Quantization method and apparatus

Abstract

The present application provides a quantization method that can be applied to an application scenario in which two devices share the same quantization dictionary. This method proposes the idea of ​​dividing information to be quantized into multiple parts and quantizing the quantization parts individually. A representation format for the quantization dictionary is provided. The quantization dictionary can be represented by using dictionary usage and dictionary content. In this representation format, a correspondence is actually established between R quantization parts of the information to be quantized and M quantization sub-dictionaries. If a first sub-dictionary among the M quantization sub-dictionaries corresponds to a first quantization part among the R quantization parts, the first sub-dictionary is used to quantize the first quantization part or to dequantize information obtained by quantizing the first quantization part, and each dictionary element in the first sub-dictionary is a candidate for quantization information of the first quantization part. The use of a representation format for the quantization dictionary can reduce the overhead of storing the quantization dictionary by a device.

Description

[Technical Field]

[0001] FIELD Embodiments of the present application relate to the field of wireless communication technology, and more particularly to quantization methods and apparatus. [Background technology]

[0002] In many scenarios in the field of wireless communication technology, a user equipment (UE) needs to feed back channel state information (CSI) to a base station. Generally, a UE measures a downlink reference signal transmitted by a base station, estimates a downlink channel matrix based on the received downlink reference signal, and finally generates CSI and feeds the CSI back to the base station. The more accurate the CSI fed back by the UE, the richer the information. The more accurate the channel information recovered by the base station based on the CSI, and the more accurate the precoding matrix determined by the base station. However, as the scale of antenna arrays in communication systems increases, the number of supported antenna ports also increases. Because the size of the complete channel matrix is ​​directly proportional to the number of antenna ports, in large-scale multiple-input multiple-output (MIMO) systems, the UE feeds back the complete channel matrix to the base station through CSI. This means that the feedback overhead is very large. Machine learning methods have stronger nonlinear feature extraction capabilities and can more effectively extract channel matrix features. Compared with traditional CSI feedback solutions, machine learning methods can include more channel information in the same scale of feedback, thereby reducing the information loss of CSI compressed feedback and improving the channel recovery accuracy of the base station. Compared with traditional solutions, machine learning methods can use less feedback to represent the same channel information, further reducing feedback overhead.

[0003] In the existing deep learning-based channel information compensation feedback mechanism, compressed channel information processing is performed on the channel information by using a deep learning-based encoder at the UE side, the information obtained through the compression processing is quantized and represented by using a quantization dictionary, and the index of the quantized representation in the quantization dictionary is fed back to the base station. The base station maintains the same quantization dictionary as the UE. Based on the quantization dictionary, the base station can restore the received index to a quantized representation and input the quantized representation into a deep learning-based decoding process to recover the channel information.

[0004] In the above mechanism, the quantization dictionary is usually large, resulting in a very large storage overhead at the base station side and the UE side. Summary of the Invention

[0005] The present application provides a quantization method and apparatus to reduce the overhead of maintaining a quantization dictionary.

[0006] According to a first aspect, there is provided a quantization method, the method being applied to a communication device (e.g., the first device of the present application), the method comprising: A first device obtaining a first quantization dictionary; The first device quantizes the information to be quantized by using a first quantization dictionary, the first quantization dictionary including M sub-dictionaries, the M sub-dictionaries being used to quantize R quantized portions of the information to be quantized, where M and R are both positive integers greater than 1; or The first device dequantizes the received information obtained by quantizing the R quantization portions by using a first quantization dictionary, the first quantization dictionary including M sub-dictionaries, the M sub-dictionaries being used to dequantize the received information obtained by quantizing the R quantization portions, and M and R are both positive integers greater than 1; Includes:

[0007] The first quantization dictionary includes a dictionary usage and dictionary contents of each of the M sub-dictionaries. The dictionary usage indicates a correspondence between the M sub-dictionaries and the R quantization portions, and a first sub-dictionary among the M sub-dictionaries corresponds to a first quantization portion among the R quantization portions, and the first sub-dictionary is used to quantize the first quantization portion or the first sub-dictionary is used to dequantize information obtained by quantizing the first quantization portion. The dictionary contents of each sub-dictionary among the M sub-dictionaries include at least one dictionary element, and each dictionary element included in the first sub-dictionary among the M sub-dictionaries is a candidate for quantization information of the first quantization portion.

[0008] This technical solution provides the idea of ​​dividing the information to be quantized into multiple quantization portions and individually quantizing each quantization portion. Based on this, a representation format of a quantization dictionary is provided. Based on the representation format, the quantization dictionary can be represented by a dictionary usage and a dictionary content. In the dictionary representation format, a correspondence is actually established between R quantization portions of the information to be quantized and M sub-dictionaries (sub-dictionaries for short). In the correspondence, if a first sub-dictionary among the M sub-dictionaries corresponds to a first quantization portion among the R quantization portions, the first sub-dictionary is used to quantize the first quantization portion, or the first sub-dictionary is used to dequantize information obtained by quantizing the first quantization portion, and each dictionary element in the dictionary content of the first sub-dictionary is a candidate for the first quantization portion. Based on the representation format of the quantization dictionary provided in this application, the overhead of maintaining the quantization dictionary by a device can be reduced.

[0009] Optionally, after quantizing the information to be quantized by using the first quantization dictionary to obtain the quantized information, the first device can send the quantized information to the second device.

[0010] Optionally, the information obtained by individually quantizing the R quantized portions and received by the first device is from the second device.

[0011] Furthermore, in known deep learning-based channel information compression feedback mechanisms, the base station and the UE each maintain the same quantization dictionary. The quantization dictionary is used for quantization (or dequantization) in various scenarios. In other words, the mechanism does not distinguish between scenarios. For example, the same quantization dictionary is used in high-speed scenarios, low-speed scenarios, indoor scenarios, outdoor scenarios, etc. Considering that the value parts of the information to be quantized in different scenarios may be completely different or have their own characteristics, if the same quantization dictionary is used uniformly and not distinguished between scenarios, the quantization performance will be poor.

[0012] Therefore, the present application provides another quantization method, which is designed to use different quantization dictionaries in different scenarios, so that the quantization dictionaries can adapt to the value distribution of the information to be quantized in the current scenario, so as to improve the quantization performance.

[0013] According to a second aspect, there is provided a quantization method, the method being applied to a communication device (e.g., the first device of the present application), the method comprising: A first device obtaining a first quantization dictionary; The first device quantizes the first quantization target information by using a first quantization dictionary, or the first device dequantizes the received first quantized information by using the first quantization dictionary; The first device obtains a second quantization dictionary; The first device quantizes the second quantization target information by using the second quantization dictionary, or the first device dequantizes the received second quantized information by using the second quantization dictionary; Includes:

[0014] The first quantization dictionary includes a plurality of dictionary elements, and each dictionary element included in the first quantization dictionary is a candidate for quantization target information to be generated in the first quantization scenario.

[0015] The second quantization dictionary includes a plurality of dictionary elements, and each dictionary element included in the second quantization dictionary is a candidate for quantization target information generated in the second quantization scenario.

[0016] The first quantization dictionary is different from the second quantization dictionary.

[0017] Optionally, the first quantization dictionary corresponds to a first quantization scenario and the second quantization dictionary corresponds to a second quantization scenario.

[0018] Optionally, the value distribution of the first quantization target information is different from the value distribution of the second quantization target information.

[0019] In other words, whether the value distributions of the information to be quantized are the same can be used as a basis for determining whether the information to be quantized is in a different quantization scenario.

[0020] For example, if the value distribution of the quantization target information generated by the first quantization scenario is the same as or close to the value distribution of the quantization target information generated by the second quantization scenario, for example, does not exceed a determined threshold, then the first quantization scenario and the second quantization scenario are actually the same quantization scenario. In this case, when the first quantization scenario changes to the second quantization scenario, or when the second quantization scenario changes to the first quantization scenario, the quantization dictionary does not need to be updated. In contrast, if the value distribution of the quantization target information generated by the first quantization scenario is significantly different from the value distribution of the quantization target information generated by the second quantization scenario, for example, exceeds a determined threshold, then the first quantization scenario and the second quantization scenario are considered to be different quantization scenarios. When the first quantization scenario changes to the second quantization scenario, or when the second quantization scenario changes to the first quantization scenario, a quantization dictionary adapted to the current quantization scenario needs to be selected.

[0021] For example, in some communication system positioning scenarios, when positioning information, e.g., the position coordinates of a device to be located, needs to be quantized, a change in the position distribution means a change in the positioning scenario. As another example, in a scenario in which a UE feeds back CSI to a network side, a change in the channel condition causes a change in the distribution of the CSI value part, which can be considered as a change in the CSI feedback scenario. These various scenarios are examples of the various quantization scenarios mentioned above.

[0022] Optionally, in different quantization scenarios, the value distribution of the quantization target information may be different, for example, two-dimensional, three-dimensional, or more dimensional, without being limited thereto. For example, in a positioning scenario, the positioning information may be a three-dimensional spatial coordinate.

[0023] Optionally, in the method according to the second aspect, the first device obtaining the quantization dictionary includes the first device obtaining the quantization dictionary from a quantization dictionary table. The quantization dictionary table includes at least one candidate quantization dictionary. For example, in the second aspect, the first quantization dictionary and / or the second quantization dictionary are obtained (or selected) from the quantization dictionary table. For example, the first device may select an appropriate quantization dictionary from the quantization dictionary table based on different quantization scenarios.

[0024] Referring to the second aspect, in a first implementation of the second aspect, the first quantization dictionary includes M sub-dictionaries, and the M sub-dictionaries are used to individually quantize R quantization portions of the information to be quantized, or are used to dequantize information obtained by individually quantizing the R quantization portions, where M and R are both positive integers.

[0025] The second quantization dictionary includes C sub-dictionaries, which are used to individually quantize S quantization portions of the information to be quantized or to dequantize information obtained by individually quantizing the S quantization portions, where S and C are both positive integers.

[0026] The first quantization dictionary includes a dictionary usage and dictionary contents of each of the M sub-dictionaries. The dictionary usage indicates a correspondence between the M sub-dictionaries and the R quantization portions, and a first sub-dictionary among the M sub-dictionaries corresponds to a first quantization portion among the R quantization portions, and the first sub-dictionary is used to quantize the first quantization portion or the first sub-dictionary is used to dequantize information obtained by quantizing the first quantization portion. The dictionary contents of each sub-dictionary among the M sub-dictionaries include at least one dictionary element, and each dictionary element included in the first sub-dictionary among the M sub-dictionaries is a candidate for quantization information of the first quantization portion.

[0027] The second quantization dictionary includes a dictionary usage and dictionary contents of each of the C sub-dictionaries. The dictionary usage indicates a correspondence between the C sub-dictionaries and the S quantization portions, and a first sub-dictionary of the C sub-dictionaries corresponds to a first quantization portion of the S quantization portions, and the first sub-dictionary is used to quantize the first quantization portion or the first sub-dictionary is used to dequantize information obtained by quantizing the first quantization portion. The dictionary contents of each sub-dictionary of the C sub-dictionaries include at least one dictionary element, and each dictionary element included in the first sub-dictionary of the C sub-dictionaries is a candidate for quantization information of the first quantization portion.

[0028] Although the number of sub-dictionaries or the dictionary contents of one or more sub-dictionaries may differ, the structure of the second quantization dictionary is similar to the structure of the first quantization dictionary, so the following description of the first quantization dictionary may also be applicable to the second quantization dictionary. For simplicity, the following uses the first quantization dictionary as an example for explanation.

[0029] Furthermore, for the dictionary content of the quantization dictionary in the second aspect and the specific implementation and procedure of obtaining the quantization dictionary by the first device, please refer to the specific implementation and procedure of obtaining the quantization dictionary by the first device in the first aspect, and the details will not be described again in this specification.

[0030] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, each dictionary element included in the first sub-dictionary corresponds to one piece of quantization information and one index; Each dictionary element included in the first sub-dictionary is a candidate for quantization information of the first quantization part, The quantization information corresponding to each dictionary element included in the first sub-dictionary is a candidate for quantization information of the first quantization part.

[0031] In this implementation, based on the representation format of the quantization dictionary, the dictionary content of each sub-dictionary in the quantization dictionary includes dictionary elements, and each dictionary element corresponds to one quantization information and one index. The index of a dictionary element can uniquely indicate the dictionary element in the sub-dictionary and further can uniquely correspond to the quantization information corresponding to the dictionary element. Therefore, when an updated portion of the quantization dictionary is transmitted, if only some of the dictionary elements in the sub-dictionary are updated, only the updated dictionary elements in the sub-dictionary need to be transmitted, and not the complete quantization dictionary, so the overhead of transmitting the quantization dictionary can be reduced.

[0032] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device obtaining the first quantization dictionary includes: The first device updates the second quantization dictionary to obtain the first quantization dictionary; The method is: The method further includes the first device sending all or a portion of the first quantization dictionary to the second device.

[0033] In this implementation, the first device updates the quantization dictionary and provides the updated quantization dictionary for use by the second device.

[0034] Optionally, the first device is a terminal device, and the second device is an access network device. Since the function of the quantization dictionary is to quantize the output of the encoder at the terminal side, the terminal device can obtain a quantization dictionary with higher quantization precision based on the output distribution of the encoder.

[0035] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the M sub-dictionaries include P sub-dictionaries and Q sub-dictionaries.

[0036] The P sub-dictionaries are selected by the first device from a quantization dictionary table stored in the first device, the quantization dictionary table including at least the P candidate sub-dictionaries.

[0037] The Q sub-dictionaries are obtained by the first device through dictionary training.

[0038] 0≦P≦M, 0≦Q≦M, P+Q=M, and P and Q are integers.

[0039] In this implementation, the first quantization dictionary obtained by updating by the first device includes M sub-dictionaries. The M sub-dictionaries may be selected from a quantization dictionary table, obtained through dictionary training, or obtained by a combination of the two methods. The first device updates the quantization dictionary based on changes in application scenarios (e.g., from a slow scenario to a fast scenario), so that consistently excellent quantization performance can be obtained in various different scenarios.

[0040] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device sending a portion of the first quantization dictionary to the second device includes: The method includes the first device sending a first message to the second device, the first message indicating an updated portion of the first quantization dictionary compared to the second quantization dictionary.

[0041] Optionally, the first message "indicating" the updated portion can alternatively be expressed as the first message "carrying" the updated portion, and the specific method of indicating the updated portion is not limited. Specifically, the indication may be performed by including the updated portion or by other information corresponding to the updated portion. The correspondence between the updated portion and the other information may be predefined in the protocol. For example, the correspondence may be pre-stored in the first device and the second device or pre-configured in the first device and / or the second device. The pre-configuration may be performed by the first device and / or the second device, or by another device other than the first device and the second device.

[0042] Optionally, the first quantization dictionary corresponds to a first quantization scenario and the second quantization dictionary corresponds to a second quantization scenario.

[0043] Optionally, the value distribution of the first quantization target information is different from the value distribution of the second quantization target information.

[0044] In other words, whether the value distributions of the information to be quantized are the same can be used as a basis for determining whether the information to be quantized is in a different quantization scenario.

[0045] For example, if the value distribution of the quantization target information generated by the first quantization scenario is the same as or close to the value distribution of the quantization target information generated by the second quantization scenario, for example, does not exceed a determined threshold, then the first quantization scenario and the second quantization scenario are actually the same quantization scenario. In this case, when the first quantization scenario changes to the second quantization scenario, or when the second quantization scenario changes to the first quantization scenario, the quantization dictionary does not need to be updated. In contrast, if the value distribution of the quantization target information generated by the first quantization scenario is significantly different from the value distribution of the quantization target information generated by the second quantization scenario, for example, exceeds a determined threshold, then the first quantization scenario and the second quantization scenario are considered to be different quantization scenarios. When the first quantization scenario changes to the second quantization scenario, or when the second quantization scenario changes to the first quantization scenario, a quantization dictionary adapted to the current quantization scenario needs to be selected.

[0046] For example, in some communication system positioning scenarios, when positioning information, e.g., the position coordinates of a device to be located, needs to be quantized, a change in the position distribution means a change in the positioning scenario. As another example, in a scenario in which a UE feeds back CSI to a network side, a change in the channel condition causes a change in the distribution of the CSI value part, which can be considered as a change in the CSI feedback scenario. These various scenarios are examples of the various quantization scenarios mentioned above.

[0047] Optionally, in different quantization scenarios, the value distribution of the quantization target information may be different, for example, two-dimensional, three-dimensional, or more dimensional, without being limited thereto. For example, in a positioning scenario, the positioning information may be a three-dimensional spatial coordinate.

[0048] In this implementation, since the representation format of the quantization dictionary provided in the present application is used, after the quantization dictionary is updated, the first device only needs to transmit the updated portion of the quantization dictionary to the second device, rather than the entire updated quantization dictionary, to convey the updated quantization dictionary from the first device to the second device. Because the complete quantization dictionary is very large, transmitting only the updated portion can reduce the overhead of transmitting the quantization dictionary.

[0049] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the updated part and / or comprising a correspondence between L quantization portions of the R quantization portions and the sub-dictionaries, where 1≦L≦R and L is an integer; and / or It contains dictionary contents of sub-dictionaries corresponding to Z quantization portions out of the R quantization portions, where 1≦Z≦R and Z is an integer.

[0050] In this implementation, updating the quantization dictionary may include updating the dictionary usage, updating the dictionary content, or updating both the dictionary usage and the dictionary content. The overhead of transmitting the quantization dictionary may be reduced because when a first device transmits an updated portion to a second device, only some or all of the updated correspondence and / or some or all of the updated dictionary content need be transmitted.

[0051] It should be understood that the above updated portion including correspondence between L quantization portions and sub-dictionaries indicates that the dictionary usage of the first quantization dictionary is updated, and the updated portion including dictionary content of sub-dictionaries corresponding to Z quantization portions indicates that the dictionary content of the first quantization dictionary is updated.

[0052] It should be understood that a portion of the correspondences updated refers to the fact that the correspondences between only some of the R quantization portions and the sub-dictionaries have been updated, but not all of the correspondences between the R quantization portions and the sub-dictionaries have been updated. Similarly, a portion of the dictionary content updated refers to the fact that some of the dictionary content in the quantization dictionary has been updated, for example, dictionary elements in some sub-dictionaries have been updated, but not the dictionary content in all sub-dictionaries. Furthermore, for sub-dictionaries whose dictionary content has been updated, only some of the dictionary elements may be updated, but not every dictionary element. Compared to transmitting the complete quantization dictionary, this implementation can reduce transmission overhead unless all correspondences are updated and all dictionary elements in all sub-dictionaries are updated. Furthermore, the smaller the updated portion of the quantization dictionary, the greater the reduction in transmission overhead compared to transmitting the complete quantization dictionary.

[0053] With reference to the first aspect, in some implementations of the first aspect, when the updated portion includes correspondences between L quantization portions and sub-dictionaries among the R quantization portions, the first message indicates identifiers of the L quantization portions and the sub-dictionaries that respectively correspond to the L quantization portions; and / or If the updated portion includes dictionary contents of sub-dictionaries corresponding to Z quantization portions of the R quantization portions, the first message indicates identifiers of the Z sub-dictionaries and updated dictionary elements and indices of the Z sub-dictionaries.

[0054] Optionally, in this implementation, there may be more than one first message, in other words, the information contained in the first message may be carried in the same message or in more than one message.

[0055] For example, the first message may include multiple messages, such as first message A including L quantized portions, and first message B carrying identifiers of sub-dictionaries corresponding to the L quantized portions, in a preset order.

[0056] Alternatively, as another example, each of the multiple messages carries the quantization portions and the identifiers corresponding to the quantization portions in a predetermined order. For example, the predetermined order includes the order of the quantization portions in the information. For example, a first message a includes a first quantization portion of the L quantization portions and an identifier of a sub-dictionary corresponding to the first quantization portion, a first message b includes a second quantization portion of the L quantization portions and an identifier of a sub-dictionary corresponding to the second quantization portion, and so on.

[0057] Additionally, the updated dictionary elements and indexes include one or more of the following: Newly added dictionary elements and indexes; and The dictionary element that was changed and the index that corresponds to the dictionary element.

[0058] Optionally, a sub-dictionary may be considered a newly added sub-dictionary if its identifier is a newly added identifier in the quantization dictionary. If the dictionary usage update included in the updated portion includes an identifier of a newly added sub-dictionary, the dictionary content of the sub-dictionary included in the updated portion may further include all dictionary elements in the sub-dictionary and the corresponding indexes.

[0059] In this implementation, the first device indicates the updated portion of the quantization dictionary to the second device, so transmission overhead may be reduced.

[0060] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, before the first device acquires the first quantization dictionary, the method includes: The method further includes the first device sending first information to the second device, the first information indicating that the first device updates the second quantization dictionary.

[0061] In this implementation, since the first device and the second device share one quantization dictionary, the quantization dictionary cannot be updated separately by the first device and the second device, but needs to be completed on one side and then transmitted to the other side. When the first information indicates that the first device updates the quantization dictionary, the second device does not change the contents of the quantization dictionary in the process of updating the encoder or decoder of the second device, thereby ensuring that the quantization dictionaries held by the first device and the second device are always consistent. Optionally, in a scenario in which the terminal device feeds back CSI to the network device, the accuracy of the CSI recovered on the network side can be ensured.

[0062] Referring to the first aspect, in some implementations of the first aspect, after the first device sends all or a portion of the first quantization dictionary to the second device, the method includes: The method further includes the first device receiving second information from the second device, the second information being used to request that the first quantization dictionary be enabled.

[0063] In implementation, the first device sends a first quantization dictionary to the second device. If the first quantization dictionary is received correctly and the second device determines to use the first quantization dictionary, the second device requests the first device to enable the first quantization dictionary, thereby improving the reliability of synchronously updating the quantization dictionaries between the two devices. For example, if the second device does not receive the first quantization dictionary correctly, or if the quantization performance does not satisfy requirements after evaluation after the second device receives the first quantization dictionary, the second device may not accept the new quantization dictionary to be enabled.

[0064] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the method comprises: The method further includes the first device sending third information to the second device, the third information indicating that the first quantization dictionary is enabled.

[0065] In this implementation, to maintain synchronized updates of the quantization dictionary between the first device and the second device, the first device sends third information to the second device, based on information that the second device has requested that a new quantization dictionary be enabled, to instruct the second device to enable the new quantization dictionary.

[0066] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the quantization of the quantization target information by the first device using the first quantization dictionary includes: the first device enabling the first quantization dictionary to quantize the to-be-quantized information after a time interval T from the moment the third information is sent, the time interval T being indicated by using the third information or agreed upon based on a protocol; or The first device dequantizes the received information obtained by quantizing the R quantization portions by using the first quantization dictionary. The method includes the first device enabling the first quantization dictionary to dequantize the received information obtained by quantizing the R quantization portions after a time interval T from the moment the third information is sent, the time interval T being indicated to the second device by using the third information or agreed upon based on a protocol.

[0067] Optionally, the time interval T may be determined by the second device and transmitted to the first device. For example, the time interval T may be carried in the second information used to request that the first quantization dictionary be enabled. This is not a limitation. Alternatively, the time interval T may be determined by the first device and the second device through negotiation using other signaling or messages, and is not limited to being carried or indicated by the second information or the third information.

[0068] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device obtaining the first quantization dictionary includes: The first device receives all or a portion of the first quantization dictionary from the second device.

[0069] In this implementation, the quantization dictionary is updated by the second device, and the new updated quantization dictionary is communicated to the first device. According to the quantization dictionary representation format provided herein, the first device receives the first quantization dictionary obtained by the update. Specifically, the first device receives an updated portion of the dictionary usage and an updated portion of the dictionary content in the first quantization dictionary. Compared to transmitting the complete quantization dictionary, the overhead of transmitting the quantization dictionary can be reduced.

[0070] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, receiving a portion of the first quantization dictionary from the second device by the first device includes: The method includes the first device receiving a second message from the second device, the second message indicating an updated portion of the first quantization dictionary compared with a second quantization dictionary, the first quantization dictionary being obtained by updating the second quantization dictionary by the second device.

[0071] With reference to the first aspect, in some implementations of the first aspect, the updated part and / or comprising a correspondence between L quantization portions of the R quantization portions and the sub-dictionaries, where 1≦L≦R and L is an integer; and / or It contains dictionary contents of sub-dictionaries corresponding to Z quantization portions out of the R quantization portions, where 1≦Z≦R and Z is an integer.

[0072] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, when the updated portion includes correspondences between L quantization portions and sub-dictionaries among the R quantization portions, the second message indicates identifiers of the L quantization portions and the sub-dictionaries respectively corresponding to the L quantization portions; and / or If the updated portion includes dictionary contents of sub-dictionaries corresponding to Z quantization portions of the R quantization portions, the second message indicates identifiers of the Z sub-dictionaries and updated dictionary elements and indices of the Z sub-dictionaries.

[0073] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the method comprises: If the first device determines that the obtained first quantization dictionary is defective, the first device sends fourth information to the second device, the fourth information instructing the second device to send a defective portion of the first quantization dictionary, the defective portion including a dictionary usage defect and / or a dictionary content defect of the first quantization dictionary; the first device receiving the defective portion from the second device; Further includes:

[0074] In this implementation, a device that receives a quantization dictionary from another party needs to determine whether the quantization dictionary has been completely received. If the quantization dictionary is defective, the other party can be requested to send the defective portion, so that the reliability of transmitting the quantization dictionary can be ensured.

[0075] Referring to the first aspect, in some implementations of the first aspect, before the first device acquires all or a portion of the first quantization dictionary, the method includes: The method further includes the first device receiving fifth information from the second device, the fifth information indicating that the second device updates the quantization dictionary.

[0076] Referring to the first aspect, in some implementations of the first aspect, after the first device acquires all or a portion of the first quantization dictionary, the method includes: The method further includes the first device sending sixth information to the second device, where the sixth information is used to request that the first quantization dictionary be enabled.

[0077] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the method comprises: The method further includes the first device receiving seventh information from the second device, the seventh information indicating enabling the first quantization dictionary.

[0078] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the quantization of the quantization target information by the first device using the first quantization dictionary includes: the first device enabling the first quantization dictionary to quantize the to-be-quantized information after a time interval T from the moment the seventh information is received, the time interval T being indicated by using the seventh information or agreed upon based on a protocol; or The first device dequantizes the received information obtained by quantizing the R quantization portions by using the first quantization dictionary. The method includes the first device enabling the first quantization dictionary to dequantize the received information obtained by quantizing the R quantization portions after a time interval T from the moment the seventh information is received, the time interval T being indicated by using the seventh information or agreed upon based on the protocol.

[0079] Optionally, the time interval T may be determined by the first device and transmitted to the second device. For example, the time interval T may be carried in the sixth information used to request that the first quantization dictionary be enabled. This is not a limitation. Alternatively, the time interval T may be determined by the first device and the second device through negotiation using other signaling or messages, and is not limited to being carried or indicated by the sixth or seventh information.

[0080] In some of the above implementations, the second device updates the quantization dictionary and transmits the updated quantization dictionary to the first device. The technical effects of these implementations are equivalent to the technical effects of the corresponding implementations in which the first device updates the quantization dictionary and transmits the updated quantization dictionary to the second device. Details will not be described again.

[0081] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device is an access network device and the second device is a terminal device.

[0082] After the first device obtains the first quantization dictionary, the method includes: evaluating, by the first device, a quantization performance of the first quantization dictionary; When it is determined that the decoder of the first device does not conform to the first quantization dictionary, the first device updates the decoder based on the first quantization dictionary. Further includes:

[0083] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device is an access network device and the second device is a terminal device.

[0084] Before the first device enables the first quantization dictionary, the method includes: receiving, by the first device, the first quantized information from the second device; the first device dequantizing the first quantized information based on the first quantized information and the second quantization dictionary; Further includes:

[0085] In this implementation, before activating the updated quantization dictionary, the access network device dequantizes the received quantized information (or recovers the channel information) by using the pre-update quantization dictionary, ensuring that the access network device and the terminal device always use the same quantization dictionary in a synchronized manner.

[0086] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device is an access network device and the second device is a terminal device.

[0087] After the first device enables the first quantization dictionary, the method includes: receiving, by the first device, second quantized information from the second device; the first device dequantizing the second quantized information based on the second quantized information and the first quantization dictionary; Further includes:

[0088] In this implementation, after activating the new quantization dictionary, the access network device dequantizes the received quantized information by using the new quantization dictionary, ensuring that the access network device and the terminal device always use the same quantization dictionary in synchronization.

[0089] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device is a terminal device and the second device is an access network device.

[0090] After the first device obtains the first quantization dictionary, the method includes: evaluating, by the first device, a quantization performance of the first quantization dictionary; and when it is determined that an output of the encoder of the first device does not conform to the first quantization dictionary, the first device updates the encoder based on the first quantization dictionary. Further includes:

[0091] In this implementation, the quantization dictionary is updated by the base station, and the updated quantization dictionary is communicated to the terminal device. The terminal device evaluates the quantization performance of the quantization dictionary, and if the encoder does not adapt to the new quantization dictionary, updates the encoder based on the new quantization dictionary, so that the encoder can better adapt to the new quantization dictionary, thereby improving the quantization performance.

[0092] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device is a terminal device and the second device is an access network device.

[0093] Before the first device enables the first quantization dictionary, the method includes: The first device determines quantized information to be fed back of the first channel information based on the second quantization dictionary; the first device sending quantized information of the first channel information to the second device; Further includes:

[0094] In this implementation, before activating the updated quantization dictionary, the terminal device quantizes the channel information by using the pre-updated quantization dictionary, thereby ensuring that the terminal device and the network device always use the same quantization dictionary in synchronization.

[0095] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the first device is a terminal device and the second device is an access network device.

[0096] After the first device enables the first quantization dictionary, the method includes: The first device determines quantized information to be fed back of the second channel information based on the first quantization dictionary; the first device sending quantized information of the second channel information to the second device; Further includes:

[0097] In this implementation, after activating the updated quantization dictionary, the terminal device quantizes the channel information by using the updated quantization dictionary, thereby ensuring that the terminal device and the network device always use the same quantization dictionary in synchronization.

[0098] It should be understood that both the first channel information and the second channel information in this implementation are examples of quantization target information. In other words, an example of quantization target information can be channel information, for example, CSI. Specifically, for example, the channel information can be output information of an encoder on the UE side, for example, compressed information of CSI.

[0099] With reference to the first implementation of the first aspect or the second aspect, in some implementations of the first aspect or the second aspect, the quantization of the quantization target information by the first device using the first quantization dictionary includes: When the first device is a terminal device and the information to be quantized is output information of an encoder, the first device quantizes the output information of the encoder by using a first quantization dictionary; or When the first device is an access network device, the first device dequantizes the received information obtained by quantizing the R quantization portions by using a first quantization dictionary.

[0100] According to a third aspect, there is provided a quantization or dequantization device, which has a function of performing the method of the first aspect or the second aspect or any one of the possible implementations of the first aspect or the second aspect. The function may be performed by hardware or by executing corresponding software by the hardware. The hardware or software includes one or more units corresponding to the above functions.

[0101] Optionally, the apparatus may be an access network device, a device, module, chip (or chip system) etc. within the access network device, or an apparatus usable with the access network device. Alternatively, the apparatus may be a terminal device, a device, module, chip (or chip system) within the terminal device, or an apparatus usable with the access network device.

[0102] In one design, the apparatus may include modules that correspond one-to-one to the methods / operations / steps / actions described in the first aspect. The modules may be implemented by hardware circuitry, software, or a combination of hardware circuitry and software. In another design, the apparatus may include a processing module and a communication module.

[0103] For example, the processing module is configured to perform the processes / operations / steps / actions performed in the device in the first or second aspect or any one of the possible implementations of the first or second aspect. The communication module is configured to perform the operations such as sending / receiving / transmitting performed by the device. For details, please refer to the description of the device embodiments. Alternatively, the communication module may be referred to as a transceiver module and includes receiving and / or transmitting functions.

[0104] According to a fourth aspect, there is provided a quantization or dequantization device. The device includes a processor and a memory. Optionally, the quantization or dequantization device may further include a transceiver. The memory is configured to store a computer program. The processor is configured to invoke and load the computer program stored in the memory and control the transceiver to receive and transmit signals, thereby enabling the communication device to perform a method according to either the first aspect or the second aspect or any one of possible implementations of the first aspect or the second aspect.

[0105] According to a fifth aspect, there is provided a quantization or dequantization device. The device includes a processor and a communication interface. The communication interface is configured to receive data and / or information and convey the received data and / or information to the processor. The processor processes the data and / or information. Furthermore, the communication interface is further configured to output the data and / or information processed by the processor to enable a method according to either the first aspect or the second aspect or any one of possible implementations of the first aspect or the second aspect to be performed.

[0106] For example, the communication interface may be a transceiver, a circuit, a bus, a module, a pin, or other type of communication interface.

[0107] According to a sixth aspect, there is provided a quantization or dequantization apparatus, the apparatus including at least one processor, coupled to at least one memory, configured to execute computer programs or instructions stored in the at least one memory to enable the communications apparatus to perform a method according to either the first aspect or the second aspect or any one of possible implementations of the first or second aspect.

[0108] Optionally, the device may be a chip or a chip system.

[0109] According to a seventh aspect, there is provided a computer-readable storage medium having stored thereon computer instructions which, when executed by a computer, perform a method according to either the first or second aspect or any one of possible implementations of the first or second aspect.

[0110] According to an eighth aspect, there is provided a computer program product comprising computer program code which, when run on a computer, performs a method according to the first aspect or the second aspect or any one of possible implementations of the first aspect or the second aspect.

[0111] According to a ninth aspect, there is provided a chip configured to read a computer program stored in a memory to perform a method according to the first aspect or the second aspect or any one of possible implementations of the first or second aspect. Alternatively, the chip includes circuitry configured to perform a method according to the first aspect or the second aspect or any one of possible implementations of the first or second aspect.

[0112] According to a tenth aspect, there is also provided a chip system. The chip system includes a processor configured to assist an apparatus in performing a method according to the first aspect or the second aspect or any one of possible implementations of the first aspect or the second aspect. In a possible design, the chip system further includes a memory configured to store programs and data required by the apparatus. The chip system may include a chip, or may include a chip and other discrete elements.

[0113] According to an eleventh aspect, there is provided a wireless communication system. The system includes a quantization or dequantization apparatus according to the third, fourth, fifth, or sixth aspect. For example, the system may include a first device and a second device, a first device and a third device, or a second device and a third device according to an embodiment of the present application. [Brief explanation of the drawings]

[0114] [Figure 1] FIG. 1 is a diagram of a communication scenario applicable to embodiments of the present application. [Figure 2] 1 is an example of deep learning based channel compression feedback applicable to embodiments of the present application; [Figure 3] FIG. 1 is a diagram of a quantization method applicable to the present application. [Figure 4] FIG. 1 is a diagram of updating the quantization dictionary by using the cluster method. [Figure 5] 1 is an example of a quantization method according to the present application. [Figure 6] 1 is an example of a quantization method according to the present application. [Figure 7] 1 is a schematic block diagram of a quantization or dequantization device according to the present application; [Figure 8] 1 is a diagram of the structure of a quantization or dequantization device according to the present application; DETAILED DESCRIPTION OF THE INVENTION

[0115] The following describes the technical solution of the present application with reference to the accompanying drawings.

[0116] The technical solutions in the embodiments of the present application are applicable to various communication systems, including, but not limited to, the 5th generation (5G) system or new radio (NR) system, long term evolution (LTE) system, long term evolution-advanced (LTE-A) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD) system, etc. The technical solutions are further applicable to future communication systems, for example, 6th generation mobile communication systems. Furthermore, the technical solution may be further applied to device-to-device (D2D) communication, vehicle-to-every (V2X) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), Internet of Things (IoT) communication systems, satellite communication systems, or other communication systems. Furthermore, the technical solution may be further extended to similar wireless communication systems, such as, but not limited to, wireless fidelity (Wi-Fi), worldwide interoperability for microwave access (Wi-MAX), and communication systems related to the 3rd generation partnership project (3GPP).

[0117] A network element in a communication system can transmit signals to or receive signals from another network element. The signals can include information, signaling, data, etc. A network element can also be substituted by an entity, a network entity, a device, a communication device, a communication module, a node, a communication node, etc. In this application, a device is used as an example for explanation.

[0118] A communication system applicable to the present application may include one or more first devices and one or more second devices. Optionally, one of the first device and the second device may be a terminal device, and the other of the first device and the second device may be a network device. Alternatively, both the first device and the second device may be a network device or a terminal device.

[0119] FIG. 1 is a diagram of a communication scenario applicable to an embodiment of the present application. As shown in FIG. 1, this embodiment of the present application is mainly applied to a scenario in which a terminal device sends information to a network device. For example, this embodiment of the present application is applied to a scenario in which a terminal device sends channel state information (CSI) to a network device. In this embodiment of the present application, an example in which a terminal device sends CSI to a network device is used as an example to describe the technical solution of the present application. Based on the example, those skilled in the art may apply the technical solution to other scenarios, such as a scenario in which one device sends information to another device. This is not limited thereto. The number and types of devices in the communication system shown in FIG. 1 are merely used as an example, and the present disclosure is not limited thereto. In a practical application, the communication system may further include more terminal devices and more access network devices, and may further include other network elements, such as core network devices and / or network elements configured to implement artificial intelligence functions.

[0120] For example, a terminal device may also be referred to as user equipment (UE), access terminal, subscriber unit, subscriber station, mobile station, mobile console, mobile terminal (MT), remote station, remote terminal, mobile device, user terminal, terminal, drone, wireless communication device, user agent, or user equipment. A terminal device in this embodiment of the present application may be a device that provides voice and / or data connectivity for a user and may be configured to connect people, things, and machines, such as handheld devices or vehicle-mounted devices, with wireless connectivity capabilities. The terminal device in this embodiment of the present application may be a mobile phone, a tablet computer (pad), a notebook computer, a palmtop computer, a mobile internet device (MID), a wearable device, a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical surgery, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home, etc.

[0121] In this embodiment of the present application, a device configured to implement the functions of a terminal can be a terminal, or can be a device that can help a terminal implement functions, such as a chip system or chip, a hardware circuit, a software module, or a combination of a hardware circuit and a software module. The device can be installed in a terminal or used with a terminal. In this embodiment of the present application, a chip system can include a chip, or may include a chip and another discrete element.

[0122] For example, a network device may be a device equipped with wireless transceiver capabilities. The network device can be a device that provides wireless communication function services and is usually located on the network side, including, but not limited to, a next generation NodeB (gNodeB, gNB) in a fifth generation (5G) communication system, a base station in a sixth generation (6G) mobile communication system, a base station in a future mobile communication system, or an access node in a wireless fidelity (Wi-Fi) system, an evolved NodeB (eNB) in a long term evolution (LTE) system, a radio network controller (RNC), a NodeB (NodeB, NB), a base station controller (BSC), a home base station (e.g., a home evolved NodeB or home NodeB, HNB), a base band unit (BBU), a transmission reception point (TRP), a transmission point (TP), a base transceiver station (BTS), a satellite, a drone, etc. In the network structure, the network device may include a central unit (CU) node or a distributed unit (DU) node, or may be a RAN device including a CU node and a DU node, or a RAN device including a CU control plane node, a CU user plane node, and a DU node. Alternatively, the network device may be a radio controller, a relay station, an in-vehicle device, a wearable device, etc. in a cloud radio access network (CRAN) scenario.Furthermore, the base station can be a macro base station, a micro base station, a relay node, a donor node, or a combination thereof. Alternatively, the base station can be a communication module, a modem, or a chip disposed in the above device or apparatus. Alternatively, the base station can be a mobile switching center, a device having base station functionality in D2D, V2X, and M2M communications, a network side device in a 6G network, a device having base station functionality in a future communication system, etc. The base station can support networks of the same access technology or different access technologies. This is not limited.

[0123] In this embodiment of the present application, a device configured to implement the functionality of a network device can be the network device itself, or can be a device that can help the network device implement the functionality, such as a chip system or chip, a hardware circuit, a software module, or a combination of a hardware circuit and a software module. The device can be installed in the network device or used in conjunction with the network device. In this embodiment of the present application, a chip system can include a chip, or may include a chip and another discrete element.

[0124] With the development of wireless communication technology, the number of supported services is increasing, imposing higher requirements on communication systems in terms of system capacity, communication delay, and other criteria. Communication rates can be effectively improved by expanding the available bandwidth of a UE. To support wider available bandwidth, multiple contiguous or non-contiguous frequency bands can be allocated to a UE for communication. To support effective communication in each frequency band, a base station needs to acquire channel state information (CSI) for each frequency band. However, in communication systems based on frequency division duplex (FDD), which is currently widely used, the uplink and downlink channels are not interdependent, and the base station needs to acquire downlink CSI through uplink feedback from the UE. In other words, the base station transmits a downlink reference signal to the UE, and the UE receives the downlink reference signal. Since the UE knows the transmission information of the downlink reference signal, the UE can estimate (measure) the downlink channel through which the downlink reference signal travels based on the received downlink reference signal. The UE generates CSI based on the measured downlink channel matrix and feeds the CSI back to the base station.

[0125] In an FDD system, an important part of CSI feedback is a precoding matrix indicator (PMI). In other words, bits 0 and 1 in the CSI can be used to quantize the channel matrix or precoding matrix. PMI design (also called codebook design) is a fundamental challenge in mobile communication systems. In traditional codebook design methods, a set of precoding matrices and corresponding numbers are predefined (agreed upon) in a protocol, and these precoding matrices are called codewords. The channel matrix or precoding matrix can be approximated by using a predefined codeword or a linear combination of multiple predefined codewords. Therefore, the UE can feedback one or more of the numbers and weighting coefficients corresponding to the codewords to the base station by using the PMI, so that the base station can recover the channel matrix or precoding matrix. The more accurate the CSI fed back by the UE, the richer the information, the more accurate the channel information recovered by the base station based on the CSI, and the more accurate the precoding matrix determined by the base station. As a result, the better the downlink spatial multiplexing performance, the higher the signal-to-interference and noise ratio received by the UE, and the higher the system capacity. However, since the size of the complete channel matrix is ​​directly proportional to the number of antenna ports, the larger the size of the antenna array in the communication system, the more antenna ports it can support. In a massive MIMO system, the feedback overhead of the UE feeding back the complete channel matrix to the base station through CSI is also very large.

[0126] Machine learning methods have stronger nonlinear feature extraction capabilities and can extract channel matrix features more effectively. Compared with traditional solutions, feedback of the same scale can contain more channel information, reducing information loss in CSI compression feedback and improving channel recovery accuracy at the base station side. Known technologies introduce artificial intelligence (AI) into wireless communication networks, providing a CSI feedback solution based on an AI model. A terminal device compresses and feeds back CSI by using an AI model, and an access network device recovers the compressed CSI by using the AI ​​model. Compared with traditional solutions, this solution requires less feedback to represent the same channel information, further reducing feedback overhead. The AI ​​model may include a neural network.

[0127] FIG. 2 is an example of deep learning-based channel compression feedback applied to an embodiment of the present application. As shown in FIG. 2, at the UE side, channel information is compressed by a deep learning-based encoder, the compressed information is quantized and represented by using a quantization dictionary (the compressed information is quantized and represented by using the element in the quantization dictionary that is closest to the compressed information), and the index of the quantized representation in the quantization dictionary is fed back to the base station by using finite bits. The base station has the same quantization dictionary as the UE side. Therefore, the fed back index can be restored to the quantized representation of the compressed information, and then the quantized representation of the compressed information is input to a deep learning-based decoder to restore the channel information. In the above process, the UE and the base station need to share the same quantization dictionary. However, since the quantization dictionary is usually very large, the overhead of the UE and the base station maintaining the quantization dictionary is also very large.

[0128] Furthermore, when the quantization dictionary needs to be updated due to changes in the environment or configuration, the quantization dictionary of the UE and the quantization dictionary of the base station need to be updated synchronously, and the usage time of the updated quantization dictionary also needs to be synchronized. In many current applications, the encoding process on the UE side, the decoding process on the base station side, and the quantization process use the entire AI model for joint training. After training is completed, the trained AI model is configured on the UE side and the base station side. If the AI ​​model or the quantization process needs to be updated later, the updated AI model also needs to be sent to the base station side or the UE side, and the complete AI model needs to be transmitted every time it is updated, resulting in a very large transmission overhead.

[0129] Therefore, the present application provides a quantization method to reduce the overhead of maintaining a quantization dictionary. Furthermore, when the quantization dictionary is updated, the complete quantization dictionary does not need to be transmitted, but only the updated part needs to be transmitted, thereby reducing the overhead of transmitting the quantization dictionary.

[0130] The following describes the technical solutions provided in this application.

[0131] The quantization methods provided herein apply to any scenario in which information needs to be quantized, for example, one of two devices needs to quantize information and transmit the quantized information to another device, which needs to dequantize the received quantized information.

[0132] Furthermore, in the solution provided herein, the method for quantizing information and the method for dequantizing received information may be separated for use. However, there is no limitation on whether the quantization process and the dequantization process are used together. For example, in many communication scenarios, quantization and dequantization are usually used together. For example, in the communication systems listed above, such as in an FDD communication scenario, the UE needs to feed back channel information, such as CSI information, to a base station. In this scenario, the UE needs to quantize compressed information to be fed back as CSI information and feed back the quantized information to the base station. The base station needs to receive the quantized information from the UE and dequantize the quantized information to finally obtain the CSI information. As another example, in some communication positioning scenarios, one device needs to feed back location information to another device, and the location information may be quantized. The other device needs to receive the quantized information and dequantize it to obtain the location information. Therefore, specific application scenarios are not limited in this application, and the above-listed quantization of channel information, such as CSI information, and location information is used merely as an example.

[0133] FIG. 3 is a diagram of the quantization method applied in this application.

[0134] 310: The first device obtains a first quantization dictionary.

[0135] Optionally, the first quantization dictionary may be initially generated by the first device, or may be obtained by the first device by updating the second quantization dictionary, as is not limited thereto.

[0136] In this application, a quantization dictionary is represented by using the dictionary usage of the quantization dictionary and the dictionary content of each of the M sub-dictionaries included in the quantization dictionary. In other words, the quantization dictionary includes the dictionary usage and the dictionary content. The dictionary content includes the dictionary content of each of the M sub-dictionaries included in the quantization dictionary.

[0137] The dictionary usage of the first quantization dictionary indicates a correspondence between the M sub-dictionaries and the R quantization portions, where the M sub-dictionaries include a first sub-dictionary, and the first sub-dictionary corresponds to the first quantization portion of the R quantization portions, where the first sub-dictionary is used to quantize the first quantization portion, and the first sub-dictionary is any one of the M sub-dictionaries.

[0138] The first quantization dictionary is used to quantize the information to be quantized. Specifically, the first quantization dictionary includes M sub-dictionaries, which are used to quantize R quantized portions of the information to be quantized, where M and R are both positive integers. Alternatively, the first quantization dictionary includes M sub-dictionaries, which correspond to the R quantized portions of the information to be quantized.

[0139] Optionally, M and R are both positive integers greater than one.

[0140] It can be understood that dequantization is the inverse process of quantization. Therefore, the M sub-dictionaries included in the first quantization dictionary are alternatively used to dequantize information obtained by quantizing R quantization portions, where M and R are positive integers greater than 1. The information obtained by quantization, i.e., the quantized information, may be an index of the quantization information corresponding to the quantization portion in the sub-dictionary. For simplicity of explanation, the information obtained by quantization may also be referred to as a "quantization index" in this specification.

[0141] For example, R=M, and specifically, M sub-dictionaries are used to quantize M quantization portions. In this case, the M sub-dictionaries are in one-to-one correspondence with the M quantization portions, and each of the M sub-dictionaries is used to quantize a corresponding quantization portion, as shown in Table 1 as an example. [Table 1]

[0142] As another example, R is greater than M. In this case, some of the R quantization portions may correspond to the same sub-dictionary among the M sub-dictionaries, as shown in Table 2 as an example. [Table 2]

[0143] As described above, the dictionary usage of the first quantization dictionary indicates the correspondence between the M sub-dictionaries and the R quantization portions of the information to be quantized, and each of the M sub-dictionaries is used to quantize a corresponding quantization portion. For example, the M sub-dictionaries include the first sub-dictionary. The first sub-dictionary corresponds to a first quantization portion of the R quantization portions, and the first sub-dictionary is used to quantize the first quantization portion.

[0144] For example, in Table 1, the first quantization dictionary includes three sub-dictionaries: Y1, Y2, and Y3. The quantization target information includes three parts: quantization part 1, quantization part 2, and quantization part 3. Therefore, Y1 corresponds to quantization part 1 of the three quantization parts. Y2 corresponds to quantization part 2 of the three quantization parts. Y3 corresponds to quantization part 3 of the three quantization parts. It can be seen that the dictionary usage indicates the quantization part that corresponds to the sub-dictionary.

[0145] As another example, in Table 2, the first quantization dictionary includes two sub-dictionaries: Y1 and Y2, and the quantization target information includes three quantization portions. The sub-dictionary Y1 corresponds to quantization portion 1 and quantization portion 3 of the three quantization portions. Therefore, the sub-dictionary Y1 is used to quantize quantization portion 1 and quantization portion 3 of the quantization target information.

[0146] The dictionary content of each of the M sub-dictionaries includes at least one dictionary element. Taking the first sub-dictionary as an example, at least one dictionary element included in the dictionary content of the first sub-dictionary is a candidate for quantization information of the first quantization portion. In other words, the dictionary elements included in the dictionary content of the first sub-dictionary are candidates for quantization information of the quantization portion corresponding to the first sub-dictionary. For example, in Table 1, each dictionary element included in the dictionary content of Y1 is a candidate for quantization information of quantization portion 1, and each dictionary element included in the dictionary content of Y2 is a candidate for quantization information of quantization portion 2. As another example, in Table 2, Y1 corresponds to quantization part 1 and quantization part 3, so each dictionary element in the dictionary content of Y1 is not only a candidate for quantization information for the quantization part, but also a candidate for quantization information for quantization part 3.

[0147] The first quantization dictionary includes M sub-dictionaries, and the dictionary content of each of the M sub-dictionaries includes at least one dictionary element. Furthermore, each dictionary element corresponds to one piece of quantization information and one index. Optionally, the "quantization information" corresponding to each dictionary element can alternatively be referred to as a "quantization representation." Therefore, in other words, when the first quantization portion of the information to be quantized corresponds to the first sub-dictionary among the M sub-dictionaries, the quantization representations corresponding to each dictionary element in the first sub-dictionary are candidates for the quantization representation of the first quantization portion, as shown in the example of Table 3. [Table 3]

[0148] As shown in Table 3, the dictionary content of sub-dictionary Y1 includes four dictionary elements: a1, a2, a3, and a4. The dictionary content of sub-dictionary Y2 includes eight dictionary elements: b1, b2, ..., and b8. The dictionary content of sub-dictionary Y3 includes two dictionary elements: c1 and c2. From the above description of the dictionary elements, it can be seen that dictionary elements a1, a2, a3, and a4 are all candidates for quantization information of quantization portion 1. Alternatively, when quantization portion 1 of the quantization target information needs to be quantized, dictionary elements can be selected from dictionary elements a1, a2, a3, and a4 in Y1 as quantization information of quantization portion 1. When the quantization information of the quantization portion is selected, the quantization information is selected according to a specific rule. For example, a dictionary element with the closest Euclidean distance to quantization portion 1 may be selected from Y1 as the quantization information of quantization portion 1. If dictionary element a3 has the closest Euclidean distance to quantization portion 1, it is assumed that dictionary element a3 is the quantization information of quantization portion 1, that is, in other words, dictionary element a3 is a quantized representation of quantization portion 1.

[0149] Specific examples are shown below.

[0150] It is assumed that the information to be quantized is a real vector with a length of N. If the real vector is divided into multiple quantization sections without repetition or omission, the multiple quantization sections can be quantized individually. For example, a quantization section can be expressed in the form [x1, x2). x1 represents the index of the start point of a quantization section, and x2 represents the index of the start point of the next quantization section. For example, x1 represents the index of the start point of quantization section 1, and x2 represents the index of the start point of quantization section 2. Therefore, the dictionary usage of the quantization dictionary can be shown in Table 4. As shown in Table 4, the quantization section [x1, x2) of the encoder output information corresponds to the sub-dictionary Y1, and Y1 is used to quantize the quantization section [x1, x2). The quantized portion [x2, x3) corresponds to sub-dictionary Y2, which is used to quantize the quantized portion [x2, x3), and so on. [Table 4]

[0151] As an example, it is assumed that the information to be quantized is a real vector with length N of 8, and that the real vector is divided into four quantized parts of equal length with no repetitions or omissions. For the quantization of a real vector, a possible dictionary usage is shown in Table 5. [Table 5]

[0152] Table 5 shows that a real vector of length N=8 is divided into four equal-length quantized portions, and the first quantized portion (the 0th and 1st values ​​of the real vector) and the third quantized portion (the 4th and 5th values ​​of the real vector) are quantized by using the sub-dictionary identified as Y1. The second quantized portion (the 2nd and 3rd values ​​of the real vector) and the fourth quantized portion (the 6th and 7th values ​​of the real vector) are quantized by using the sub-dictionary identified as Y2.

[0153] For example, it is assumed that the quantization target information is a real vector with a length N of 8, specifically, [0.6, 0.2, -0.7, -0.1, 0.3, 0.8, -0.1, -0.2]. The real vector is divided into quantization parts according to Table 5, and the quantization parts are individually quantized. Specifically, the first quantization part (the 0th and 1st values, i.e., [0.6, 0.2]) and the third quantization part (the 4th and 5th values, i.e., [0.3, 0.8]) are quantized using the quantization dictionary Y1, and the second quantization part (the 2nd and 3rd values, i.e., [-0.7, -0.1]) and the fourth quantization part (the 6th and 7th values, i.e., [-0.1, -0.2]) are quantized using the quantization dictionary Y2.

[0154] It is assumed that the sub-dictionary Y1 includes four dictionary elements, each of which corresponds to one piece of quantization information. The four pieces of quantization information corresponding to the four dictionary elements are a1=[0.1,0.1], a2=[0.1,0.6], a3=[0.6,0.1], and a4=[0.6,0.6], and the indices corresponding to the four dictionary elements in the sub-dictionary Y1 are (0,1,2,3), in order.

[0155] It is assumed that sub-dictionary Y2 includes four dictionary elements, and the quantization information corresponding to the four dictionary elements is, in order, b1 = [-0.1, -0.1], b2 = [-0.1, -0.6], b3 = [-0.6, -0.1], and b4 = [-0.6, -0.6]. The indices corresponding to the four dictionary elements in sub-dictionary Y2 are, in order, (0, 1, 2, 3).

[0156] Therefore, based on Table 5 and the dictionary contents of the sub-dictionaries Y1 and Y2, the quantization information of the four quantization parts of the real vector [0.6, 0.2, -0.7, -0.1, 0.3, 0.8, -0.1, -0.2] can be obtained as [0.6, 0.1], [-0.6, -0.1], [0.1, 0.6], and [-0.1, -0.1]. The four quantization information correspond to index 2 of Y1, index 2 of Y2, index 1 of Y1, and index 0 of Y2, respectively. The indexes 2, 2, 1, and 0 are fed back to the base station, so that the base station can search for Table 5 from the corresponding sub-dictionary and obtain the quantization information of the four quantization parts, which are [0.6, 0.1], [-0.6, -0.1], [0.1, 0.6], and [-0.1, -0.1].

[0157] For this example, the quantization error is the difference between the quantized information [0.6,0.1], [-0.6,-0.1], [0.1,0.6], and [-0.1,-0.1] and the pre-quantization information [0.6,0.2,-0.7,-0.1,0.3,0.8,-0.1,-0.2]. As an example, a method for calculating the quantization error is to calculate the mean square error of the two pieces of information, i.e., quantization error = ((0.6-0.6) 2 +(0.1-0.2) 2 +(-0.6+0.7) 2 +(-0.1+0.1) 2 +(0.1-0.3) 2 +(0.6-0.8) 2 +(-0.1+0.1) 2 +(-0.1+0.2) 2 ) / 8.

[0158] For each quantization information, the quantization error can be calculated based on the above steps, and the average value of the multiple quantization errors can be used as a criterion for evaluating the quantization performance of the quantization dictionary. If the average quantization error exceeds a threshold, it can be considered that the quantization dictionary cannot properly quantize the information to be quantized. In contrast, if the average quantization error is not greater than (equal to or less than) the threshold, it can be considered that the quantization dictionary can properly quantize the information to be quantized, and the quantization performance can meet the quantization requirements.

[0159] In this application, the first quantization dictionary includes M sub-dictionaries, and each sub-dictionary among the M sub-dictionaries can be determined by using two parameters (K, B). K indicates that the sub-dictionary is used to quantize a quantization portion having a length of K (i.e., the quantization portion includes K vectors), and B indicates that B bits are used to quantize each input of the quantization portion corresponding to the sub-dictionary. Different values ​​of the B bits can correspond to different candidate quantization information. In other words, different values ​​of the B bits are indices of different candidate quantization information. A sub-dictionary with parameters (K, B) can have up to 2 B It can be understood that a sub-dictionary Y1 may contain dictionary elements. Note that for different sub-dictionaries, the values ​​of K and B may be the same or different. For example, sub-dictionary Y1 may be determined by (K1, B1), and sub-dictionary Y2 may be determined by (K2, B2). [Table 6]

[0160] When a sub-dictionary is used to perform vector quantization, for a sub-dictionary with parameters (K, B), for example, the sub-dictionary can have a maximum of 2 BThe sub-dictionary may include dictionary elements (elements for short), each of which corresponds to a real vector of length K and an index. The real vector corresponding to each element is a candidate for quantization information for the quantization part corresponding to the sub-dictionary. Different elements in the sub-dictionary correspond to different indexes. The indexes are used to uniquely identify elements in the sub-dictionary. For example, the quantization target information is the output information of the encoder of the UE. It is assumed that the quantization part of length K, which is the output information of the encoder, is quantized by using a sub-dictionary with parameters (K, B). At one output of the encoder, the value of the quantization part of length K is represented as o1, and the element closest to o1 in distance (which can be said to be Euclidean distance) is found from the sub-dictionary as an approximation of o1. In this case, a real vector having a length of K corresponding to an element is the quantization information of o1, and an index corresponding to an element can be used to uniquely identify the element in the sub-dictionary, so that the index can also uniquely correspond to the quantization information of o1. Based on the index of the quantization information, the base station can find a corresponding real vector having a length of K from the same sub-dictionary as an approximation of o1.

[0161] Optionally, if the dictionary is used to implement other quantization methods, the sub-dictionary with parameters (K, B) may be stored in the form of a code or model, and the code or model includes a quantization function. The input of the quantization function is a real vector to be quantized with a length of K, and the output of the quantization function is a binary bit string of B bits. The quantization function corresponds to an inverse quantization function. The input of the inverse quantization function is a binary bit string of B bits, and the output of the inverse quantization function is a real vector of a length of K. In this application, inverse quantization and dequantization are interchangeable.

[0162] The above describes in detail the representation format of the quantization dictionary in this application.

[0163] It can be understood that the above table is a set of some correspondences, and the correspondences may alternatively be expressed in other formats, such as strings or functions, which is not limited here.

[0164] Optionally, in step 310, the first device obtaining the first quantization dictionary may include multiple possible implementations.

[0165] Implementation 1 The first device updates the second quantization dictionary to obtain the first quantization dictionary.

[0166] It can be understood that the second quantization dictionary may be a quantization dictionary used by the first device and the second device before the first device acquires the first quantization dictionary by updating, and used to quantize or dequantize information exchanged between the first device and the second device.

[0167] As described above, the first device and the second device use the same quantization dictionary, so when the quantization dictionary is obtained by the first device through an update, the first device needs to send the first quantization dictionary obtained through the update to the second device.

[0168] Optionally, the first device and the second device may separately maintain multiple sub-dictionaries for quantization in different application scenarios (also called quantization scenarios), or the first device and the second device may be used to quantize different quantization portions in the information to be quantized. For example, the first device may maintain a quantization dictionary table, which includes one or more candidate sub-dictionaries. The first device updates the second quantization dictionary to obtain the first quantization dictionary. Optionally, compared with the second quantization dictionary, the first quantization dictionary may have updated dictionary usage, updated dictionary content of sub-dictionaries corresponding to one or more quantization portions of the information to be quantized, or updated both dictionary usage and dictionary content in the first quantization dictionary. It should be understood that updating the second quantization dictionary to the first quantization dictionary may include changing the number of quantization portions of the quantization target information, for example, increasing the number of quantization portions or decreasing the number of quantization portions. Alternatively, the number of quantization portions does not change. This is not a limitation. Updating the number of quantization portions is also included in updating the dictionary usage or updating the dictionary content.

[0169] For example, the first device may select one or more candidate sub-dictionaries from the candidate sub-dictionaries in the quantization dictionary table as some of the M sub-dictionaries included in the first quantization dictionary obtained by updating. Alternatively, if none of the candidate sub-dictionaries included in the quantization dictionary table is suitable for the current application, the first device may perform quantization dictionary training to obtain a new sub-dictionary. In the present application, the sub-dictionary obtained by the first device through quantization dictionary training is not included by default in the quantization dictionary table held by the first device. Alternatively, the first device obtains the new sub-dictionary through training. The first device may then add the new sub-dictionary to the quantization dictionary table. Optionally, the first device may assign a new identifier to the new sub-dictionary to distinguish it from existing sub-dictionaries in the quantization dictionary table. For example, if the quantization dictionary table holds sub-dictionaries Y1, Y2, and Y3, the first device assigns identifier Y4 to the new sub-dictionary obtained by training and adds Y4 to the quantization dictionary table. In this case, the quantization dictionary table includes sub-dictionaries Y1, Y2, Y3, and Y4. Optionally, the first device may alternatively assign an existing identifier to the new sub-dictionary, or may replace an existing sub-dictionary with the new sub-dictionary. For example, the first device assigns identifier Y3 to the new sub-dictionary obtained by training. Thus, the sub-dictionaries held in the quantization dictionary table are the existing sub-dictionaries Y1 and Y2 and the new sub-dictionary obtained by training (identified as Y3).

[0170] Optionally, the first quantization dictionary obtained by the first device includes M sub-dictionaries, and the M sub-dictionaries include P sub-dictionaries and Q sub-dictionaries, where the P sub-dictionaries are selected by the first device from a quantization dictionary table maintained in the first device, the quantization dictionary table includes at least P sub-dictionaries, and the Q sub-dictionaries are obtained by the first device by performing quantization dictionary training, where 0≦P≦M, 0≦Q≦M, P+Q=M, and P and Q are integers.

[0171] It should be understood that when P=0, it indicates that all M sub-dictionaries included in the first quantization dictionary are obtained by the first device through quantization dictionary training. When Q=0, it indicates that all M sub-dictionaries included in the first quantization dictionary are selected by the first device from the quantization dictionary table. When P>0 and Q>0, it indicates that P sub-dictionaries of the M sub-dictionaries included in the first quantization dictionary are selected from candidate sub-dictionaries in the quantization dictionary table, and the remaining Q sub-dictionaries are obtained by the first device through training.

[0172] After obtaining the first quantization dictionary through updating, the first device needs to send the first quantization dictionary to the second device. According to the above description of the format of the quantization dictionary, transmitting the complete quantization dictionary includes transmitting the dictionary usage and the dictionary content of the quantization dictionary. Because the dictionary content is usually large and requires a large transmission overhead, the number of times the complete dictionary is transmitted after the dictionary is updated should be reduced as much as possible.

[0173] Based on the representation format of the quantization dictionary defined in this specification, the quantization dictionary can be updated in the following several scenarios: In different scenarios, different transmission methods of the updated quantization dictionary can be used, and the transmission overhead is reduced.

[0174] (1) The dictionary usage of the quantization dictionary is updated, but the dictionary content of the quantization dictionary is not updated.

[0175] For example, in the AI ​​model deployment phase, multiple sub-dictionaries may be deployed on the first device and the second device. The multiple sub-dictionaries may be used to adapt to quantization requirements in different scenarios or to adapt to quantization requirements for different quantization portions of the information to be quantized. When a scenario changes and the AI ​​model / quantization dictionary needs to be updated, the first device may select a sub-dictionary from the multiple existing sub-dictionaries that is more suitable for the current scenario for the quantization or dequantization process. For example, multiple sub-dictionaries may be deployed on the first device and the second device to satisfy the quantization requirements of a low-speed scenario and a high-speed scenario, respectively. When the user speed changes, only the dictionary usage of the quantization dictionary needs to be updated. In this scenario, the first device does not need to transmit the dictionary content of the updated quantization dictionary, but only needs to transmit the updated part of the dictionary usage of the updated quantization dictionary.

[0176] As an example, assume that sub-dictionaries Y1, Y2, Y3, and Y4 are deployed on a first device and a second device. Assume that the second quantization dictionary before the update is shown in Table 6, and the first quantization dictionary after the update is shown in Table 7. [Table 7]

[0177] In this scenario, it can be seen that the update of the quantization dictionary can be only an update of the dictionary usage, that is, the correspondence between the quantization portion of the quantization target information and the sub-dictionaries is updated. Optionally, the correspondence indicated by using the dictionary usage of the first quantization dictionary can be partially or completely changed compared to the correspondence indicated by using the dictionary usage of the second quantization dictionary. This is not limiting. An example of a partially changed correspondence is used in Table 7. For example, the correspondence between quantization portion 1 and sub-dictionary Y1 is not changed, and only the correspondence between quantization portion 2 and quantization portion 3 and the sub-dictionaries is changed. However, since Y3 and Y4 are existing sub-dictionaries stored in the quantization dictionary table, the update of the quantization dictionary can be only an update of the dictionary usage. The first device can indicate the updated portion of the quantization dictionary to the second device. As an example, a first device sends a first message to a second device. The first message indicates an updated portion of the first quantization dictionary compared to the second quantization dictionary. Tables 6 and 7 are used as examples. The first message indicates the correspondence between quantization portion 2 and sub-dictionary Y3, and the correspondence between quantization portion 3 and sub-dictionary Y4. For example, the first message includes an identifier of sub-dictionary Y3 and quantization portion 2, and an identifier of sub-dictionary Y4 and quantization portion 3. The identifier of Y3 corresponds to quantization portion 2, and the identifier of Y4 corresponds to quantization portion 3.

[0178] As another example, the quantization dictionary tables of the first and second devices hold sub-dictionaries Y1, Y2, Y3, Y4, and Y5. The second quantization dictionary before updating is shown in Table 6, and the first quantization dictionary after updating is shown in Table 8. [Table 8]

[0179] In this case, the dictionary usage of the quantization dictionary is updated. Specifically, the correspondence between three quantization parts (quantization part 1, quantization part 2, and quantization part 3) and three sub-dictionaries (Y1, Y2, Y3) indicated by using the second quantization dictionary is updated to a correspondence between four quantization parts (quantization part 1 to quantization part 4) and four sub-dictionaries (Y1, Y2, Y3, Y4). It can be seen that when the dictionary usage is updated, the quantization target information is subdivided into multiple quantization parts, and the number of quantization parts increases. In this example, the first message may include identifiers of Y1, Y3, Y4, and Y5, and quantization parts corresponding to Y1, Y3, Y4, and Y5, respectively.

[0180] (2) The dictionary usage of the quantization dictionary is not updated, and the dictionary content of the quantization dictionary is updated.

[0181] In some cases, when the application scenario changes and the AI ​​model or the quantization dictionary needs to be updated, if the existing sub-dictionary stored in the quantization dictionary table cannot adapt to the current application scenario, the first device needs to train the quantization dictionary. It should be understood that if the existing sub-dictionary cannot adapt to the current application scenario, this indicates that neither the sub-dictionary before the update nor the sub-dictionary in the quantization dictionary table is applicable to the current application scenario, and the first device needs to obtain a new quantization dictionary by retraining.

[0182] As an example, a method for training and updating a quantization dictionary is to update the quantization dictionary by using a clustering method. Figure 4 is a diagram of updating a quantization dictionary by using a clustering method. The quantization dictionary shown in Table 5 is used as an example. The quantization dictionary includes four sub-dictionaries, and four equal-length quantization parts are obtained by dividing a real vector (an example of information to be quantized) with a length K of 8. For a quantization part [0,2), it is assumed that the value distribution of the quantization part [0,2] in the information to be quantized is shown in the left diagram of Figure 4. When 2 bits (B=2) are used to quantize the quantization part, the value space of the quantization part is 2. B The sub-dictionary can be divided into 4 blocks, and the center position of the value part in each block is used as a dictionary element to obtain the right diagram of FIG. 4. In this way, the value of the dictionary element in the sub-dictionary can be very close to the quantization target information. Only one quantization part of the quantization target information is used here as an example for explanation. The update of the sub-dictionaries corresponding to the remaining three quantization parts is similar, and the details will not be described again.

[0183] Therefore, if the dictionary usage of the quantization dictionary remains unchanged, the dictionary elements may be updated based on the distribution of the quantization target information. Here, updating the dictionary elements based on the distribution of the quantization target information is simply an implementation of updating the dictionary content. If the value distribution of only some quantized portions of the quantization target information changes, only the dictionary content of the sub-dictionary corresponding to the quantized portions whose value distribution changes may be updated.

[0184] (3) The dictionary usage of the quantization dictionary is updated and the dictionary content may also be updated.

[0185] When an application scenario changes and the AI ​​model and / or quantization dictionary is updated, if all existing sub-dictionaries in the quantization dictionary table cannot adapt to the new application scenario, the quantization dictionary may be retrained.

[0186] In this case, when a quantization dictionary is updated, both the dictionary usage of the quantization dictionary and the dictionary content can be considered to be updated. Optionally, the dictionary usage update can include re-division of the quantization part and updating of the identifiers of the sub-dictionaries corresponding to each quantization dictionary.

[0187] In implementation 1, to update the quantization dictionary, the first device selects a sub-dictionary from the quantization dictionary table, and / or the first device performs dictionary retraining to obtain a new sub-dictionary so as to obtain M sub-dictionaries included in the first quantization dictionary obtained by updating.

[0188] When the first device updates the quantization dictionary, optionally, the first device sends first information to the second device before updating. The first information indicates that the first device will update the quantization dictionary. After obtaining the first quantization dictionary through the update, the first device sends the first quantization dictionary to the second device. Optionally, in another implementation, the first device and the second device may agree that the quantization dictionary will be updated by the first device or the second device. In this implementation, before the quantization dictionary is updated, the first device and the second device do not need to exchange information about who will update the quantization dictionary, and after the agreeing party completes the quantization dictionary update, the updated quantization dictionary is sent directly to the other party.

[0189] The above has described several possibilities for updating a quantization dictionary. After the update is completed, when the first device sends the first quantization dictionary to the second device, the first device sends the second device an updated portion of the first quantization dictionary compared with the second quantization dictionary. Specifically, the updated portion is sent in the form of dictionary usage or dictionary content. For example, the first device sends a first message to the second device. The first message indicates the updated portion. Optionally, there may be one or more first messages. This is not limited.

[0190] For example, the first quantization dictionary includes M sub-dictionaries. If the update of the quantization dictionary is only an update of the dictionary usage, the updated portion includes correspondences between L quantization portions of the R quantization portions and the sub-dictionaries, where 1≦L≦R, and L is an integer. When the first device notifies the second device of the updated portion, for example, the first device sends a first message to the second device. The first message indicates the L quantization portions and the correspondences between the L quantization portions and the sub-dictionaries. It should be understood that the L quantization portions are quantization portions whose correspondences with the sub-dictionaries change.

[0191] If the update of the quantization dictionary is only an update of dictionary content, the updated portion includes dictionary content of sub-dictionaries corresponding to Z quantization portions of the R quantization portions, where 1≦Z≦R and Z is an integer. When the first device notifies the second device of the updated portion, the first device sends a first message to the second device. The first message indicates identifiers of the Z sub-dictionaries and updated dictionary elements and indexes of the Z sub-dictionaries. The Z quantization portions are quantization portions for which the dictionary content of the corresponding sub-dictionaries changes.

[0192] Therefore, when receiving the first quantization dictionary from the first device, the second device can specifically receive the dictionary usage and dictionary content of the first quantization dictionary, which are updated dictionary usage and updated dictionary content, thereby updating the second quantization dictionary synchronously with the first quantization dictionary based on the second quantization dictionary.

[0193] Optionally, after obtaining the first quantization dictionary, the second device sends second information to the first device. The second information is used to request that the first quantization dictionary be enabled. After receiving the second information, the first device sends third information to the second device. The third information instructs the first quantization dictionary to be enabled. Optionally, the third information may carry a time interval T. If the time error between the moment the first device sends the third information and the moment the second device receives the third information is ignored, the first device enables the first quantization dictionary after the time interval T from the moment the third information is sent, and the second device starts to enable a new quantization dictionary, i.e., the first quantization dictionary, after the time interval T from the moment the third information is received. Optionally, the time interval T may also be agreed upon in the protocol. By default, both parties enable the new quantization dictionary after the time interval T from the moment the second device receives the third information.

[0194] Further, optionally, in the process of receiving a new quantization dictionary from the first device, the second device parses the received first quantization dictionary according to the representation format of the quantization dictionary to determine whether the received first quantization dictionary is complete, specifically, whether both the dictionary content and dictionary usage of the first quantization dictionary are intact. If it is determined that the received first quantization dictionary has a defective portion, the second device may send fourth information to the first device. The fourth information instructs the first device to resend the defective portion. After receiving the defective portion resent by the first device, the second device may obtain a complete first quantization dictionary.

[0195] Implementation 2 The first device receives the first quantization dictionary from the second device.

[0196] Optionally, in implementation 2, the first quantization dictionary may be initially generated by the second device or may be obtained by updating the second quantization dictionary. After the second device obtains the first quantization dictionary, the second device sends the first quantization dictionary to the first device.

[0197] The above describes in detail the process in which the first device updates the quantization dictionary, and the process in which the first device sends the new quantization dictionary to the second device after the first device obtains the new quantization dictionary through updating. It can be understood that in Implementation 2 in which the second device updates the quantization dictionary, the process or operation performed by the second device can be considered to be the same as the process or operation performed by the first device in Implementation 1. To avoid repetition, the details will not be described again.

[0198] Optionally, in Implementation 2, before updating the quantization dictionary, the second device sends fifth information to the first device. The fifth information indicates that the second device will update the quantization dictionary. Similarly, after the first device obtains a new quantization dictionary from the second device, the first device sends sixth information to the second device. The sixth information indicates that the new quantization dictionary is requested to be enabled. Thereafter, the second device sends seventh information to the first device based on the sixth information. The seventh information instructs the first device to enable the new quantization dictionary. Optionally, when a time error between the moment the second device sends the seventh information and the moment the first device receives the seventh information is taken into account, the first device enables the new quantization dictionary by default after a time interval T1 from the moment the seventh information is received, and the second device enables the new quantization dictionary by default after a time interval T2 from the moment the seventh information is sent. Considering that there is a time difference between the moment the second device sends the seventh information and the moment the first device receives the seventh information, it should be understood that the first device and the second device enable the new quantization dictionary after different time intervals to ensure that the first device and the second device enable the new quantization dictionary synchronously. For example, the above time difference can be taken into account in the design of T1 and T2, so that the first device and the second device enable the new quantization dictionary simultaneously. Alternatively, the second device indicates the time interval T1 in the seventh information, so that the first device determines the moment to enable the new quantization dictionary based on the moment the seventh information is received and the time interval T1.

[0199] Optionally, in implementation 2 in which the first device receives a new quantization dictionary from the second device, after the first device receives the first quantization dictionary and before the first device sends the sixth information used to request that the first quantization dictionary be enabled, the first device may evaluate the quantization performance of the first quantization dictionary.

[0200] For example, the first device inputs pre-stored channel information of a test set to an encoder, quantizes output information of the encoder by using a first quantization dictionary, and calculates a quantization error to evaluate the quantization performance of the first quantization dictionary. If the quantization performance satisfies the quantization requirement, for example, if the quantization error is equal to or less than a threshold, the quantization requirement is deemed to be satisfied, and the first device sends sixth information to the second device to request that the first quantization dictionary be enabled.

[0201] Optionally, for example, in Implementation 2, the first device may further update the encoder (the first device is a UE, and the second device is an access network device) or the decoder (the first device is an access network device, and the second device is a UE) based on the new quantization dictionary. For example, the first device is an access network device. If the access network device determines that the decoder does not match the first quantization dictionary obtained from the UE, the access network device updates the decoder based on the first quantization dictionary, for example, updating the decoder parameters. For example, the access network device inputs quantization information in the quantization dictionary as candidates to the decoder to obtain output information, calculates the difference between the probability distribution of the output information and the probability distribution of channel information in the training set, for example, mutual information, and updates the decoder by using the difference as a loss function. For example, the first device is a UE. If the UE determines that the encoder output does not match the first quantization dictionary obtained from the access network device, the UE updates the encoder based on the first quantization dictionary, for example, updating the encoder parameters. The UE uses the training set, inputs the channel data in the training set to the encoder to obtain an output, and quantizes the encoder output by using the new dictionary to calculate an average quantization error for the training set, which is used as a loss function. The loss function is a function of the encoder weights, which is an AI model. If the loss function is less than a threshold, training is stopped; otherwise, the encoder weights are updated to reduce the loss function; for example, the gradient of the loss function, which is the quantization error, with respect to the encoder weights is calculated, and the encoder is updated by using gradient descent. If the encoder is implemented by a neural network, the encoder weights may include the weights of each neuron in the neural network.

[0202] 320: The first device quantizes the quantization target information by using the first quantization dictionary, or the first device dequantizes the received information obtained by quantizing the R quantization parts by using the first quantization dictionary.

[0203] In step 310, the first device obtains a first quantization dictionary. Then, the first device performs quantization by using the first quantization dictionary, or performs inverse quantization by using the first quantization dictionary. For example, the information obtained by quantizing the R quantization parts and received by the first device may be R quantization indexes.

[0204] For example, in a scenario in which a UE reports CSI to a network side, for a terminal device reporting CSI to the network side, a first quantization dictionary is used to quantize the compressed information of the CSI, and then the UE reports the quantized information to the network side. Specifically, the quantization information is an index of the quantization information corresponding to the compressed information in a corresponding sub-dictionary (e.g., a quantization index for short). For the network side, the network side receives the index of the quantization information (i.e., the quantization index) reported by the terminal device, determines the information corresponding to the index based on the first quantization dictionary, recovers the quantization information corresponding to the compressed information of the CSI, decodes the compressed information, and finally obtains the channel information. Therefore, for example, if the first device is a UE, the UE quantizes the compressed information of the channel information by using the first quantization dictionary. When the first device is an access network device, the access network device dequantizes the received quantized information, for example, an index corresponding to the quantized information, by using the first quantization dictionary to recover the compressed information of the channel information.

[0205] Furthermore, it can be understood that in the process of updating the quantization dictionary, before the new first quantization dictionary is enabled, the first device continues to use the second quantization dictionary to quantize the quantization target information, and the second device continues to use the second quantization dictionary to dequantize the received quantized information, for example, the index corresponding to the quantized information.

[0206] For example, when the first device is an access network device, before the first quantization dictionary is enabled, the access network device receives first quantized information from the UE, and the access network device dequantizes the first quantized information by using the second quantization dictionary. After the first quantization dictionary is enabled, when the access network device receives second quantized information from the UE, the access network device dequantizes the second quantized information by using the first quantization dictionary.

[0207] If the first device is a UE, before the first quantization dictionary is enabled, if the UE needs to feed back first channel information to the access network device, the UE quantizes the first channel information by using the second quantization dictionary and sends the quantized information of the first channel information. After the first quantization dictionary is enabled, if the UE needs to feed back channel information to the access network device, the UE quantizes the second channel information by using the first quantization dictionary and sends the quantized information of the second channel information.

[0208] The above has described in detail the representation format of the quantization dictionary provided in the present application and the process by which a device obtains the quantization dictionary. The following provides two examples of the process of updating the quantization dictionary with reference to Figures 5 and 6. Note that Figures 5 and 6 are intended to provide a complete example description of the procedure or mechanism for updating the quantization dictionary. For a detailed description of the relevant information or steps, please refer to the relevant description above. Details will not be described in this section.

[0209] In the following examples of updating a quantization dictionary, it is assumed that the quantization dictionary initially used or previously used by the base station and the UE is the second quantization dictionary.

[0210] FIG. 5 is an example of a quantization method according to the present application.

[0211] 501: Optionally, a base station sends first information to a UE, where the first information indicates that the base station updates a quantization dictionary.

[0212] 502: The base station updates the second quantization dictionary to obtain the first quantization dictionary.

[0213] 503: The base station sends a first message to the UE, where the first message indicates an updated portion of the quantization dictionary.

[0214] 504: The UE receives the updated portion of the quantization dictionary.

[0215] 505: The UE parses the updated portion of the quantization dictionary to obtain a first quantization dictionary.

[0216] 506: Optionally, when the UE determines that the received first quantization dictionary has a defective portion according to the representation format of the quantization dictionary, the UE sends fourth information to the base station, where the fourth information is used to request the base station to retransmit the defective portion of the first quantization dictionary.

[0217] 507: Optionally, the base station retransmits the defective portion to the UE.

[0218] After receiving the defective portion, the UE obtains the complete first quantization dictionary.

[0219] It should be understood that steps 506 and 507 are interrelated steps, and that method 500 may include both steps 506 and 507 or may not include steps 506 and 507 .

[0220] 508: Optionally, the UE evaluates the quantization performance of the first quantization dictionary.

[0221] After evaluating the quantization performance of the first quantization dictionary, if the quantization requirements are met, the UE may request that the first quantization dictionary be enabled, as shown in step 510.

[0222] Optionally, before requesting that the first quantization dictionary be enabled, the UE may alternatively test the degree of match between the first quantization dictionary and the encoder and update the encoder based on the first quantization dictionary, as shown in step 509.

[0223] 509: Optionally, if it is determined that the output of the encoder does not match the first quantization dictionary, the UE updates the encoder based on the first quantization dictionary.

[0224] 510: Optionally, the UE sends second information to the base station, where the second information is used to request that the first quantization dictionary be enabled.

[0225] The base station receives second information from the UE.

[0226] 511: The base station sends third information to the UE, where the third information instructs the UE to enable the first quantization dictionary.

[0227] The UE receives the third information from the base station.

[0228] 512: After the first quantization dictionary is enabled, the base station dequantizes the received quantized information by using the first quantization dictionary, and the UE quantizes the quantization target information by using the first quantization dictionary.

[0229] Optionally, by way of example and not limitation, the information that needs to be quantized at the UE side may be CSI or other information that needs to be reported to the base station.

[0230] The procedure shown in Figure 5 is an example of updating the quantization dictionary by the base station. The following further provides an example of a procedure for updating the quantization dictionary by the UE, as shown in Figure 6.

[0231] FIG. 6 is an example of a quantization method according to the present application.

[0232] 601: Optionally, the UE sends first information to the base station, where the first information indicates that the UE updates a quantization dictionary.

[0233] 602: The UE updates the second quantization dictionary to obtain the first quantization dictionary.

[0234] 603: The UE sends a first message to the base station, where the first message indicates an updated portion of the first quantization dictionary compared with the second quantization dictionary.

[0235] 604: The base station receives the updated portion.

[0236] 605: The base station parses the updated portion to obtain a first quantization dictionary.

[0237] Optionally, if the base station determines that there is a defective portion in the first quantization dictionary, the base station may request the UE to retransmit the defective portion. After receiving the defective portion, the base station obtains a complete first quantization dictionary.

[0238] Optionally, the base station may evaluate the quantization performance of the first quantization dictionary, and when the evaluation is completed and the quantization requirements are satisfied, the base station requests the UE to enable the first quantization dictionary, as shown in step 606.

[0239] Optionally, before requesting that the first quantization dictionary be enabled, the base station may alternatively determine a degree of compatibility between the decoder and the first dictionary, and if the decoder does not match the first quantization dictionary, the base station updates the decoder based on the first quantization dictionary, for example, updates parameters of the decoder.

[0240] 606: Optionally, the base station sends second information to the UE, where the second information is used to request enabling the first quantization dictionary.

[0241] The UE receives the second information, and in response to the second information, the UE sends third information to the base station.

[0242] 607: Optionally, the base station receives third information from the UE, where the third information instructs the UE to enable the first quantization dictionary.

[0243] 608: After the first quantization dictionary is enabled, the base station dequantizes the received quantized information by using the first quantization dictionary, and the UE quantizes the quantization target information by using the first quantization dictionary.

[0244] The above is described by using an example in which one of the quantization device (e.g., the first device) and the inverse quantization device (e.g., the second device) initially generates and / or updates the quantization dictionary. It can be understood that the initially generated quantization dictionary or the updated quantization dictionary may alternatively be provided by a third device other than the quantization device or the inverse quantization device. For example, both the quantization device and the dequantization device receive all or part of the updated quantization dictionary or the originally generated quantization dictionary from the third device, or the quantization device receives all or part of the updated quantization dictionary or the originally generated quantization dictionary from the third device and sends all or part of the updated quantization dictionary or the originally generated quantization dictionary to the dequantization device, or the dequantization device receives all or part of the updated quantization dictionary or the originally generated quantization dictionary from the third device and sends all or part of the updated quantization dictionary or the originally generated quantization dictionary to the quantization device. For example, the third device may be an over-the-top server (OTT server).

[0245] The above has described in detail the quantization method provided in the present application. The following describes the quantization or dequantization device provided in the present application.

[0246] 7 is a schematic block diagram of a quantization or dequantization apparatus according to the present invention. As shown in FIG. 7, the apparatus 1000 includes a processing unit 1100, a storage unit 1200, and a transceiver unit 1300.

[0247] The processing unit 1100 Get the first quantization dictionary, quantizing the quantization target information by using a first quantization dictionary, the first quantization dictionary including M sub-dictionaries, the M sub-dictionaries being used to quantize R quantization portions of the quantization target information, where M and R are both positive integers greater than 1; or dequantizing received information obtained by quantizing the R quantization portions by using the first quantization dictionary, the dequantized information being used as an input of a decoder, the first quantization dictionary including M sub-dictionaries, the M sub-dictionaries being used to dequantize the received information obtained by quantizing the R quantization portions, where M and R are both positive integers greater than 1; The first quantization dictionary includes a dictionary usage and dictionary content of each of the M sub-dictionaries, the dictionary usage indicating a correspondence between the M sub-dictionaries and the R quantization portions, a first sub-dictionary among the M sub-dictionaries corresponds to a first quantization portion among the R quantization portions, the first sub-dictionary is used to quantize the first quantization portion or the first sub-dictionary is used to dequantize information obtained by quantizing the first quantization portion, the dictionary content of each sub-dictionary among the M sub-dictionaries includes at least one dictionary element, each dictionary element included in the first sub-dictionary among the M sub-dictionaries is a candidate for quantization information of the first quantization portion, and M is a positive integer; Quantizing or recovering channel information by using a first quantization dictionary It is configured as follows.

[0248] Optionally, in an embodiment, each dictionary element included in the first sub-dictionary corresponds to one quantization information and one index; Each dictionary element included in the first sub-dictionary is a candidate for quantization information of the first quantization part, The quantization information corresponding to each dictionary element included in the first sub-dictionary is a candidate for quantization information of the first quantization part.

[0249] Optionally, in an embodiment, the processing unit 1100 is configured to update the second quantization dictionary to obtain the first quantization dictionary.

[0250] The transceiver unit 1300 is further configured to send all or a portion of the first quantization dictionary to the second device.

[0251] Optionally, in an embodiment, the M sub-dictionaries include P sub-dictionaries and Q sub-dictionaries.

[0252] The P sub-dictionaries are selected by the first device from a quantization dictionary table maintained at the first device, the quantization dictionary table including at least P candidate sub-dictionaries.

[0253] The Q sub-dictionaries are obtained by the first device through dictionary training.

[0254] 0≦P≦M, 0≦Q≦M, P+Q=M, and P and Q are integers.

[0255] Optionally, in an embodiment, the transceiver unit 1300 comprises: A first message is configured to be sent to the second device, the first message indicating an updated portion of the first quantization dictionary compared to the second quantization dictionary.

[0256] Optionally, in an embodiment, the updated portion comprises: and / or comprising a correspondence between L quantization portions of the R quantization portions and the sub-dictionaries, where 1≦L≦R and L is an integer; and / or It contains dictionary contents of sub-dictionaries corresponding to Z quantization portions out of the R quantization portions, where 1≦Z≦R and Z is an integer.

[0257] Optionally, in an embodiment, if the updated portion includes correspondences between L quantization portions and sub-dictionaries among the R quantization portions, the first message indicates identifiers of the L quantization portions and the sub-dictionaries corresponding to the L quantization portions, respectively; and / or If the updated portion includes dictionary contents of sub-dictionaries corresponding to Z quantization portions of the R quantization portions, the first message indicates identifiers of the Z sub-dictionaries and updated dictionary elements and indices of the Z sub-dictionaries.

[0258] Optionally, in an embodiment, the transceiver unit 1300 comprises: The device is further configured to send first information to the second device, the first information indicating that the first device updates the second quantization dictionary.

[0259] Optionally, in an embodiment, the transceiver unit 1300 comprises: The device is further configured to receive second information from the second device, the second information being used to request that the first quantization dictionary be enabled.

[0260] Optionally, in an embodiment, the transceiver unit 1300 comprises: The device is further configured to send third information to the second device, the third information indicating that the first quantization dictionary is to be enabled.

[0261] Optionally, the processing unit 1100 activating the first quantization dictionary for quantizing the to-be-quantized information after a time interval T from the moment the third information is sent, the time interval T being indicated by using the third information or agreed upon based on a protocol; or The device is configured to enable the first quantization dictionary to dequantize the received information obtained by quantizing the R quantization portions after a time interval T from the moment the third information is sent, the time interval T being indicated to the second device by using the third information or agreed upon based on a protocol.

[0262] Optionally, in an embodiment, the transceiver unit 1300 is configured to receive dictionary usage and dictionary content of the first quantization dictionary from the second device.

[0263] The processing unit 1100 is configured to receive all or a portion of the first quantization dictionary from the second device.

[0264] Optionally, in an embodiment, the transceiver unit 1300 comprises: and configured to receive a second message from the second device, the second message indicating an updated portion of the first quantization dictionary compared with a second quantization dictionary, the first quantization dictionary being obtained by updating the second quantization dictionary by the second device.

[0265] Optionally, in an embodiment, the updated portion comprises: and / or comprising a correspondence between L quantization portions of the R quantization portions and the sub-dictionaries, where 1≦L≦R and L is an integer; and / or It contains dictionary contents of sub-dictionaries corresponding to Z quantization portions out of the R quantization portions, where 1≦Z≦R and Z is an integer.

[0266] Optionally, in an embodiment, if the updated portion includes correspondences between L quantization portions and sub-dictionaries among the R quantization portions, the second message indicates identifiers of the L quantization portions and the sub-dictionaries corresponding to the L quantization portions, respectively; and / or If the updated portion includes dictionary contents of sub-dictionaries corresponding to Z quantization portions of the R quantization portions, the second message indicates identifiers of the Z sub-dictionaries and updated dictionary elements and indices of the Z sub-dictionaries.

[0267] Optionally, in an embodiment, the processing unit 1100 is configured to determine that the obtained first quantization dictionary is defective.

[0268] The transceiver unit 1300 includes: sending fourth information to the second device, the fourth information instructing the second device to send a defective portion of the first quantization dictionary, the defective portion including a dictionary usage defect and / or a dictionary content defect of the first quantization dictionary; Receive the defective part from the second device It is configured as follows.

[0269] Optionally, in an embodiment, the transceiver unit 1300 is configured to receive fifth information from the second device, the fifth information indicating that the second device updates the quantization dictionary.

[0270] Optionally, in an embodiment, the transceiver unit 1300 comprises: The sixth information is further configured to send to the second device, the sixth information being used to request that the first quantization dictionary be enabled.

[0271] Optionally, in an embodiment, the transceiver unit 1300 comprises: The device is further configured to receive seventh information from the second device, the seventh information indicating that the first quantization dictionary is to be enabled.

[0272] Optionally, in an embodiment, the processing unit 1100 the transceiver unit 1300 is configured to enable the first quantization dictionary for quantizing the to-be-quantized information after a time interval T from the moment the seventh information is received, the time interval T being indicated by using the seventh information or agreed upon based on a protocol, or The transceiver unit 1300 is configured to enable the first quantization dictionary to dequantize the received information obtained by quantizing the R quantization portions after a time interval T from the moment the transceiver unit 1300 receives the seventh information, the time interval T being indicated by using the seventh information or agreed upon based on a protocol.

[0273] Optionally, in an embodiment, the apparatus is an access network device or a chip (or chip system) configured in the access network device, the second device is a terminal device, the information to be quantized is output information of an encoder, and the first quantization dictionary is used by the apparatus to dequantize received information obtained by quantizing the R quantization parts to obtain input for a decoder.

[0274] The processing unit 1100 Evaluate the quantization performance of the first quantization dictionary; When it is determined that the decoder of the device does not conform to the first quantization dictionary, the decoder is updated based on the first quantization dictionary. It is configured as follows.

[0275] Optionally, in an embodiment, the apparatus is an access network device or a chip (or chip system) configured in an access network device.

[0276] The transceiver unit 1300 is configured to receive the first quantized information from the second device.

[0277] The processing unit 1100 is configured to dequantize the first quantized information based on the first quantized information and the second quantization dictionary.

[0278] Optionally, in an embodiment, the apparatus is an access network device or a chip (or chip system) configured in an access network device.

[0279] The transceiver unit 1300 is configured to receive second quantized information from a second device.

[0280] The processing unit 1100 is configured to dequantize the second quantized information based on the second quantized information and the first quantization dictionary.

[0281] Optionally, in an embodiment, the apparatus is a terminal device or a chip (or chip system) configured in the terminal device, the second device is an access network device, the information to be quantized is output information of an encoder, and the first quantization dictionary is used by the apparatus to quantize the output information of the encoder.

[0282] The processing unit 1100 Evaluate the quantization performance of the first quantization dictionary; When it is determined that the output of the encoder of the first device does not conform to the first quantization dictionary, the encoder is updated based on the first quantization dictionary. It is configured as follows.

[0283] Optionally, in an embodiment, the apparatus is a terminal device or a chip (or chip system) configured in a terminal device.

[0284] The processing unit 1100 is configured to determine quantized information to be fed back of the first channel information based on the second quantization dictionary.

[0285] The transceiver unit 1300 is configured to send quantized information of the first channel information to the second device.

[0286] Optionally, in an embodiment, the apparatus is a terminal device or a chip (or chip system) configured in a terminal device.

[0287] The processing unit 1100 is configured to determine quantized information to be fed back of the second channel information based on the first quantization dictionary.

[0288] The transceiver unit 1300 is configured to send quantized information of the second channel information to the second device.

[0289] Optionally, in an embodiment, when the apparatus is a terminal device or a chip or chip system configured in a terminal device, the information to be quantized is output information of an encoder, and the processing unit 1100 is specifically configured to quantize the output information of the encoder by using a first quantization dictionary.

[0290] Alternatively, when the apparatus is an access network device or a chip or chip system configured in an access network device, the processing unit 1100 is specifically configured to dequantize the received information obtained by quantizing the R quantization portions by using a first quantization dictionary.

[0291] In the above device embodiments, the processing unit 1100 is configured to perform processes and / or operations implemented in the first device in addition to transmitting and receiving operations. The transceiver unit 1300 is configured to perform receiving or transmitting operations of the first device. The storage unit 1200 is configured to store information such as a quantization dictionary and a quantization dictionary table.

[0292] 3, the processing unit 1100 is configured to perform a quantization dictionary update process to obtain the first quantization dictionary. Alternatively, the transceiver unit 1300 is configured to perform an operation of receiving the first quantization dictionary from the second device to obtain the first quantization dictionary. The processing unit 1100 is further configured to perform step 320.

[0293] As another example, in Fig. 5, transceiver unit 1300 is configured to perform steps 501, 503, 506, and 507, as well as steps 510 and 511, which are performed by a base station. Processing unit 1100 is configured to perform steps 502, 512, etc., which are performed by a base station. In Fig. 6, transceiver unit 1300 is configured to perform steps 601, 603, and steps 606 and 607, which are performed by a UE. Processing unit 1100 is configured to perform steps 602 and 608, which are performed by a UE.

[0294] In this application, the division into units (or modules) is an example and is merely a logical division of functions. In actual implementation, other division patterns may exist. Furthermore, the functional units in the examples of this disclosure may be incorporated into one processor, or may exist physically alone, or two or more units may be integrated into one unit. The integrated unit may be implemented in the form of hardware, or may be implemented in the form of a software functional module.

[0295] Based on the same technical concept, the present application further provides a quantization or dequantization device, as shown in FIG.

[0296] 8 is a diagram of the structure of a quantization or dequantization device according to the present application. As shown in FIG. 8, the device 10 includes one or more processors 11, one or more memories 12, and one or more communication interfaces 13. The processor 11 is configured to control the communication interfaces 13 to send and receive signals, the memory 12 is configured to store computer programs, and the processor 11 is configured to call and execute the computer programs in the memory 12 to enable the device 10 to perform the processing performed by the first device in the method embodiments of the present application.

[0297] For example, the processor 11 may have the functionality of the processing unit 1100 shown in Figure 7, the memory 12 may have the functionality of the storage unit 1200, and the communication interface 13 may have the functionality of the transceiver unit 1300. Specifically, the processor 11 may be configured to perform processes or operations executed within the device, the memory may be configured to perform operations of caching and deleting information units from a cache, and the communication interface 13 is configured to perform transmitting and / or receiving operations of the device.

[0298] In a possible implementation, the apparatus 10 may be used in an access network. Specifically, the apparatus 10 may be an access network device, or may be a device that can help an access network device implement the functionality of the access network device in any one of the above examples.

[0299] For example, the apparatus 10 may be a first device in a method embodiment. In this implementation, the communication interface 13 may be a transceiver of the first device. The transceiver may include a receiver and / or a transmitter. Optionally, the processor 11 may be a baseband device of the first device, and the communication interface 13 may be a radio frequency device.

[0300] For example, the apparatus 10 may be a chip (or a chip system) installed in the first device. In this implementation, the communication interface 13 may be an interface circuit or an input / output interface.

[0301] In FIG. 8, a dashed box behind a device (eg, a processor, memory, or a communication interface) indicates that there may be more than one device present.

[0302] Optionally, the memory and the processor in the above device embodiments may be physically separate units, or the memory and the processor may be integrated, which is not limited herein.

[0303] Additionally, the present application further provides a computer-readable storage medium having stored thereon computer instructions that, when executed by a computer, cause the actions and / or processes performed by the first device in the method embodiments of the present application to be performed.

[0304] Additionally, the present application further provides a computer program product, which includes computer program code and instructions that, when executed by a computer, perform the actions and / or processes performed by the first device in the method embodiments of the present application.

[0305] The present application further provides a chip, the chip including a processor, and a memory configured to store a computer program located independently of the chip, the processor configured to execute the computer program stored in the memory to enable an apparatus in which the chip is installed to perform the operations and / or processes performed by the first device in any one of the method embodiments.

[0306] Furthermore, the chip may include a communication interface, which may be an input / output interface, an interface circuit, etc. Furthermore, the chip may include a memory.

[0307] The present application further provides a communication device (which may be, for example, a chip or a chip system) including a memory, a processor, and a communication interface, wherein the communication interface is configured to receive first information and transmit the first information to the processor, and the processor is configured to process the first information element cached in the memory based on the first information.

[0308] Additionally, the present application further provides a communications apparatus including at least one processor coupled to at least one memory, the at least one processor configured to execute computer programs or instructions stored in the at least one memory to enable the communications apparatus to perform the actions and / or processes performed by the first device in any one of the method embodiments.

[0309] Additionally, the present application further provides a wireless communication system including a first device and a second device in the method embodiment of the present application.

[0310] In the embodiments of the present application, "at least one" refers to one or more. "plural" refers to two or more than two. The term "and / or" indicates an association relationship describing related objects, and indicates that three relationships may exist. For example, A and / or B may indicate three cases: only A exists, both A and B exist, and only B exists. The character " / " generally indicates an "OR" relationship between related objects. Furthermore, although terms such as "first" and "second" are sometimes used in this disclosure to describe objects, it should be understood that these objects are not limited by these terms. These terms are used merely to distinguish objects from one another.

[0311] The term "comprises" and other variations thereof used in the embodiments of this application are intended to cover non-exhaustive inclusions. For example, a process, method, system, product, or device comprising a series of steps or units is not limited to the listed steps or units, and may optionally further include other steps or units not listed, or may optionally further include other specific steps or units of the process, method, product, or device. Note that terms such as "example" or "for example" are used herein to indicate providing an example, illustration, or explanation. Any method or design solution described herein as an "example" or "for example" should not be described as preferred or advantageous over other methods or design solutions. Strictly speaking, the use of expressions such as "example" or "for example" is intended to present the relevant concept in a concrete manner.

[0312] The processors described in the embodiments of this application may be central processing units (CPUs), or may be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may be any conventional processor, etc.

[0313] The memory in the present embodiments may be volatile or nonvolatile memory, or may include both volatile and nonvolatile memory. The nonvolatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory may be random access memory (RAM) and is used as an external cache. By way of example, and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM). It should be noted that the memory of the systems and methods described herein includes, but is not limited to, these and any other suitable types of memory.

[0314] Those skilled in the art may realize that, in combination with the examples described in the embodiments disclosed herein, the units and algorithm steps may be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether a function is implemented by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use various methods to implement the described functions for each specific application, but it should not be considered that the implementation goes beyond the scope of the present application.

[0315] As can be clearly understood by those skilled in the art, for the purpose of convenient and concise description, for the detailed operation processes of the above systems, devices and units, please refer to the corresponding processes in the above method embodiments, and the details will not be described again here.

[0316] It should be understood that the disclosed systems, devices, and methods in some embodiments provided herein may be implemented in other ways. For example, the above-described device embodiments are merely examples. For example, the division of units is merely a logical division of functions, and other divisions may be used in actual implementation. For example, multiple units or components may be combined or integrated into other systems, or some features may be omitted or not implemented. Furthermore, the mutual couplings or direct couplings or communication connections shown or discussed may be implemented by using some interfaces. Indirect couplings or communication connections between devices or units may be implemented in electronic, mechanical, or other forms.

[0317] The units described as separate parts may or may not be physically separated, and the parts shown as units may or may not be physical units, and may be located in one place or distributed over multiple network units. Some or all of the units may be selected based on actual requirements to achieve the objectives of the solutions of the embodiments.

[0318] Furthermore, the functional units in the embodiments of the present application may be integrated into one processing unit, each of the units may exist physically alone, or two or more units may be integrated into one unit.

[0319] When a function is implemented in the form of a software functional unit and sold or used as an independent product, the function may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may essentially be implemented in the form of a software product, or a portion of the technical solution, or a portion of the technical solution. A computer software product is stored in a storage medium and includes instructions for instructing a computer device (such as a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method described in the embodiments of the present application. The storage medium includes any medium capable of storing program code, such as a USB flash drive, a removable hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0320] The above description is merely a specific implementation of the present application and is not intended to limit the scope of protection of the present application. Any modifications or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present application should fall within the scope of protection of the present application. Therefore, the scope of protection of the present application should be governed by the scope of protection of the claims.

[0321] This application claims priority from Chinese Patent Application No. 202210971983.8, filed with the State Intellectual Property Office of the People's Republic of China on August 12, 2022, entitled "QUANTIZATION METHOD AND APPARATUS," which is hereby incorporated by reference in its entirety.

Claims

1. 1. A quantization method performed by a first device or a chip used in said first device, comprising: Obtaining a first quantization dictionary; quantizing information to be quantized by using the first quantization dictionary, wherein the first quantization dictionary includes M sub-dictionaries, and the M sub-dictionaries are used to quantize R quantization portions of the information to be quantized, where M and R are both positive integers greater than 1; or dequantizing received information obtained by quantizing the R quantization portions by using the first quantization dictionary, wherein the first quantization dictionary includes M sub-dictionaries, and the M sub-dictionaries are used to dequantize the received information obtained by quantizing the R quantization portions, where M and R are both positive integers greater than 1; and and the first quantization dictionary includes a dictionary usage and dictionary content of each of the M sub-dictionaries, the dictionary usage indicating a correspondence between the M sub-dictionaries and the R quantization portions, a first sub-dictionary among the M sub-dictionaries corresponds to a first quantization portion among the R quantization portions, the first sub-dictionary is used to quantize the first quantization portion or the first sub-dictionary is used to dequantize information obtained by quantizing the first quantization portion, the dictionary content of each sub-dictionary among the M sub-dictionaries includes at least one dictionary element, and each dictionary element included in the first sub-dictionary among the M sub-dictionaries is a candidate for quantization information of the first quantization portion. method.

2. Each dictionary element included in the first sub-dictionary corresponds to one piece of quantization information and one index; Each dictionary element included in the first sub-dictionary is a candidate for quantization information of the first quantization part, quantization information corresponding to each dictionary element included in the first sub-dictionary is a candidate for quantization information of the first quantization part. The method of claim 1.

3. The aforementioned obtaining of the first quantization dictionary includes: updating a second quantization dictionary to obtain the first quantization dictionary; The method comprises: sending all or a portion of the first quantization dictionary to a second device.

3. The method according to claim 1 or 2.

4. The M sub-dictionaries include P sub-dictionaries and Q sub-dictionaries; the P sub-dictionaries are selected from a stored quantization dictionary table, the quantization dictionary table including at least P candidate sub-dictionaries; The Q sub-dictionaries are obtained through dictionary training; 0≦P≦M, 0≦Q≦M, P+Q=M, and P and Q are integers; The method of claim 3.

5. The sending of the portion of the first quantization dictionary to the second device includes: sending a first message to the second device; the first message indicates an updated portion of the first quantization dictionary compared with the second quantization dictionary, the first quantization dictionary being obtained based on updating the second quantization dictionary; The method according to claim 3 or 4.

6. The updated portion is a correspondence between L quantization portions of the R quantization portions and sub-dictionaries, where 1≦L≦R and L is an integer; and / or the dictionary contents of sub-dictionaries corresponding to Z quantization portions among the R quantization portions, where 1≦Z≦R and Z is an integer; The method of claim 5.

7. If the updated portion includes the correspondence between the L quantization portions of the R quantization portions and the sub-dictionaries, the first message indicates the L quantization portions and identifiers of the sub-dictionaries respectively corresponding to the L quantization portions; and / or If the updated portion includes the dictionary contents of the sub-dictionaries corresponding to the Z quantization portions of the R quantization portions, the first message indicates identifiers of the Z sub-dictionaries and updated dictionary elements and indices of the Z sub-dictionaries. The method of claim 6.

8. Before obtaining the first quantization dictionary, the method further comprises: and sending first information to the second device, the first information indicating that the first device updates the second quantization dictionary.

8. The method according to any one of claims 3 to 7.

9. After transmitting all or a portion of the first quantization dictionary to the second device, the method further comprises: and receiving second information from the second device, the second information being used to request enabling the first quantization dictionary.

9. The method according to any one of claims 3 to 8.

10. The method comprises: and sending third information to the second device, the third information instructing the second device to enable the first quantization dictionary.

10. The method according to any one of claims 3 to 9.

11. Quantizing the information to be quantized by using the first quantization dictionary as described above includes: enabling the first quantization dictionary to quantize the to-be-quantized information after a time interval T from the moment the third information is sent, the time interval T being indicated by using the third information or agreed upon based on a protocol; or The inverse quantization of the received information obtained by quantizing the R quantization portions by using the first quantization dictionary includes: and enabling the first quantization dictionary to dequantize the received information obtained by quantizing the R quantized portions after a time interval T from the moment the third information is sent, the time interval T being indicated to the second device by using the third information or agreed upon based on a protocol. The method of claim 10.

12. The aforementioned obtaining of the first quantization dictionary includes: receiving all or a portion of the first quantization dictionary from a second device; parsing the dictionary usage and / or the dictionary content of the first quantization dictionary to obtain the first quantization dictionary; having 3. The method according to claim 1 or 2.

13. The step of receiving a portion of the first quantization dictionary from the second device includes: receiving a second message from the second device, the second message indicating an updated portion of the first quantization dictionary compared to a second quantization dictionary, the first quantization dictionary being obtained based on updating the second quantization dictionary; The method of claim 12.

14. The updated portion is a correspondence between L quantization portions of the R quantization portions and sub-dictionaries, where 1≦L≦R and L is an integer; and / or the dictionary contents of sub-dictionaries corresponding to Z quantization portions among the R quantization portions, where 1≦Z≦R and Z is an integer; The method of claim 13.

15. If the updated portion includes the correspondence between the L quantization portions of the R quantization portions and the sub-dictionaries, the second message indicates the L quantization portions and identifiers of the sub-dictionaries respectively corresponding to the L quantization portions; and / or If the updated portion includes the dictionary contents of the sub-dictionaries corresponding to the Z quantization portions of the R quantization portions, the second message indicates identifiers of the Z sub-dictionaries and updated dictionary elements and indices of the Z sub-dictionaries.

15. The method of claim 14.

16. The method comprises: sending fourth information to the second device if it is determined that the obtained first quantization dictionary is defective, the fourth information instructing the second device to send a defective portion of the first quantization dictionary, the defective portion including a defect in the dictionary usage and / or a defect in the dictionary content of the first quantization dictionary; receiving the defective part from the second device; Further comprising:

16. The method according to any one of claims 12 to 15.

17. Before obtaining all or a portion of the first quantization dictionary, the method further comprises: and receiving fifth information from the second device, the fifth information indicating that the second device updates a quantization dictionary.

17. The method according to any one of claims 12 to 16.

18. After obtaining all or part of the first quantization dictionary, the method further comprises: and sending sixth information to the second device, the sixth information being used to request that the first quantization dictionary be enabled.

18. The method according to any one of claims 12 to 17.

19. The method comprises: and receiving seventh information from the second device, the seventh information indicating that the first quantization dictionary is enabled.

19. The method of any one of claims 12 to 18.

20. Quantizing the information to be quantized by using the first quantization dictionary as described above includes: activating the first quantization dictionary to quantize the to-be-quantized information after a time interval T from the moment the seventh information is received, the time interval T being indicated by using the seventh information or agreed upon based on a protocol; or The inverse quantization of the received information obtained by quantizing the R quantization portions by using the first quantization dictionary includes: and enabling the first quantization dictionary to dequantize the received information obtained by quantizing the R quantized portions after a time interval T from the moment the seventh information is received, the time interval T being indicated to the second device by using the seventh information or agreed upon based on a protocol.

20. The method of claim 19.

21. the first device is an access network device, the second device is a terminal device, the quantization target information is output information of an encoder, and the first quantization dictionary is used by the first device to dequantize the received information obtained by quantizing the R quantization parts to obtain an input of a decoder; After obtaining the first quantization dictionary, the method further comprises: evaluating the quantization performance of the first quantization dictionary; updating the decoder based on the first quantization dictionary if it is determined that the decoder of the first device does not match the first quantization dictionary; Further comprising:

21. The method of any one of claims 12 to 20.

22. Before enabling the first quantization dictionary, the method includes: receiving first quantized information from the second device; dequantizing the first quantized information based on the first quantized information and the second quantization dictionary; wherein the first quantization dictionary is updated based on the second quantization dictionary.

22. The method of claim 21.

23. After validating the first quantization dictionary, the method includes: receiving second quantized information from the second device; dequantizing the second quantized information based on the second quantized information and the first quantization dictionary; Further comprising:

23. The method of claim 21 or 22.

24. The first device is a terminal device, the second device is an access network device, the quantization target information is output information of an encoder, and the first quantization dictionary is used by the first device to quantize the output information of the encoder; After obtaining the first quantization dictionary, the method further comprises: evaluating the quantization performance of the first quantization dictionary; updating the encoder based on the first quantization dictionary when it is determined that the output of the encoder of the first device does not match the first quantization dictionary; Further comprising:

21. The method of any one of claims 12 to 20.

25. Before enabling the first quantization dictionary, the method includes: determining quantized information to be fed back of the first channel information based on the second quantization dictionary; sending the quantized information of the first channel information to the second device; Further comprising:

25. The method of claim 24.

26. After validating the first quantization dictionary, the method includes: determining quantized information to be fed back of second channel information based on the first quantization dictionary; sending the quantized information of the second channel information to the second device; and Further comprising:

26. The method of claim 24 or 25.

27. Quantizing the information to be quantized by using the first quantization dictionary as described above includes: When the first device is a terminal device and the information to be quantized is output information of an encoder, the first device quantizes the output information of the encoder by using the first quantization dictionary; or The inverse quantization of the received information obtained by quantizing the R quantization portions by using the first quantization dictionary includes: and when the first device is an access network device, dequantizing the received information obtained by quantizing the R quantization portions by using the first quantization dictionary by the first device.

21. The method of any one of claims 1 to 20.

28. 1. A quantization method performed by a first device or a chip used in said first device, comprising: Obtaining a first quantization dictionary; Quantizing first quantization target information by using the first quantization dictionary, or dequantizing the received first quantized information by using the first quantization dictionary; Obtaining a second quantization dictionary; quantizing second quantization target information by using the second quantization dictionary, or dequantizing the received second quantized information by using the second quantization dictionary; and the first quantization dictionary includes a plurality of dictionary elements, and each dictionary element included in the first quantization dictionary is a candidate for quantization target information generated in a first quantization scenario; the second quantization dictionary includes a plurality of dictionary elements, and each dictionary element included in the second quantization dictionary is a candidate for quantization target information generated in a second quantization scenario; the first quantization dictionary is different from the second quantization dictionary; method.

29. the first quantization dictionary corresponds to the first quantization scenario, and the second quantization dictionary corresponds to the second quantization scenario.

29. The method of claim 28.

30. a value distribution of the first quantization target information is different from a value distribution of the second quantization target information; 30. The method of claim 28 or 29.

31. the first quantization dictionary includes M sub-dictionaries, and the M sub-dictionaries are used to individually quantize R quantization portions of the information to be quantized or to dequantize information obtained by individually quantizing the R quantization portions, where M and R are both positive integers; the second quantization dictionary includes C sub-dictionaries, and the C sub-dictionaries are used to individually quantize S quantization portions of the information to be quantized or to dequantize information obtained by individually quantizing the S quantization portions, where S and C are both positive integers; the first quantization dictionary includes a dictionary usage and dictionary content of each of the M sub-dictionaries, the dictionary usage indicating a correspondence between the M sub-dictionaries and the R quantization portions, a first sub-dictionary among the M sub-dictionaries corresponds to a first quantization portion among the R quantization portions, the first sub-dictionary is used to quantize the first quantization portion or the first sub-dictionary is used to dequantize information obtained by quantizing the first quantization portion, the dictionary content of each sub-dictionary among the M sub-dictionaries includes at least one dictionary element, and each dictionary element included in the first sub-dictionary among the M sub-dictionaries is a candidate for quantization information of the first quantization portion; the second quantization dictionary includes a dictionary usage and dictionary content of each of the C sub-dictionaries, the dictionary usage indicating a correspondence between the C sub-dictionaries and the S quantization portions, a first sub-dictionary among the C sub-dictionaries corresponds to a first quantization portion among the S quantization portions, the first sub-dictionary is used to quantize the first quantization portion or the first sub-dictionary is used to dequantize information obtained by quantizing the first quantization portion, the dictionary content of each sub-dictionary among the C sub-dictionaries includes at least one dictionary element, and each dictionary element included in the first sub-dictionary among the C sub-dictionaries is a candidate for quantization information of the first quantization portion.

31. The method of any one of claims 28 to 30.

32. A communication device configured to perform a method according to any one of claims 1 to 31.

33. 1. A communications device having at least one processor, the at least one processor is coupled to at least one memory, the at least one processor being configured to execute computer programs or instructions stored in the at least one memory to enable the communication device to perform the method of any one of claims 1 to 31. Communication equipment.

34. 1. A quantization or dequantization device having a processor and a communication interface, In order to enable the device to perform the method of any one of claims 1 to 31, the communication interface is configured to receive data and / or information and to transmit the received data and / or information to the processor, the processor processing the data and / or information, and the communication interface is further configured to output the data and / or information processed by the processor. Quantization or dequantization device.

35. storing computer instructions; The computer instructions, when executed on a computer, perform the method of any one of claims 1 to 31. A computer-readable storage medium.

36. Contains computer program code, The computer program code, when executed on a computer, performs the method of any one of claims 1 to 31. Computer program products.

37. A wireless communication system comprising a communication device according to claim 32 or 33 or a device according to claim 34.

Images