four intervals correspond to the four child nodes of the root node in the basic tree structure. For each interval in the four intervals, when the interval includes an element with the ability to express the corresponding sub-data, the value of the child node is the second value, and when the interval does not contain an element with the ability to express the corresponding sub-data, the value of the child node is the third value; further, the interval is divided into four sub-intervals by two mutually perpendicular lines, and the four sub-intervals correspond to the four child nodes of the next level of the child node corresponding to the interval in the basic tree structure. Similarly, when the sub-interval includes an element with the ability to express the corresponding sub-data, the value of the child node corresponding to the sub-interval in the basic tree structure is the second value, otherwise it is the third value, and so on until the minimum interval is reached to terminate the construction of the tree structure. The second value can be 1 and the third value can be 0, or the second value can be 0 and the third value can be 1, or the second value and the third value can be any two different values, which is not limited in this application.

[0197] For example, the size of the first sparse matrix of point cloud data is 3×16. In the process of constructing the basic tree structure, it is divided into four intervals by two mutually perpendicular lines. The four intervals correspond to the four child nodes of the root node in the tree structure, and then each interval is divided into four sub-intervals. The four sub-intervals under the interval correspond to the four child nodes of the next level of the child node corresponding to the interval in the basic tree structure. And so on. After 4 interval divisions (or 4 recursions), 16×16 minimum intervals are obtained, and the basic tree structure of the first sparse matrix is ​​constructed at the same time. Some child nodes in the basic tree structure can correspond one-to-one to elements in the first sparse matrix. For example, the 3×16 minimum intervals located in the upper left corner of the basic two-dimensional plane respectively correspond to child nodes in the basic tree structure indicating 3×16 elements of the first sparse matrix. Optionally, the first position indication information may be the basic tree structure, and the values ​​of the nodes corresponding to the intervals in the basic tree structure that do not include elements in the first sparse matrix are all third numerical values; or the tree structure in the first position indication information may include some child nodes in the basic tree structure, and the some child nodes are the child nodes corresponding to the intervals in the basic two-dimensional plane that include elements in the first sparse matrix.

[0198] In the above example, the divided minimum interval includes one element in the first sparse matrix. This application does not exclude the situation where the minimum interval includes multiple elements in the first sparse matrix or multiple minimum intervals correspond to one element in the first sparse matrix. When the minimum interval includes one element, the position indicated by the first position indication information is more accurate, and thus the data constructed by the second communication device based on the first position indication information is more accurate. When the minimum interval includes multiple elements, the transmission resources occupied by the first position indication information are smaller, saving resource overhead.

[0199] The above exemplary description of the first position indication information can be applied to the description of the position indication information in the following related embodiments. The implementation methods are the same or similar and will not be repeated for the sake of brevity.

[0200] When the first device is implemented as a chip or a chip system, the above S220 can be replaced by the first device outputting the compressed data of M first sparse matrices and sending the compressed data of the M first sparse matrices through the terminal device or network device deployed by the first device.

[0201] As mentioned above, the M first data may correspond to multiple groups of sub-data, and each group of sub-data is expressed by the corresponding M first sparse matrices based on the same first dictionary matrix. The first device may encapsulate the compressed data of the first sparse matrices corresponding to the multiple groups of sub-data together and send them, or the first device may encapsulate the compressed data of the M first sparse matrices corresponding to one group of sub-data together and send them, or the first device may encapsulate the compressed data of each first sparse matrix separately and send them.

[0202] The first device may also send the first dictionary matrix to the second device. Optionally, when the first device is implemented as a chip or a chip system, the first device outputs the first dictionary matrix, and the terminal device or network device deployed by the first device sends the first dictionary matrix.

[0203] After the second device receives the compressed data of the M first sparse matrices sent by the first device, it can decompress the compressed data of the M first sparse matrices to obtain decompressed data. It should be noted that decompressing the M first sparse matrices can also be expressed as restoring the M first sparse matrices, or constructing the M first sparse matrices, and for the convenience of expression, the meaning expressed after the word order is changed remains the same, for example, recovering data of the M first sparse matrices, or constructing data of the M first sparse matrices.

[0204] The second device can construct M first sparse matrices based on the first dictionary matrix and the compressed data.

[0205] In some embodiments, the first device may perform compression processing on the compressed data of the first dictionary matrix and / or the M first sparse matrices, where the compression processing includes but is not limited to quantization (such as scalar quantization or vector quantization) and / or entropy coding. Accordingly, the second device receives the compressed data of the first dictionary matrix and / or the M first sparse matrices after the compression processing and needs to perform corresponding decompression processing on them.

[0206] Therefore, in an embodiment of the present application, dictionary learning is performed on the M sub-data corresponding to the M first data based on a first dictionary matrix to obtain a sparse expression of each sub-data, thereby achieving effective and reliable data compression transmission in transmission scenarios with large data volumes.

[0207] To further improve the compression ratio, the embodiment of the present application can jointly compress the M first sparse matrices, as exemplified by the following two implementations.

[0208] Implementation method 1: Jointly compress the M first sparse matrices based on a low-rank approximation method.

[0209] In the first implementation, the first device may determine the first matrix based on the M first sparse matrices and perform low-rank approximation on the first matrix to obtain first compressed data of the M first sparse matrices. The method for obtaining the M first sparse matrices has been described in the previous example and will not be repeated here.

[0210] In the first example of the above implementation mode 1, the first device may combine M first sparse matrices to obtain a first matrix. For example, the M first sparse matrices (X1 to X M ) Get the first matrix by row stacking Alternatively, the M first sparse matrices (X1 to X M ) The first matrix X=[X1,X2…X M ].

[0211] In the second example of the above-mentioned implementation method one, the first device may first perform data compression on at least one first sparse matrix among the M first sparse matrices, and combine the M first sparse matrices after data compression to obtain a first matrix.

[0212] Continuing with the second example above, take the mth first sparse matrix X m For example, the first sparse matrix after data compression Among them, B m is the first position indication information corresponding to the mth first sparse matrix, that is, according to the first position indication information B m , the value of the first element in the mth first sparse matrix is ​​set to the first value to achieve data compression. The first sparse matrix after data compression can be understood as an element sequence composed of elements in the first sparse matrix that have the ability to express the corresponding sub-data. Optionally, the first sparse matrix X m The data compression may further include obtaining the compressed first sparse matrix After that, for the first sparse matrix Quantization (including scalar quantization or vector quantization) is performed. The first position indication information and the first element have been described in the above example and are not repeated here. It should be further explained that the first device can perform the above data compression on one or more first sparse matrices among the M first sparse matrices.

[0213] When the first device performs the above data compression on multiple first sparse matrices in the M first sparse matrices, the first position indication information (ie, B1=B2=...=B M) performs data compression on the multiple first sparse matrices, that is, the first device sets the value of the first element in each first sparse matrix to the first numerical value according to the same first position indication information. In this case, the first position indication information can be determined based on one sparse matrix among the M first sparse matrices, and the one sparse matrix can be the first sparse matrix corresponding to the first sub-data in the M first sparse matrices, or can be any one of the M first sparse matrices. For example, the first device determines the first position indication information and element sequence corresponding to the one first sparse matrix. Furthermore, when the first device performs data compression on the M-1 sparse matrices other than the one first sparse matrix, it can use the first position indication information to screen the elements in each first sparse matrix that have the ability to express the corresponding sub-data, that is, to set the value of the element that does not have the ability to express the corresponding sub-data to the first numerical value, and obtain the element sequence of each first sparse matrix.

[0214] When the first device performs the above-mentioned data compression on multiple first sparse matrices among M first sparse matrices, it can first determine the first position indication information corresponding to each first sparse matrix among the multiple first sparse matrices, and then, for each first sparse matrix among the multiple first sparse matrices, perform data compression on the first sparse matrix using the first position indication information corresponding to the first sparse matrix.

[0215] Optionally, the first device may determine the first position indication information based on the comparison result between the capability threshold and each element in the first sparse matrix. For example, bit 1 in the first position indication information is used to indicate the position of the element whose absolute value is greater than the capability threshold, that is, the element with the ability to express the corresponding sub-data. Bit 0 in the first position indication information is used to indicate the position of the element whose absolute value is less than the capability threshold, that is, the element without the ability to express the corresponding sub-data. For elements whose absolute value is equal to the capability threshold, it can be determined as an element with the ability to express the corresponding sub-data or an element without the ability to express the corresponding sub-data. This application does not limit this. The first device may also determine the first position indication information based on the first W elements with the largest absolute value in the first sparse matrix. For example, the first device determines the first W elements with the largest absolute value in the first sparse matrix as elements with the ability to express the corresponding sub-data, and determines the elements in the first sparse matrix other than the W elements as elements without the ability to express the corresponding sub-data, and then indicates the position of the element with the ability to express the corresponding sub-data by bit 1 in the first position indication information, and indicates the position of the element without the ability to express the corresponding sub-data by bit 0. It should be understood that the present application does not limit the bit value for indicating the position of an element that has or does not have the ability to express the corresponding sub-data. For example, the position of an element that does not have the ability to express the corresponding sub-data can be indicated by bit 1, and the position of an element that has the ability to express the corresponding sub-data can be indicated by bit 0.

[0216] Continuing with the second example above, the first device may combine the first sparse matrix after data compression, such as the element sequence of the first sparse matrix obtained after element screening based on expression ability, to obtain the first matrix. For example, the first sparse matrix after M data compression ( to ) Get the first matrix by row stacking Alternatively, the M first sparse matrices after data compression ( to ) The first matrix is ​​obtained by column stacking

[0217] In both the first and second examples of the above-mentioned implementation method 1, the first matrix is ​​a high-dimensional matrix, and directly transmitting the first matrix will result in a large communication overhead. Therefore, the first device can perform singular value decomposition on the first matrix to obtain K eigenvalues ​​(or singular values) and eigenvectors (or singular value vectors) corresponding to the K eigenvalues. The K eigenvalues ​​and the eigenvectors corresponding to the K eigenvalues ​​can express the above-mentioned first matrix. For example, the first matrix Convert to USV after singular value decomposition T, where U is the left singular matrix, S is the singular value matrix, V T is a right singular matrix, the left singular matrix may include K columns, the singular value matrix may include K singular values, and the right singular matrix may include K rows. Further, the first device uses K0 eigenvalues ​​of the K eigenvalues ​​and the eigenvectors corresponding to the K0 eigenvalues ​​(including the K0 column vectors of the left singular matrix and the K0 row vectors of the right singular matrix) as the first compressed data of the M first sparse matrices. The low rank approximation process can be expressed as in, represents the elements of all rows from columns 1 to K0 in the left singular matrix, represents the first K0 eigenvalues ​​of the singular value matrix, Represents the elements of all columns from rows 1 to K0 in the right singular matrix. It can be understood that the more forward singular values ​​in the singular value matrix have a stronger ability to express the first matrix. Correspondingly, the more forward column vectors in the left singular matrix have a stronger ability to express the first matrix, and the more forward row vectors in the right singular matrix have a stronger ability to express the first matrix. Therefore, by expressing the first matrix through the first K0 eigenvalues ​​and the eigenvectors corresponding to the K0 eigenvalues, a low-rank approximation of the first matrix is ​​achieved, thereby achieving a low-rank approximation of the M first data.

[0218] As previously mentioned, the M first data may correspond to multiple groups of sub-data, each group of sub-data being expressed by corresponding M first sparse matrices based on the same first dictionary matrix. The M first sparse matrices expressing the corresponding sub-data based on the same first dictionary matrix form one first compressed data. The first device may perform protocol encapsulation on the first compressed data corresponding to the multiple groups of sub-data before sending them together, or the first device may perform protocol encapsulation on each first compressed data separately before sending them.

[0219] In some embodiments, the first device transmits compressed data of M first data in an incremental manner. In this case, the above-mentioned first compressed data may be the compressed data of the M first data transmitted during the initial transmission process. During the incremental transmission process, the first device may send second compressed data of M first sparse matrices to the second device. The second compressed data may include K1 eigenvalues ​​other than K0 eigenvalues ​​among the K eigenvalues ​​obtained by the above-mentioned singular value decomposition, and the eigenvectors corresponding to the K1 eigenvalues ​​respectively. Among them, the K0 eigenvalues ​​may be the first K0 eigenvalues ​​among the K eigenvalues, and the K1 eigenvalues ​​may be the first K1 eigenvalues ​​other than the first K0 eigenvalues ​​among the K eigenvalues, such as the K0+1th eigenvalue to the K0+K1th eigenvalue. Accordingly, the eigenvectors corresponding to the K1 eigenvalues ​​may include the K0+1th to K0+K1th column vectors of the left singular matrix and the K0+1th to K0+K1th row vectors of the right singular matrix.

[0220] It is understood that the first device may perform incremental transmission on one or more groups of sub-data among the multiple groups of sub-data. When performing incremental transmission on at least two groups of sub-data, the first device may perform protocol encapsulation on the second compressed data corresponding to the multiple groups of sub-data before sending the data, or the first device may perform protocol encapsulation on each second compressed data separately before sending the data.

[0221] It should also be understood that the first device can perform one or more incremental transmissions. For example, during the second incremental transmission, the first device can send third compressed data to the second device. The third compressed data can include K2 eigenvalues ​​in addition to the above-mentioned K0+K1 eigenvalues ​​among the K eigenvalues, and the eigenvectors corresponding to the K2 eigenvalues ​​respectively.

[0222] When the first device is implemented as a chip or a chip system, the first device can output the above-mentioned second compressed data and send the second compressed data through a transceiver via a terminal device or a network device where the first device is deployed.

[0223] Implementation method 2: Jointly compress the M first sparse matrices based on the residual method.

[0224] In the above-mentioned second implementation method, the first device can sequentially send the compressed data of each first sparse matrix among the M first sparse matrices, and each time the compressed data of the first sparse matrix is ​​sent, the compressed data of the first sparse matrix currently being sent can be determined based on the compressed data of the first sparse matrix previously sent. For ease of explanation, taking the jth first sparse matrix among the M first sparse matrices currently being sent as an example, the first device determines the first residual information based on the information of the jth first sparse matrix and the i-th first sparse matrix, and sends the first residual information as the compressed data of the j-th first sparse matrix to the second device. Wherein, i is less than j, and both i and j are positive integers. The information of the i-th first sparse matrix can be obtained by decompressing the compressed data of the i-th first sparse matrix.

[0225] In some embodiments, any one of the M first sparse matrices is used as a basic sparse matrix, and the first device may use the first position indication information and element sequence of the basic sparse matrix as compressed data of the basic sparse matrix. Moreover, for each of the M-1 first sparse matrices other than the basic sparse matrix, the first device compresses and transmits each first sparse matrix according to the compression transmission method of the j-th first sparse matrix. The basic sparse matrix may be the first sparse matrix compressed and transmitted for the first time among the M first sparse matrices.

[0226] In other embodiments, M first sparse matrices are divided into multiple sparse matrix groups, and any first sparse matrix in each sparse matrix group is used as the basic sparse matrix of the group. The first device can use the first position indication information and element sequence of the basic sparse matrix as the compressed data of the basic sparse matrix, and for the first sparse matrices in the group other than the basic sparse matrix, the first device compresses and transmits each first sparse matrix in the group according to the compression transmission method of the jth first sparse matrix. The basic sparse matrix can be the first sparse matrix in the sparse matrix group that is compressed and transmitted for the first time.

[0227] Exemplarily, the first device can determine the residual matrix based on the information of the jth first sparse matrix and the i-th first sparse matrix, and then determine the first residual information based on the residual matrix. The first residual information may include a first residual element sequence, and the first residual element sequence includes residual elements in the residual matrix whose absolute values ​​are greater than or equal to the first residual threshold, and the elements whose absolute values ​​are greater than or equal to the first residual threshold have the ability to express the residual between the information of the jth first sparse matrix and the i-th first sparse matrix. The second device can construct the j-th first sparse matrix based on the first residual element sequence and the information of the i-th first sparse matrix. Optionally, when calculating the residual matrix between the j-th first sparse matrix and the matrix obtained by decompressing the compressed data of the i-th first sparse matrix, the values ​​of the elements at each corresponding position in the two matrices can be differenced to obtain the residual matrix.

[0228] Continuing with the above example, the first residual information may also include second position indication information, which is used to indicate the position of the residual element whose absolute value is greater than the first residual threshold in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, that is, the position of the element in the residual matrix that has the ability to express the residual between the information of the j-th first sparse matrix and the i-th first sparse matrix. It should be noted that the second position indication information may also indicate the position of the residual element whose absolute value is greater than the first residual threshold in the residual matrix, that is, the position of the element in the residual matrix that does not have the ability to express the residual between the information of the j-th first sparse matrix and the i-th first sparse matrix. The element in the residual matrix whose absolute value is equal to the first residual threshold can be considered to have the ability to express the residual or not to have the ability to increase the residual, and this application does not limit this. The bits of the two residual elements indicating the ability to express the residual between the information of the j-th first sparse matrix and the i-th first sparse matrix may be different.

[0229] Exemplarily, the first device may obtain a residual matrix based on the difference between the jth first sparse matrix and the restored ith first sparse matrix; or the first device may obtain a residual matrix based on the jth first sparse matrix after data compression (such as And the restored i-th first sparse matrix after data compression (like in, represents the i-th first sparse matrix after data compression, Represents the difference of the restored i-th first sparse matrix) to obtain the residual matrix.

[0230] In order to further improve the compression rate of the M first sparse matrices, in some embodiments of the above-mentioned implementation method two, the first device can determine the similarity between the jth first sparse matrix and the i-th first sparse matrix. When the similarity between the jth first sparse matrix and the i-th first sparse matrix is ​​less than the second similarity threshold, the first device sends the first residual information to the second device. When the similarity between the jth first sparse matrix and the i-th sparse matrix is ​​greater than or equal to the second similarity threshold, the first device does not send the compressed data of the j-th first sparse matrix. When the similarity between the j-th first sparse matrix and the i-th sparse matrix is ​​equal to the second similarity threshold, the compressed data of the j-th first sparse matrix may be sent or not. This application does not limit this.

[0231] When the first device does not send the compressed data of the j-th first sparse matrix, the second device uses the compressed data of the ith first sparse matrix as the compressed data of the j-th first sparse matrix, or in other words, the second device uses the constructed ith first sparse matrix as the j-th first sparse matrix.

[0232] The similarity between the jth first sparse matrix and the ith first sparse matrix can be determined based on the residual matrix between the jth first sparse matrix and the ith first sparse matrix. For example, based on the modulus value of the residual matrix (such as || X j -X i || or || X j ⊙B j -X i ⊙B i ||) determines the similarity between the jth first sparse matrix and the ith first sparse matrix, the larger the modulus value of the residual matrix, the lower the similarity between the jth first sparse matrix and the ith first sparse matrix; for another example, the similarity between the jth first sparse matrix and the ith first sparse matrix can be determined based on the ratio of the modulus value of the residual matrix to the first modulus value, wherein the first modulus value can be the modulus value of the ith first sparse matrix (such as ||X i || or | | X i ⊙B i ||), or the modulus of the jth first sparse matrix (such as ||X j || or || X j ⊙Bj ||), the larger the ratio of the modulus value of the residual matrix to the first modulus value, the lower the similarity between the j-th first sparse matrix and the i-th first sparse matrix.

[0233] It should be understood that before the first device sends the compressed data of the jth first sparse matrix to the second device, it can send the compressed data of at least one first sparse matrix to the second device. Then the i-th first sparse matrix can be any one of the at least one first sparse matrix. In order to facilitate compressed transmission and data recovery, the i-th first sparse matrix can be the first sparse matrix transmitted in the most recent transmission of the at least one first sparse matrix. When the i-th first sparse matrix is ​​the first first sparse matrix, in this implementation method 2, the compressed data of the i-th first sparse matrix includes the first position indication information and element sequence of the i-th first sparse matrix; when the i-th first sparse matrix is ​​not the first first sparse matrix, in this implementation method 2, the compressed data of the i-th first sparse matrix includes the second position indication information and the first residual element sequence of the i-th first sparse matrix.

[0234] As mentioned above, the M first data may correspond to multiple groups of sub-data, and each group of sub-data is expressed by the corresponding M first sparse matrices based on the same first dictionary matrix. The compressed information of the first sparse matrices corresponding to the sub-data in each group can be packaged together and sent after the protocol is completed. For example, the nth sub-data Y in the first data Y1 (1,n , the nth sub-data Y in the second first data Y2 (2,n) ...the Mth first data Y M The nth sub-data Y in (M,n) They can be sent together after being packaged with a protocol, or the compressed data of the first sparse matrix corresponding to the sub-data in each group can be sent after being packaged with a protocol separately.

[0235] When the first device is implemented as a chip or a chip system, the first device can output the compressed data of the above-mentioned M first sparse matrices, and send the second compressed data through a transceiver via a terminal device or network device where the first device is deployed.

[0236] In some embodiments, the first device transmits compressed data of M first data in an incremental manner. In this case, the first residual information sent in the above embodiment may be the compressed data of the jth first sparse matrix transmitted during the initial transmission process. During the incremental transmission process, the first device may send the second residual information of the jth first sparse matrix to the second device. The second residual information is also determined based on the information of the jth first sparse matrix and the i-th first sparse matrix. The difference from the first residual information is that the third position indication information in the second residual information is used to indicate the residual elements in the residual matrix between the jth first sparse matrix and the information of the i-th first sparse matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold, and the second residual element sequence in the second residual information includes the residual elements in the residual matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold. The second residual information transmitted in the incremental process can supplement the first residual information to enrich the residual information, so that the second device can construct the jth first sparse matrix based on the first residual information and the second residual information with higher accuracy.

[0237] Furthermore, the first device can transmit third residual information when it transmits incrementally again during the incremental transmission process. The third residual information is also determined based on the information of the j-th first sparse matrix and the i-th first sparse matrix. The fourth position indication information in the third residual information is used to indicate the residual elements in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix that are less than or equal to the second residual threshold and greater than or equal to the third residual threshold, so as to supplement the first residual information and the second residual information. Of course, the first device can also transmit more residual information, and this application does not limit the number of incremental transmissions during the incremental transmission process.

[0238] It should be noted that the first device can incrementally transmit the first sparse matrix corresponding to one or more sub-data among the M sub-data, and / or the first device can incrementally transmit one or more groups of sub-data among the multiple groups of sub-data corresponding to the M first data.

[0239] In the above-mentioned implementation manners 1 and 2, the first device may perform compression processing on the compressed data of the M first sparse matrices, such as performing at least one of scalar quantization, vector quantization, and entropy coding, and send the compressed first residual information to the second device, and the second device performs corresponding decompression processing on the compressed data. The compressed data of the M first sparse matrices includes the first compressed data and / or the second compressed data in implementation manner 1, or includes the first residual information and / or the second residual information in implementation manner 2.

[0240] The various thresholds in the above-mentioned implementation method one and implementation method two, such as the capability threshold, similarity threshold, and residual threshold (including the first residual threshold to the fourth residual threshold) can all be preset or pre-configured, where the preset can be, for example, protocol-defined or pre-stored in the device, and the pre-configuration can be, for example, pre-configured by the network device to the terminal device.

[0241] FIG7 is a schematic diagram of an interactive process of another data compression transmission method provided by an embodiment of the present application. As shown in FIG7 , the method 300 includes some or all of the following processes from S310 to S390:

[0242] S310: The first device sends a compressed transmission request to the second device, where the compressed transmission request carries the data type of the data to be transmitted. Correspondingly, the second device receives the compressed transmission request sent by the first device.

[0243] S320: The second device sends first configuration information to the first device, where the first configuration information is used to indicate first time-frequency resources of M first data. Correspondingly, the first device receives the first configuration information and / or first indication information sent by the second device.

[0244] S330: The first device transmits, on the first time-frequency resource, compressed data of the M first sparse matrices and first indication information, where the first indication information is used to indicate a first compression parameter. Correspondingly, the second device receives, on the first time-frequency resource, the compressed data of the M first sparse matrices and the first indication information transmitted by the first device.

[0245] S340, the second device constructs M first data according to the compressed data of the M first sparse matrices and the first indication information.

[0246] S350: The second device sends second configuration information to the first device, where the second configuration information is used to indicate the second time-frequency resource of the compressed data of the M first sparse matrices. Correspondingly, the first device receives the second configuration information sent by the second device.

[0247] S360: The first device sends the compressed data of the M first sparse matrices and the second indication information in the second time-frequency resource. Correspondingly, the second device receives the compressed data of the M first sparse matrices and the second indication information sent by the first device in the second time-frequency resource.

[0248] S370: The second device updates the constructed M first data.

[0249] It should be noted that the initial transmission process is the basis for the incremental transmission process. In the absence of incremental transmission, the initial transmission process is a complete compressed transmission process. When the above method 300 includes S350 to S370, S310 to S340 are the initial transmission process, and S350 to S370 are the incremental transmission process. When the above method 300 does not include S350 and S370, S310 to S340 are the compressed transmission process.

[0250] When S310 to S340 are implemented as the initial transmission process, the first time-frequency resources are the time-frequency resources occupied by the initial transmission process, and the M first data are the compressed data of the initial transmission process transmitted on the first time-frequency resources, such as the first compressed data in the aforementioned implementation method one and the first residual information in the aforementioned implementation method two, and may also include the compressed data of the first dictionary matrix, and the first compression parameter is the compression parameter of the initial transmission process; the second time-frequency resources are the time-frequency resources occupied by the incremental transmission process, and the compressed data of the M first sparse matrices are the compressed data of the incremental transmission process transmitted on the second time-frequency resources, such as the second compressed data in the aforementioned implementation method one and the second residual information in the aforementioned implementation method two.

[0251] Optionally, the data type includes point cloud data or AI model data.

[0252] Optionally, the compressed transmission request may be a status report (SR) or a buffer status report (BSR). For example, the first device may send a compressed transmission request via an SR to indicate the data type, or the first communication device may send a compressed transmission request via a BSR to report the size of the data to be transmitted (e.g., the M first sparse matrices and / or the first dictionary matrix).

[0253] In the above S310, the first device sends a compressed transmission request to the second device to request compressed transmission of the M first sparse matrices. The compressed transmission request may carry the data type and / or size of the data to be transmitted.

[0254] The second device may determine the first time-frequency resource for transmitting the compressed data of the M first sparse matrices based on the data type of the data to be transmitted and / or the size of the data to be transmitted. Optionally, if the first data is large and the compressed data of the M first sparse matrices needs to be transmitted in an incremental transmission manner, the second device may first determine the first time-frequency resource for the compressed data of the M first sparse matrices during the initial transmission process.

[0255] Before the above S330, this embodiment may also determine the compressed data of the M first sparse matrices based on the method of any of the above embodiments, which will not be described again for the sake of brevity.

[0256] Based on any of the foregoing embodiments, the first device may determine a first compression parameter based on the first time-frequency resource, and then perform data compression based on the first compression parameter; wherein the first compression parameter includes at least one of the following:

[0257] 1) A capability threshold, which is used to determine whether an element in the first sparse matrix has the capability to express a sub-data in the corresponding first data;

[0258] 2) a first residual threshold, the first residual threshold being used to determine a first residual element sequence, the first residual element sequence being used to express a residual matrix between information of the j-th first sparse matrix and information of the i-th first sparse matrix, where i is less than j, and both i and j are positive integers;

[0259] 3) Compression processing parameters, including at least one of quantization precision, quantization codebook, and encoding mode. The compression processing parameters may indicate the encoding mode, quantization precision, quantization codebook, etc., of the first dictionary matrix, the first compressed data, and the first residual information, respectively. Optionally, the encoding mode may include scalar quantization, vector quantization, entropy coding, etc.

[0260] Optionally, in a scalar quantization mode, the compression processing parameters may further include a quantization range of the scalar quantization; in a vector quantization mode, the compression processing parameters may further include a quantization codebook corresponding to the vector quantization.

[0261] In the above-mentioned S330, the first indication information can be sent together with the first compressed data on the first time-frequency resource, but this application is not limited to this. For example, the first indication information can be sent separately, and the order in which the first device sends the first indication information and sends the first compressed data is not limited. Alternatively, the first indication information can be sent by the second device to the first device to indicate data compression for the first device. Alternatively, the content indicated by the first indication information can be agreed upon by the protocol.

[0262] The first indication information is used to indicate at least one of the following:

[0263] 1) Number of features of low-rank approximation K0. For example, based on the data division according to the spatial dimension in the above-mentioned method 2, K0 can be equal to 4 or 8, but this application is not limited to this. The value of K0 can be adaptively adjusted according to the application scenario or communication service.

[0264] 2) The number M of the first data. The number M of the first data can be adaptively adjusted according to the application scenario or communication service. Optionally, the number M of the first data can be associated with K0. For example, based on the data division according to the time dimension in the aforementioned method 2, when K0 is equal to 4, M can be equal to 2, and when K0 is equal to 8, M can be equal to 4; based on the data division according to the time dimension in the aforementioned method 1, when the data to be transmitted is AI data, the number M of the first data can be equal to 10. When the data to be transmitted is point cloud data, the number M of the first data is associated with the scanning frequency during the acquisition of the point cloud data. For example, the rotation scanning frequency of the lidar is usually 20Hz, and the number of the first data can be equal to 5.

[0265] 3) Capacity threshold.

[0266] 4) First residual threshold.

[0267] 5) The proportion of elements in one of the M first sparse matrices that are capable of expressing a sub-data in the corresponding first data.

[0268] 6) Data loss of the first compressed data relative to the M first sparse matrices.

[0269] 7) Parameters of compression processing.

[0270] 8) Whether to send the first residual information.

[0271] Among them, 1) the number of low-rank approximation features K0, 2) the capability threshold, 4) the first residual threshold, 5) the proportion of elements in the first sparse matrix that have the ability to express a sub-data in the corresponding first data, and 6) the data loss of the first compressed data can all be used by the second device to determine whether it is necessary to compress and transmit M first data (or M first sparse matrices) in an incremental transmission manner; 4) the first residual threshold, 7) the parameters of the compression processing can be used to instruct the first device to perform data compression and the second device to decompress the first compressed data; 1) the number of low-rank approximation features K0, 2) the number of first data M, 3) the capability threshold, 4) the first residual threshold, 8) whether to send the first residual information can all be used by the first device to perform data compression and the second device to perform data construction.

[0272] For example, in the aforementioned implementation manner 1, the first indication information may include one or more fields as shown in Table 1:

[0273] Table 1

[0274] Among them, the compression status is used by the second device to determine whether it is necessary to compress and transmit the first data in an incremental manner. The compression status field may include: 1) the number of features K0 of the low-rank approximation, 2) the capability threshold, 4) the first residual threshold, 5) the proportion of elements in the first sparse matrix that have the ability to express a sub-data in the corresponding first data, and 6) at least one of the data loss of the first compressed data.

[0275] Optionally, in the first indication information shown in Table 1, the compression indication field of the first dictionary matrix can indicate the compression accuracy of the first dictionary matrix. For example, a group of sub-data in the M first data share the first dictionary matrix. Therefore, it can be indicated that the first dictionary matrix is ​​quantized with a high-precision quantizer (e.g., 32 bits). The compression indication of the K0 eigenvalues ​​is similar to the compression indication of the first dictionary matrix and is not further described. And / or, the compression indication field of the first dictionary matrix can indicate the compression method of the first dictionary matrix, such as indicating a compression method such as quantization or entropy coding.

[0276] Optionally, in the first indication information shown in Table 1, the compression indication field of the feature vector corresponding to the K0 eigenvalues ​​may indicate a compression method for the feature vector, such as quantization, entropy coding, dictionary learning, or other compression methods. Furthermore, at least one of quantization accuracy, quantization codebook, and encoding method may be indicated.

[0277] Optionally, the data partitioning mode field may indicate data partitioning according to the time dimension in the above-mentioned method 1, or indicate data partitioning according to the space dimension in the above-mentioned method 2.

[0278] Optionally, the data partition dimension field may include an indication of the quantity M of the first data, and / or an indication of the quantity of sub-data in the first data, etc.

[0279] In some embodiments, the compression indication field of the first dictionary matrix in Table 1, the compression indication field of the K0 eigenvalues, and the compression indication field of the eigenvectors corresponding to the K0 eigenvalues ​​are all mandatory fields, and the data partitioning mode field, the data partitioning dimension field, and the compression status field are all optional fields.

[0280] For example, in the aforementioned implementation manner 2, the first indication information may include one or more fields as shown in Table 2:

[0281] Table 2

[0282] Among them, the data partition mode field, the data partition dimension field, and the compression status field are similar to those in Table 1 and are not repeated here.

[0283] Optionally, any of the above-mentioned compression indication fields, such as the compression indication field of the first dictionary matrix, the compression indication field of the first position indication information, the compression indication field of the second position indication information, and the compression indication field of the first residual element sequence, can indicate the compression method and / or compression accuracy of the corresponding data. For example, the compression indication field of the first dictionary matrix indicates that the first dictionary matrix is ​​quantized with 32-bit high precision; for another example, the compression indication field of the first position indication information indicates that the first position indication information is quantized according to (Lempel–Ziv 77, LZ77) or (Lempel–Ziv–Markov chain algorithm, LZMA), the compression indication field of the second position indication information indicates that the second position indication information is compressed in the manner of LZ77 or LZMA, wherein LZ77 is a dictionary-based, "sliding window" lossless compression algorithm invented by Lempel-Ziv in 1977, and LZMA is a compression algorithm improved based on LZ77, which has the characteristics of high compression ratio, high decompression speed, low memory consumption, etc.; for another example, the compression indication field of the element sequence indicates the quantization precision and / or quantization range when scalar quantization is performed on the element sequence, or the compression indication field of the element sequence indicates the quantization precision and / or quantization codebook when vector quantization is performed on the element sequence, or the compression indication field of the element sequence indicates the encoding method when entropy coding is performed on the element sequence, and the first residual element sequence is similar and will not be repeated.

[0284] The above compression method is only an example and not a limitation. The compression method of the first position indication information may also include any other compression algorithm, which will not be listed one by one.

[0285] The first position indication information and element sequence are determined based on the basic sparse matrix, and the second position indication information and the first residual element sequence are determined based on information of the j-th first sparse matrix and the i-th first sparse matrix.

[0286] Optionally, whether to send the first residual information field can be indicated by 1 bit whether to send the first residual information corresponding to the j-th first sparse matrix. For example, when the whether to send the first residual information field is 0, it indicates that the first residual information corresponding to the j-th first sparse matrix is ​​not sent. When the whether to send the first residual information field is 1, it indicates that the first residual information corresponding to the i-th first sparse matrix is ​​sent. Of course, this application does not limit the value of whether to send the first residual information field.

[0287] In scenarios where latency requirements are high, in order to reduce the impact of latency on communication services, the first indication information may indicate the use of the above-mentioned residual-based method to jointly compress the M first sparse matrices; in scenarios where latency requirements are low, the first indication information may indicate the use of the above-mentioned low-rank approximation-based method to jointly compress the M first sparse matrices.

[0288] In the above S340, the second device decompresses the compressed data of the M first sparse matrices according to the instructions of the first indication information, and then constructs M sub-data corresponding to the M first sparse matrices, and constructs data for other sub-data in the M first data based on a similar method to obtain M first data.

[0289] In the above S350, the second device can determine whether incremental transmission is required based on the construction quality of the M first data constructed in the above S340, and / or the compression state of the compressed data of the M first sparse matrices, and then send the second configuration information to the first device when incremental transmission is required. For example, the second device generates and sends the second configuration information when the construction quality of the M first data (or M sub-data) is poor. For another example, the second device sends the second configuration information when the compression state field indicates that the data loss of the first compressed data relative to the M first sparse matrices exceeds a threshold. The compression state can be indicated by the above compression state field.

[0290] Based on any of the foregoing embodiments, the first device can calculate the total number of transmittable bits based on the second time-frequency resources, and then determine the second compression parameter, and then perform data compression based on the second compression parameter; wherein the second compression parameter includes a second residual threshold, so that the first device determines the residual elements to be sent in the incremental transmission process in combination with the second residual threshold, and accordingly, the second device determines the residual elements sent by the first device in the incremental process in combination with the second residual threshold.

[0291] In the above S360, the second indication information can be sent on the second time-frequency resource together with the compressed data of the M first sparse matrices (such as the first compressed data or the second residual information). However, this application does not limit this. For example, the second indication information can be sent separately, and the execution order of the first device sending the second indication information and sending the compressed data is not limited. Alternatively, the second indication information can be sent by the second device to the first device to indicate the data compression of the first device during the incremental transmission process, or the content indicated by the second indication information can be agreed upon by the protocol.

[0292] Among them, the second indication information can be used to indicate incremental transmission based on low-rank approximation. In this case, the second indication information can indicate the number of features K1 of low-rank approximation; or, the second indication information can be used to indicate incremental transmission based on residual. In this case, the second indication information can indicate a second residual threshold, such as indicating a second residual threshold.

[0293] For example, in the aforementioned implementation manner 1, the second indication information may include one or more fields as shown in Table 3:

[0294] Table 3

[0295] Optionally, the incremental transmission mode indication field may indicate whether incremental transmission is performed based on a low-rank approximation method or a residual method.

[0296] Optionally, the position indication field may indicate the first data that needs to be transmitted incrementally among the M first data, or the sub-data in the first data that needs to be transmitted incrementally.

[0297] For the compression indication field of the K1 eigenvalues, refer to the description of the compression indication field of the K0 eigenvalues ​​in the aforementioned example; for the compression indication field of the eigenvector corresponding to the K1 eigenvalues, refer to the description of the compression indication field of the eigenvector corresponding to the K0 eigenvalue in the aforementioned example. For the sake of brevity, no further details are given.

[0298] For example, in the aforementioned implementation manner 2, the second indication information may include one or more fields as shown in Table 4:

[0299] Table 4

[0300] Among them, the incremental transmission mode indication field and the position indication field can refer to the description in the above Table 3. The compression indication field of the third position indication information can refer to the description of the compression indication field of the second position indication information in the aforementioned example; the compression indication field of the second residual element sequence can refer to the description of the compression indication field of the first residual element sequence in the aforementioned example.

[0301] In the above S370, the second device can update and construct the M first data based on the supplement of the constructed M sub-data by the second compressed data or the second residual information of the M first sparse matrices.

[0302] It should be understood that the embodiment shown in Figure 7 is only illustrated by taking one incremental transmission as an example. The embodiment of the present application does not limit the number of incremental transmissions. The first device and the second device can also perform incremental transmission based on the above S350 to S370 or similar methods.

[0303] Figure 8 is a schematic block diagram of a communication device provided in an embodiment of the present application. The communication device 400 can be a terminal or a network device, or a device in a terminal device or a network device, or a device that can be used in combination with a terminal device or a network device. In one possible implementation, the communication device 400 may include a module or unit that corresponds one-to-one to the method / operation / step / action performed by the first device or the second device in the above method embodiment. The unit may be a hardware circuit, or software, or a combination of a hardware circuit and software. In one possible implementation, as shown in Figure 8, the device 400 may include: a processing module 410 and a transceiver module 420.

[0304] Optionally, the communication device 400 may correspond to the first device in the above method embodiment.

[0305] In which, when the communication device 400 is used to execute the method on the first device side, the processing module 410 can be used to obtain M first data, where a sub-data in the first data corresponds to a first sparse matrix, and the first sparse matrix expresses a sub-data in the corresponding first data based on a first dictionary matrix, and the first dictionary matrix includes the features of the M sub-data corresponding to the M first data respectively; the transceiver module 420 can be used to output compressed data of the M first sparse matrices; wherein M is an integer greater than 1.

[0306] It should be understood that the specific process executed by each module has been described in detail in the above method embodiment, and for the sake of brevity, it will not be repeated here.

[0307] Optionally, the communication device 400 may correspond to the second device in the above method embodiment.

[0308] In which, when the communication device 400 is used to execute the method on the second device side, the transceiver module 420 can be used to receive compressed data of M first sparse matrices, wherein the first sparse matrix expresses the corresponding sub-data in the first data based on the first dictionary matrix, and the first dictionary matrix includes the features of M sub-data corresponding to the M first data respectively, and one sub-data in the first data corresponds to a first sparse matrix; the processing module 410 can be used to output decompression information based on the compressed data.

[0309] It should be understood that the specific process executed by each module has been described in detail in the above method embodiment, and for the sake of brevity, it will not be repeated here.

[0310] The transceiver module 420 in the communication device 400 can be implemented by a transceiver, for example, it can correspond to the transceiver 520 in the communication device 500 shown in Figure 9, and the processing module 410 in the communication device 400 can be implemented by at least one processor, for example, it can correspond to the processor 510 in the communication device 500 shown in Figure 9.

[0311] When the communication device 400 is a chip or chip system configured in a communication device (such as a terminal device or a network device), the transceiver module 420 in the communication device 400 can be implemented through an input / output interface, circuit, etc., and the processing module 410 in the communication device 400 can be implemented through a processor, microprocessor or integrated circuit integrated on the chip or chip system.

[0312] Figure 9 is another schematic block diagram of a communication device provided in an embodiment of the present application. As shown in Figure 9, the communication device 500 may include: a processor 510. The processor 510 may be used to execute the method executed by the first device or the second device in the above method embodiment.

[0313] In some possible implementations, the communication device 500 may include a transceiver 520. The transceiver 520 may communicate with the processor 510 via an internal connection path. The processor 510 may control the transceiver 520 to send and / or receive signals.

[0314] In some possible implementations, the communication device 500 may include a memory 530. The memory 530 may communicate with the processor 510 via an internal connection path. The memory 530 and the processor 510 may be integrated or provided separately. The memory 530 may also be a memory external to the device. The memory 530 is used to store instructions, and the processor 510 is used to execute the instructions stored in the memory 530 to perform the method in the above method embodiment.

[0315] It should be understood that the communication device 500 may correspond to the terminal device or network device in the above-mentioned method embodiment, and may be used to execute the various steps and / or processes performed by the first device or the second device in the above-mentioned method embodiment. Optionally, the memory 530 may include read-only memory and random access memory, and provide instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. The memory 530 may be a separate device or integrated into the processor 510. The processor 510 may be used to execute instructions stored in the memory 530, and when the processor 510 executes the instructions stored in the memory, the processor 510 is used to execute the various steps and / or processes of the above-mentioned method embodiment corresponding to the terminal device or network device.

[0316] Optionally, the communication device 500 is the first device in the above embodiment.

[0317] Optionally, the communication device 500 is the second device in the above embodiment.

[0318] The transceiver 520 may include a transmitter and a receiver. The transceiver 520 may further include an antenna, which may be one or more. The processor 510, memory 530, and transceiver 520 may be integrated on different chips. For example, the processor 510 and memory 530 may be integrated in a baseband chip, and the transceiver 520 may be integrated in a radio frequency chip. The processor 510, memory 530, and transceiver 520 may also be integrated on the same chip. This application does not limit this.

[0319] Optionally, the communication device 500 is a component configured in a terminal device, such as a chip, a chip system, etc.

[0320] Optionally, the communication device 500 is a component configured in a network device, such as a chip, a chip system, etc.

[0321] The transceiver 520 may also be a communication interface, such as an input / output interface, a circuit, etc. The transceiver 520, the processor 510, and the memory 530 may be integrated into the same chip, such as a baseband chip.

[0322] The present application also provides a processing device, comprising at least one processor, wherein the at least one processor executes a computer program or logic circuit to cause the processing device to execute the method executed by the first device or the second device in the above method embodiment. The processing device may also include a memory for storing the computer program.

[0323] An embodiment of the present application further provides a processing device, comprising a processor and an input / output interface. The input / output interface is coupled to the processor. The input / output interface is configured to input and / or output information. The information includes at least one of instructions and data. The processor is configured to execute a computer program to cause the processing device to perform the method performed by the first or second device in the above-described method embodiment.

[0324] The present application also provides a processing device including a processor and a memory. The memory is used to store a computer program, and the processor is used to call and execute the computer program from the memory, so that the processing device executes the method executed by the first device or the second device in the above method embodiment.

[0325] It should be understood that the processing device may be one or more chips. For example, the processing device may be a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on chip (SoC), a central processor unit (CPU), a network processor (NP), a digital signal processor (DSP), a microcontroller unit (MCU), a programmable logic device (PLD), or other integrated chips.

[0326] During implementation, each step of the above method can be completed by an integrated logic circuit of the hardware in the processor or by instructions in the form of software. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in conjunction with its hardware. To avoid repetition, it will not be described in detail here.

[0327] The processor in the embodiments of the present application can be an integrated circuit chip with signal processing capabilities. During implementation, the steps of the above method embodiments can be completed by hardware integrated logic circuits in the processor or instructions in software form. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0328] It is understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus RAM (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0329] According to the method provided in the embodiments of the present application, the present application also provides a computer program product, which includes: a computer program or a set of instructions, which, when the computer program or a set of instructions is run on a computer, enables the computer to execute the method executed by the first device or the second device in the above method embodiment.

[0330] According to the method provided in the embodiment of the present application, the present application also provides a computer-readable storage medium, which stores a program. When the program is run on a computer, the computer executes the method executed by the first device or the second device in the above method embodiment.

[0331] According to the method provided in the embodiment of the present application, the present application also provides a communication system, which may include the aforementioned first device or second device.

[0332] As used in this specification, the terms "component," "module," "system," and the like are used to represent computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. By way of illustration, both an application running on a computing device and a computing device can be a component. One or more components can reside in a process and / or an execution thread, and a component can be located on a computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures stored thereon. Components can communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0333] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0334] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0335] In the several embodiments provided in this application, the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0336] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0337] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0338] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the part that essentially contributes to the technical solution of the present application or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks or optical disks.

[0339] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A data compression transmission method, characterized in that: include: Acquire M first data, wherein one sub-data in the first data corresponds to a first sparse matrix, the first sparse matrix expresses one sub-data in the first data corresponding to the first sparse matrix based on a first dictionary matrix, and the first dictionary matrix includes features of the M sub-data respectively corresponding to the M first data; Output the compressed data of the M first sparse matrices, where M is an integer greater than 1.

2. The method according to claim 1, characterized in that For the p-th first data and the q-th first data among the M first data, the similarity between a sub-data in the p-th first data and a sub-data in the q-th first data is greater than or equal to a first similarity threshold; wherein, Both p and q are positive integers, and p is not equal to q.

3. The method according to claim 1 or 2, characterized in that: The M first data are data within M time units respectively, and one first data includes N sub-data divided according to spatial position relationship, where N is a positive integer; or, The M first data are data within h time units, and the data within the h time units are sorted according to the spatial position relationship to obtain the M first data, one first data includes N sub-data, and h is a positive integer; The similarity of the spatial positions between a sub-data in the p-th first data and a sub-data in the q-th first data among the M first data is greater than or equal to a first similarity threshold.

4. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: For the kth first data Y among the M first data k , the kth first data Y k A sub-data in is decomposed into the first dictionary matrix and the kth first sparse matrix, where k is a positive integer less than or equal to M; According to the first dictionary matrix, M-1 first sparse matrices other than the kth first sparse matrix are determined.

5. The method according to any one of claims 1 to 4, characterized in that: The outputting the compressed data of the M first sparse matrices comprises: Determine a first matrix according to the M first sparse matrices; Performing low-rank approximation on the first matrix to obtain first compressed data of the M first sparse matrices; The first compressed data is output.

6. The method according to claim 5, characterized in that The determining of the first matrix according to the M first sparse matrices comprises: Combining the M first sparse matrices to obtain the first matrix; or, Data compression is performed on at least one of the M first sparse matrices, and the M first sparse matrices after data compression are combined to obtain the first matrix.

7. The method according to claim 6, characterized in that The step of compressing data of at least one of the M first sparse matrices comprises: For one of the M first sparse matrices, according to the first position indication information, the value of the first element in the one first sparse matrix is ​​set to the first value; wherein, The first position indication information is used to indicate a sub-matrix in the first sparse matrix that has an expression corresponding to The position of the element of the data capability, the first element does not have the ability to express a sub-data in the first data.

8. The method according to any one of claims 5 to 7, characterized in that: The performing low-rank approximation on the first matrix to obtain first compressed data of the M first sparse matrices includes: Performing singular value decomposition on the first matrix to obtain K eigenvalues ​​and eigenvectors corresponding to the K eigenvalues, wherein the K eigenvalues ​​and the eigenvectors corresponding to the K eigenvalues ​​are used to express the first matrix; K0 eigenvalues ​​among the K eigenvalues ​​and the eigenvectors respectively corresponding to the K0 eigenvalues ​​are used as first compressed data of the M first sparse matrices.

9. The method according to claim 8, characterized in that Also includes: Output second compressed data of the M first sparse matrices, where the second compressed data includes: K1 eigenvalues ​​except the K0 eigenvalues ​​among the K eigenvalues ​​and eigenvectors corresponding to the K1 eigenvalues ​​respectively.

10. The method according to any one of claims 1 to 4, characterized in that: The outputting the compressed data of the M first sparse matrices comprises: Outputting first residual information, where the first residual information is determined based on information of the j-th first sparse matrix and the i-th first sparse matrix among the M first sparse matrices; Among them, i is less than j, and i and j are both positive integers.

11. The method according to claim 10, characterized in that The outputting the first residual information includes: Determine the similarity between the j-th first sparse matrix and the i-th first sparse matrix among the M first sparse matrices; When the similarity is less than or equal to a second similarity threshold, the first residual information is output.

12. The method according to claim 10 or 11, characterized in that: The information of the i-th first sparse matrix is ​​obtained by decompressing the compressed data of the i-th first sparse matrix.

13. The method according to any one of claims 10 to 12, characterized in that: The first residual information includes: a first residual element sequence, the first residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers.

14. The method according to claim 13, characterized in that The first residual information also includes: second position indication information, and the second position indication information is used to indicate that the absolute value in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix is ​​greater than or equal to.

15. The method according to any one of claims 10 to 14, characterized in that Also includes: Outputting second residual information, where the second residual information is determined based on information of the j-th first sparse matrix and the i-th first sparse matrix; wherein, The second residual information includes: third position indication information and a second residual element sequence, the third position indication information is used to indicate the residual elements in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold, and the second residual element sequence includes the residual elements in the residual matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold.

16. The method according to any one of claims 1 to 15, characterized in that The method further comprises: Sending the first dictionary matrix; or, The first dictionary matrix is ​​received.

17. The method according to any one of claims 5 to 9, characterized in that The outputting the first compressed data comprises: Performing compression processing on the first compressed data, wherein the compression processing includes quantization and / or entropy coding; The first compressed data after the compression process is output.

18. The method according to claim 11, characterized in that The outputting the first residual information includes: Performing compression processing on the first residual information, wherein the compression processing includes quantization and / or entropy coding; Output the first residual information after compression processing.

19. The method according to any one of claims 1 to 18, characterized in that Also includes: Sending a first indication message; or, receiving first indication information; The first indication information is used to indicate at least one of the following: The number of features of low-rank approximation K0; The number M of first data; A capability threshold, where the capability threshold is used to determine whether an element in the first sparse matrix has the capability of expressing a sub-data in the corresponding first data; a first residual threshold, where the first residual threshold is used to determine a first residual element sequence, where the first residual element sequence is used to express a residual matrix between information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and both i and j are positive integers; a proportion of elements capable of expressing a sub-data in the corresponding first data in one of the M first sparse matrices; data loss of first compressed data relative to the M first sparse matrices, the first compressed data being determined based on the M first sparse matrices; Parameters of compression processing, the compression processing including quantization and / or entropy coding, the parameters of compression processing including at least one of quantization accuracy, quantization codebook, and coding mode; Whether to send first residual information, wherein the first residual information is determined based on information of the j-th first sparse matrix and the ith first sparse matrix among the M first sparse matrices.

20. The method according to any one of claims 1 to 19, characterized in that Also includes: Determine at least one of the following according to the first time-frequency resources of the M first data: A capability threshold, where the capability threshold is used to determine whether an element in the first sparse matrix has the capability of expressing a sub-data in the corresponding first data; a first residual threshold, where the first residual threshold is used to determine a first residual element sequence, where the first residual element sequence is used to express a residual matrix between information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and both i and j are positive integers; Parameters of compression processing, wherein the compression processing includes quantization and / or entropy coding, and the parameters of the compression processing include at least one of quantization accuracy, quantization codebook, and coding mode.

21. The method according to claim 20, characterized in that Also includes: First configuration information is received, where the first configuration information is used to configure the first time-frequency resource.

22. The method according to claim 21, characterized in that Also includes: A compressed transmission request is sent, where the compressed transmission request carries a data type of the M first data, where the data type includes point cloud data and / or artificial intelligence AI data.

23. The method according to any one of claims 1 to 22, characterized in that Also includes: Sending a second indication message; or, receiving second indication information; The second indication information is used to indicate at least one of the following: The number of features of low-rank approximation is K1; A second residual threshold, wherein the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold, and the second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and both i and j are positive integers.

24. The method according to any one of claims 1 to 23, characterized in that Also includes: A second residual threshold is determined based on the second time-frequency resource of the M first data, and the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold. The second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and i and j are both positive integers.

25. The method according to claim 24, characterized in that Also includes: Receive second configuration information, where the second configuration information is used to configure the second time-frequency resource.

26. A data compression transmission method, characterized in that: include: Receive compressed data of M first sparse matrices, wherein the first sparse matrices express corresponding sub-data in the first data based on a first dictionary matrix, the first dictionary matrix includes features of the M sub-data corresponding to the M first data, and one sub-data in the first data corresponds to one first sparse matrix; Decompression information is output according to the compressed data.

27. The method according to claim 26, characterized in that For the p-th first data and the q-th first data among the M first data, the similarity between a sub-data in the p-th first data and a sub-data in the q-th first data is greater than or equal to a first similarity threshold; wherein, Both p and q are positive integers, and p is not equal to q.

28. The method according to claim 26 or 27, characterized in that The M first data are data within M time units respectively, and one first data includes N sub-data divided according to spatial position relationship, where N is a positive integer; or, The M first data are data within h time units, and the data within the h time units are sorted according to the spatial position relationship to obtain the M first data, one first data includes N sub-data, and h is a positive integer; The similarity of the spatial positions between a sub-data in the p-th first data and a sub-data in the q-th first data among the M first data is greater than or equal to a first similarity threshold.

29. The method according to any one of claims 26 to 28, characterized in that The compressed data includes first compressed data, and outputting decompression information according to the compressed data includes: Perform low-rank matrix recovery according to the first compressed data to obtain a first matrix; Determine the M first sparse matrices according to the first matrix; Constructing the M first data according to the M first sparse matrices and the first dictionary matrix; Output the M first data.

30. The method according to claim 29, characterized in that The determining the M first sparse matrices according to the first matrix includes: Splitting the first matrix to obtain the M first sparse matrices; or, The first matrix is ​​split to obtain decompression information of the M first sparse matrices, and data decompression is performed on at least one first sparse matrix among the M first sparse matrices according to the information of the M first sparse matrices.

31. The method according to claim 30, characterized in that The step of decompressing data of at least one of the M first sparse matrices comprises: For one of the M first sparse matrices, data decompression is performed on the one first sparse matrix according to the first position indication information; wherein, The first position indication information is used to indicate the position of an element in the first sparse matrix that has the ability to express a corresponding sub-data.

32. The method according to any one of claims 29 to 31, characterized in that The first compressed data includes K0 eigenvalues ​​and eigenvectors corresponding to the K0 eigenvalues ​​respectively, and the low-rank matrix recovery is performed according to the first compressed data to obtain a first matrix, including: Low-rank matrix recovery is performed according to the K0 eigenvalues ​​and the eigenvectors respectively corresponding to the K0 eigenvalues ​​to obtain the first matrix.

33. The method according to claim 32, characterized in that The compressed data also includes second compressed data, the second compressed data includes K1 eigenvalues ​​and eigenvectors corresponding to the K1 eigenvalues ​​respectively, and the K1 eigenvalues ​​are different from the K0 eigenvalues.

34. The method according to any one of claims 26 to 28, characterized in that The compressed data includes first residual information, the first residual information is determined based on information of the j-th first sparse matrix and the i-th first sparse matrix in the M first sparse matrices, and outputting decompression information according to the compressed data includes: Constructing the j-th first sparse matrix according to the first residual information and the i-th first sparse matrix; Constructing the M first data according to the j-th first sparse matrix; Outputting the M first data; Among them, i is less than j, and i and j are both positive integers.

35. The method according to claim 34, characterized in that The information of the i-th first sparse matrix is ​​obtained by decompressing the compressed data of the i-th first sparse matrix.

36. The method according to claim 34 or 35, characterized in that The first residual information includes: a first residual element sequence, the first residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, i is less than j, and i and j are both positive integers.

37. The method according to claim 36, characterized in that The first residual information also includes: second position indication information, and the second position indication information is used to indicate the position of the residual element whose absolute value is greater than or equal to the first residual threshold in the residual matrix between the information of the jth first sparse matrix and the i-th first sparse matrix.

38. The method according to any one of claims 34 to 37, characterized in that The compressed data further includes second residual information, which is determined based on information of the j-th first sparse matrix and the i-th first sparse matrix; wherein, The second residual information includes: third position indication information and a second residual element sequence, the third position indication information is used to indicate the residual elements in the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold, and the second residual element sequence includes the residual elements in the residual matrix that are less than or equal to the first residual threshold and greater than or equal to the second residual threshold.

39. The method according to any one of claims 26 to 38, characterized in that The method further comprises: receiving the first dictionary matrix; or, The first dictionary matrix is ​​transmitted.

40. The method according to any one of claims 26 to 39, characterized in that Also includes: receiving first indication information; or, Sending first instruction information; The first indication information is used to indicate at least one of the following: The number of features of low-rank approximation K0; The number M of first data; A capability threshold, where the capability threshold is used to determine whether an element in the first sparse matrix has the capability of expressing a sub-data in the corresponding first data; a first residual threshold, where the first residual threshold is used to determine a first residual element sequence, where the first residual element sequence is used to express a residual matrix between information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and both i and j are positive integers; a proportion of elements capable of expressing a sub-data in the corresponding first data in one of the M first sparse matrices; data loss of first compressed data relative to the M first sparse matrices, the first compressed data being determined based on the M first sparse matrices; Parameters of compression processing, the compression processing including quantization and / or entropy coding, the parameters of compression processing including at least one of quantization accuracy, quantization codebook, and coding mode; Whether to send first residual information, wherein the first residual information is determined based on information of the j-th first sparse matrix and the ith first sparse matrix among the M first sparse matrices.

41. The method according to any one of claims 26 to 40, characterized in that Also includes: Send first configuration information, where the first configuration information is used to configure first time-frequency resources of the M first data.

42. The method according to claim 41, characterized in that Also includes: A compressed transmission request is received, where the compressed transmission request carries a data type of the M first data, where the data type includes point cloud data and / or artificial intelligence AI data.

43. The method according to any one of claims 26 to 42, characterized in that Also includes: receiving the second indication information sent; or, Sending second instruction information; The second indication information is used to indicate at least one of the following: The number of features of low-rank approximation is K1; A second residual threshold, wherein the second residual threshold is used to determine a second residual element sequence in combination with the first residual threshold, and the second residual element sequence is used to express the residual matrix between the information of the j-th first sparse matrix and the i-th first sparse matrix, where i is less than j, and both i and j are positive integers.

44. The method according to any one of claims 26 to 43, characterized in that Also includes: Second configuration information is received, where the second configuration information is used to configure second time-frequency resources of the M first data.

45. A communication device, characterized in that: The method comprises a module for executing the method as claimed in any one of claims 1 to 25, or comprises a module for executing the method as claimed in any one of claims 26 to 44.

46. ​​A communication device, characterized in that: include: A processor, the processor being configured to execute the method according to any one of claims 1 to 44 by running a computer program or by a logic circuit.

47. A computer-readable storage medium, characterized in that Used to store computer program instructions, the computer program causing a computer to execute the method according to any one of claims 1 to 44.

48. A computer program product, characterized in that The method comprises computer program instructions which cause a computer to execute the method as claimed in any one of claims 1 to 44.

Images