Phase correction method and system

Abstract

The present application discloses a phase correction method and system, which uses deep learning-based artificial intelligence technology to perform signal analysis on each local signal segment of a modulated signal to capture the waveform semantic features of each local signal segment, and performs forward and backward information propagation aggregation on the waveform semantic features of each local signal segment, and then achieves accurate phase correction of the modulated signal based on the matching fusion information between the forward propagation aggregation features and the backward propagation aggregation features of the signal. In this way, the performance and adaptability of the communication system, as well as the accuracy and reliability of signal transmission, can be improved through software-level optimization without relying on traditional hardware adjustment technology.

Description

Technical Field

[0001] The present application relates to the field of intelligent correction, and more specifically, to a phase correction method and system. Background Art

[0002] In modern communication systems, modulation and demodulation are key steps to ensure effective transmission of information. First, in order to enable the signal to propagate more effectively in various transmission media, at the transmitting end of the signal, the baseband signal is loaded onto a high-frequency carrier through modulation technology to generate a modulated signal. The modulated signal passes through various communication channels, such as wireless channels, optical fibers or cables, and finally reaches the receiving end. At the receiving end, the received signal needs to be demodulated to remove or reduce the carrier frequency and restore the modulated signal to the original baseband signal, so as to facilitate subsequent signal processing and decoding. However, in the actual communication process, the carrier frequencies of the transmitting and receiving ends are often unable to be fully synchronized, resulting in a phase shift of the signal, which in turn affects the demodulation quality of the signal.

[0003] In order to improve the performance of the communication system, the signal needs to be phase corrected. In the prior art, the phase correction method usually relies on hardware adjustment, such as phase-locked loop (PLL) technology. However, the hardware implementation of phase-locked loop technology is relatively complex and requires additional calibration steps, which increases the cost and maintenance difficulty of the system. In addition, the adjustment range is limited and it is difficult to adapt to the rapidly changing communication environment. In addition, due to non-ideal factors such as temperature drift and aging of components in the circuit, phase-locked loop technology has certain limitations in practical applications.

[0004] Therefore, an optimized phase correction method and system are desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a phase correction method and system, which uses artificial intelligence technology based on deep learning to perform signal analysis on each local signal segment of the modulated signal to capture the waveform semantic features of each local signal segment, and perform forward and backward information propagation aggregation on the waveform semantic features of each local signal segment, and then based on the matching fusion information between the forward propagation aggregation features and the backward propagation aggregation features of the signal, realize accurate phase correction of the modulated signal. In this way, the performance and adaptability of the communication system, as well as the accuracy and reliability of signal transmission can be improved through software-level optimization without relying on traditional hardware adjustment technology.

[0006] According to one aspect of the present application, a phase correction method is provided, which includes:

[0007] Obtaining a signal to be modulated;

[0008] Extracting waveform semantic features of the signal to be modulated based on local scale to obtain a sequence of waveform semantic feature vectors of segments of the signal to be modulated;

[0009] Inputting the sequence of semantic feature vectors of the waveform of the signal segment to be modulated into a feature optimization module based on sequence endogenous correlation analysis to obtain a sequence of optimized semantic feature vectors of the waveform of the signal segment to be modulated;

[0010] Performing bidirectional propagation correction on the sequence of waveform-optimized semantic feature vectors of the signal to be modulated to obtain a bidirectional propagation corrected signal to be modulated significant fusion representation vector;

[0011] The bidirectional propagation correction modulated signal is significantly fused with a representation vector and input into a decoder-based phase corrector to obtain a phase-corrected modulated signal.

[0012] According to another aspect of the present application, there is provided a phase correction system, comprising:

[0013] A signal acquisition module, used to acquire a signal to be modulated;

[0014] A waveform semantic feature extraction module, used for extracting waveform semantic features of the signal to be modulated based on a local scale to obtain a sequence of waveform semantic feature vectors of segments of the signal to be modulated;

[0015] A feature optimization module, used for inputting the sequence of semantic feature vectors of the waveform of the signal segment to be modulated into a feature optimization module based on sequence endogenous correlation analysis to obtain a sequence of optimized semantic feature vectors of the waveform of the signal segment to be modulated;

[0016] A two-way propagation correction module, used for performing two-way propagation correction on the sequence of waveform-optimized semantic feature vectors of the signal segment to be modulated to obtain a significant fusion representation vector of the signal to be modulated that is corrected by two-way propagation;

[0017] The phase correction modulation signal generation module is used to input the significant fusion representation vector of the two-way propagation correction modulated signal into the decoder-based phase corrector to obtain a phase correction modulation signal.

[0018] Compared with the prior art, the phase correction method and system provided by the present application uses deep learning-based artificial intelligence technology to perform signal analysis on each local signal segment of the modulated signal to capture the waveform semantic features of each local signal segment, and performs forward and backward information propagation aggregation on the waveform semantic features of each local signal segment, and then achieves accurate phase correction of the modulated signal based on the matching fusion information between the forward propagation aggregation features and the backward propagation aggregation features of the signal. In this way, the performance and adaptability of the communication system, as well as the accuracy and reliability of signal transmission, can be improved through software-level optimization without relying on traditional hardware adjustment technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 is a flow chart of a phase correction method according to an embodiment of the present application;

[0021] Figure 2 A data flow diagram of a phase correction method according to an embodiment of the present application;

[0022] Figure 3 is a flowchart of sub-step S2 of the phase correction method according to an embodiment of the present application;

[0023] Figure 4 is a block diagram of a phase correction system according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0026] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0027] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] In the prior art, phase correction methods usually rely on hardware adjustments, such as phase-locked loop (PLL) technology. However, the hardware implementation of phase-locked loop technology is complex and requires additional calibration steps, which increases the cost and maintenance difficulty of the system. In addition, the adjustment range is limited and it is difficult to adapt to the rapidly changing communication environment. In addition, due to non-ideal factors such as temperature drift and aging of components in the circuit, phase-locked loop technology has certain limitations in practical applications. Therefore, an optimized phase correction method and system are expected.

[0030] In the technical solution of the present application, a phase correction method is proposed. Figure 1 4 is a flow chart of a phase correction method according to an embodiment of the present application. Figure 2 FIG. 2 is a schematic diagram of data flow of a phase correction method according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the phase correction method according to the embodiment of the present application includes the steps of: S1, obtaining a signal to be modulated; S2, performing waveform semantic feature extraction based on a local scale on the signal to be modulated to obtain a sequence of waveform semantic feature vectors of signal segments to be modulated; S3, inputting the sequence of waveform semantic feature vectors of signal segments to be modulated into a feature optimization module based on sequence intrinsic correlation analysis to obtain a sequence of waveform optimized semantic feature vectors of signal segments to be modulated; S4, performing bidirectional propagation correction on the sequence of waveform optimized semantic feature vectors of signal segments to be modulated to obtain a bidirectional propagation corrected significant fusion representation vector of the signal to be modulated; S5, inputting the bidirectional propagation corrected significant fusion representation vector of the signal to be modulated into a decoder-based phase corrector to obtain a phase-corrected modulated signal.

[0031] In particular, the S1 and S2 obtain a signal to be modulated; and extract waveform semantic features of the signal to be modulated based on a local scale to obtain a sequence of waveform semantic feature vectors of segments of the signal to be modulated. In particular, in a specific example of the present application, Figure 3 As shown, S2 includes: S21, dividing the signal to be modulated into signal segments to obtain a sequence of signal segments to be modulated; S22, inputting each signal segment to be modulated in the sequence of signal segments to be modulated into a signal waveform feature extractor based on a 1D-CNN model to obtain a sequence of waveform semantic feature vectors of the signal segments to be modulated.

[0032] Specifically, the S21 performs signal segment segmentation on the signal to be modulated to obtain a sequence of signal segments to be modulated. The signal to be modulated is segmented into signal segments to obtain a sequence of signal segments to be modulated. It should be understood that, considering that the statistical characteristics of non-stationary signals may be different in different time periods, in order to enhance the sensitivity of the model to signal details and thus improve the accuracy of correction, the present application further performs signal segment segmentation on the signal to be modulated to obtain a sequence of signal segments to be modulated, thereby helping to better capture the local characteristics of the signal. Preferably, in order to alleviate the boundary effect and improve the continuity of the signal segments, the signal segment segmentation can be performed using a sliding window technique to ensure that there is a certain overlap between the signal segments.

[0033] Specifically, in S22, each signal segment to be modulated in the sequence of signal segments to be modulated is input into a signal waveform feature extractor based on a 1D-CNN model to obtain a sequence of semantic feature vectors of the waveform of the signal segment to be modulated. Since the phase shift of a signal refers to the relative movement of a signal waveform on the time axis, the phase angle of the waveform changes. Therefore, in order to effectively capture the signal waveform characteristics, in the technical solution of the present application, a signal waveform feature extractor based on a 1D-CNN model is used to process each signal segment to be modulated in the sequence of signal segments to be modulated, and a sliding convolution operation is performed on the time axis using a one-dimensional convolution kernel to effectively learn the local temporal correlation in each signal segment, and dig out the deep semantic information of the signal waveform, thereby obtaining a sequence of semantic feature vectors of the waveform of the signal segment to be modulated.

[0034] It is worth noting that the one-dimensional convolutional neural network (1D-CNN) is a deep learning model for processing one-dimensional data with temporal correlation. It extracts features by applying convolution operations on the one-dimensional space of the input data. The 1D-CNN architecture includes the following layers: Input layer: accepts one-dimensional input data with a shape of [Batch Size, Channels, Temporal Length], where Batch Size represents the batch size, Channels represents the number of channels, and Temporal Length represents the temporal length; Convolution layer: applies convolution kernels to slide on the input data to extract temporal features; Activation function: applies nonlinear activation functions such as ReLU or Leaky ReLU after the convolution operation; Pooling layer: performs maximum pooling or average pooling operations to reduce the length of the output sequence; Fully connected layer: flattens the output of the convolution layer and connects it to the fully connected layer for classification or regression tasks.

[0035] It is worth mentioning that in other specific examples of the present application, the waveform semantic features of the signal to be modulated can be extracted based on the local scale in other ways to obtain a sequence of waveform semantic feature vectors of the signal fragments to be modulated, for example: the signal to be modulated is divided into overlapping or non-overlapping local scale windows, and the size and step size of the local scale window should be selected according to the characteristics of the signal and the required resolution; within each local scale window, waveform semantic features are extracted, and these features may include statistical features (such as mean, variance, kurtosis), time domain features (such as zero crossing rate, peak value) and frequency domain features (such as power spectral density); the waveform semantic features extracted from each local scale window are connected into a feature vector; as the local scale window slides on the signal, a sequence of waveform semantic feature vectors of the signal fragments to be modulated is formed.

[0036] In particular, the S3 inputs the sequence of semantic feature vectors of the waveform of the signal segment to be modulated into a feature optimization module based on sequence intrinsic correlation analysis to obtain a sequence of optimized semantic feature vectors of the waveform of the signal segment to be modulated. Since each signal segment to be modulated is continuous in the time dimension, there are inherent correlations and dependencies between each signal segment. In order to make full use of this inherent correlation and dependency and perform feature optimization on each local signal segment, the present application introduces a feature optimization module based on sequence intrinsic correlation analysis to process the sequence of semantic feature vectors of the waveform of the signal segment to be modulated, and further optimizes the expressive power of the semantic features of the waveform of each signal segment by analyzing the internal relationships and dependencies of the sequence.

[0037] In an embodiment of the present application, the sequence of semantic feature vectors of the waveform of the signal segment to be modulated is input into a feature optimization module based on sequence intrinsic correlation analysis to obtain a sequence of optimized semantic feature vectors of the waveform of the signal segment to be modulated, including: first, calculating the semantic association score vector between any two semantic feature vectors of the waveform of the signal segment to be modulated in the sequence of semantic feature vectors of the waveform of the signal segment to be modulated to obtain a sequence of semantic association score vectors of the waveform of the signal segment to be modulated; that is, by calculating the semantic association between any two semantic feature vectors of the waveform of the signal segment to be modulated to quantify their similarities or differences, so as to understand the semantic structure within the sequence. Next, the mean vector of the sequence of semantic association score vectors of the waveform of the signal segment to be modulated is calculated to obtain a global representation vector of the endogenous correlation of the sequence; further, based on the global representation vector of the endogenous correlation of the sequence, the association optimization factor of each semantic feature vector of the waveform of the signal segment to be modulated in the sequence of semantic feature vectors of the waveform of the signal segment to be modulated is calculated to obtain a sequence of association optimization factors; here, the aggregate information of the set of semantic association score vectors is used as the overall context information, and the association optimization factor of each semantic feature vector of the waveform of the signal segment to be modulated is calculated to quantify the consistency or deviation between the semantic features of the running data describing each time point and the overall semantic structure of the sequence, and this is used as a guide for feature association optimization. Furthermore, the sequence of association optimization factors is input into Activate the function to obtain a sequence of associated optimization weight factors; then use each associated optimization weight factor in the sequence of associated optimization weight factors as a weight to weight each signal segment waveform semantic feature vector in the sequence of signal segment waveform semantic feature vectors to be modulated to obtain a sequence of signal segment waveform optimization semantic feature vectors to be modulated. That is, after normalizing the associated optimization factor, use it as a weight coefficient to perform weighted optimization on the original sequence of signal segment waveform semantic feature vectors to be modulated, so as to integrate the overall context information of the signal into the semantic features of each signal segment waveform, and obtain a sequence of signal segment waveform optimization semantic feature vectors to be modulated.

[0038] Among them, the process of calculating the semantic association score vector between any two semantic feature vectors of the waveform of the signal segment to be modulated in the sequence of semantic feature vectors of the waveform of the signal segment to be modulated to obtain the sequence of semantic association score vectors of the waveform of the signal segment to be modulated includes: cascading any two semantic feature vectors of the waveform of the signal segment to be modulated in the sequence of semantic feature vectors of the waveform of the signal segment to be modulated, multiplying them by a weight coefficient matrix, and then dot-adding them with a bias vector to obtain the semantic association score vector of the waveform of the signal segment to be modulated.

[0039] More specifically, based on the global representation vector of the intrinsic correlation of the sequence, the association optimization factor of each semantic feature vector of the waveform of the signal segment to be modulated in the sequence of semantic feature vectors of the waveform of the signal segment to be modulated is calculated to obtain a sequence of association optimization factors, including: multiplying the semantic feature vector of the waveform of the signal segment to be modulated and the global representation vector of the intrinsic correlation of the sequence by different weight coefficient vectors and then performing an addition operation to obtain a semantic association coefficient; adding a bias parameter to the semantic association coefficient and then passing it through a sigmoid activation function to obtain the association optimization factor.

[0040] In summary, in the above embodiment, the sequence of semantic feature vectors of the waveform of the signal segment to be modulated is input into a feature optimization module based on sequence intrinsic correlation analysis to obtain a sequence of optimized semantic feature vectors of the waveform of the signal segment to be modulated, including: processing the sequence of semantic feature vectors of the waveform of the signal segment to be modulated with the following semantic feature association optimization formula to obtain the sequence of optimized semantic feature vectors of the waveform of the signal segment to be modulated, wherein the semantic feature association optimization formula is:

[0041]

[0042]

[0043]

[0044]

[0045]

[0046] in, A sequence representing the waveform semantic feature vectors of the signal segment to be modulated, , , , and The first, second, and third sequences of the waveform semantic feature vectors of the signal segment to be modulated are represented respectively. , and The waveform semantic feature vector of the signal segment to be modulated, The value of is the number of feature vectors in the sequence of semantic feature vectors of the waveform of the signal segment to be modulated, Indicates cascade operation, is the weight coefficient matrix, is the bias vector, is the first in the sequence of the semantic association score vectors of the waveform of the signal segment to be modulated The semantic association score vector of the waveform of the signal segment to be modulated, The value of is the number of vectors in the sequence of the semantic association score vectors of the waveform of the signal segment to be modulated, is the global representation vector of the endogenous correlation of the sequence, and Represent different weight coefficient vectors, is the bias parameter, is the sigmoid activation function, is the normalized exponential function, For the The associated optimization factor, Indicates The waveform of the signal segment to be modulated is optimized using the semantic feature vector.

[0047] In particular, the S4 performs a bidirectional propagation correction on the sequence of waveform-optimized semantic feature vectors of the signal segment to be modulated to obtain a bidirectional propagation-corrected significant fusion representation vector of the signal to be modulated. In a specific example of the present application, the sequence of waveform-optimized semantic feature vectors of the signal segment to be modulated is input into a forward corrector based on a forward LSTM model to obtain a forward propagation aggregate representation vector of the signal to be modulated; the sequence of waveform-optimized semantic feature vectors of the signal segment to be modulated is input into a backward corrector based on a backward LSTM model to obtain a backward propagation aggregate representation vector of the signal to be modulated; and the forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated are input into a significant fusion network based on feature principal component query matching to obtain the bidirectional propagation-corrected significant fusion representation vector of the signal to be modulated. In order to more accurately understand the global time-domain waveform change pattern of the signal to be modulated, in the technical solution of the present application, a forward LSTM model and a backward LSTM model are used as a forward corrector and a backward corrector, respectively, to perform bidirectional propagation aggregation processing on the sequence of waveform-optimized semantic feature vectors of the signal segment to be modulated. The forward LSTM model processes signal segments in chronological order to capture the natural evolution of the signal waveform over time and generate a forward propagation aggregation representation vector of the signal to be modulated; while the backward LSTM model processes signal segments in the opposite chronological order, and by tracing back segment by segment, captures the impact of the state of the signal at the end on the previous state and generates a backward propagation aggregation representation vector of the signal to be modulated. Through this two-way propagation aggregation processing, the forward and backward context information of the signal can be obtained at the same time, which helps to more accurately understand the global time domain waveform change pattern of the signal, thereby achieving accurate phase correction of the modulated signal. Secondly, in order to make full use of the correlation and complementarity between the forward and backward propagation aggregated information of the signal to be modulated to perform phase correction and reconstruction of the signal, the present application further introduces a significant fusion network based on feature principal component query matching to match and fuse the forward propagation aggregated representation vector of the signal to be modulated and the backward propagation aggregated representation vector of the signal to be modulated. Through principal component analysis and query matching technology, the significant correlation features between the forward context information and the backward context information of the signal to be modulated are extracted and fused to obtain the corrected signal waveform feature representation, that is, the significant fusion representation vector of the bidirectionally propagated corrected signal to be modulated.

[0048] In an embodiment of the present application, the forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated are input into a significant fusion network based on feature principal component query matching to obtain the significant fusion representation vector of the bidirectional propagation corrected signal to be modulated, including: first, normalizing the forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated to obtain a standardized forward propagation aggregate representation vector of the signal to be modulated and a standardized backward propagation aggregate representation vector of the signal to be modulated to eliminate the influence of different dimensions and magnitudes; then, respectively calculating the sample co-relation of the standardized forward propagation aggregate representation vector of the signal to be modulated and the standardized backward propagation aggregate representation vector of the signal to be modulated; variance matrix to obtain the forward propagation aggregate sample covariance matrix of the signal to be modulated and the backward propagation aggregate sample covariance matrix of the signal to be modulated; then, the forward propagation aggregate sample covariance matrix of the signal to be modulated and the backward propagation aggregate sample covariance matrix of the signal to be modulated are subjected to eigenvector extraction based on matrix decomposition to obtain a set of forward propagation aggregate principal component eigenvectors of the signal to be modulated and a set of backward propagation aggregate principal component eigenvectors of the signal to be modulated; that is, the principal component analysis algorithm is used to perform sample covariance calculation and principal component feature extraction on the standardized forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated, so as to mine out the main components and remove redundant information; then, the signal to be modulated is subjected to eigenvector extraction based on matrix decomposition to obtain a set of forward propagation aggregate principal component eigenvectors of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated. The set of the forward propagation aggregated principal component feature vectors of the signal to be modulated and the set of the backward propagation aggregated principal component feature vectors of the signal to be modulated are input into the maximum approximate query matching network to obtain the set of the best matching pairs of the forward propagation aggregated principal component feature vectors of the signal to be modulated and the backward propagation aggregated principal component feature vectors of the signal to be modulated; and each best matching pair in the set of the best matching pairs of the forward propagation aggregated principal component feature vectors of the signal to be modulated and the backward propagation aggregated principal component feature vectors of the signal to be modulated is input into the semantic fine-grained joint analysis module to obtain the set of the bidirectional propagation aggregated principal component joint feature vectors of the signal to be modulated; here, based on the maximum approximate query matching between the feature vectors, the most similarity between the two principal component feature vector sets is calculated and compared, and the best matching pair is found. The best matching pairs are selected, and the significant correlation features between the two are extracted by performing fine-grained semantic joint analysis on each best matching pair, so as to give priority to feature information with high similarity in both forward and backward directions in the feature fusion process, ensure the consistency of forward and backward information, and avoid phase deviation or error that may be introduced due to unidirectional processing; then the set of joint feature vectors of the bidirectional propagation aggregated principal components of the signal to be modulated are cascaded to obtain the significant fusion representation vector of the bidirectional propagation corrected signal to be modulated; by cascading the feature joint results of each best matching pair, multiple core component joint features of the forward and backward propagation aggregated information of the aggregated signal are aggregated to generate a global significant fusion representation vector of the bidirectional propagation corrected signal to be modulated.

[0049] Among them, the process of inputting the set of aggregated principal component feature vectors of forward propagation of the signal to be modulated and the set of aggregated principal component feature vectors of backward propagation of the signal to be modulated into the maximum approximation query matching network to obtain the set of best matching pairs of the aggregated principal component feature vectors of forward propagation of the signal to be modulated and the aggregated principal component feature vectors of backward propagation of the signal to be modulated includes: calculating the cosine similarity between the i-th aggregated principal component feature vector of forward propagation of the signal to be modulated in the set of aggregated principal component feature vectors of forward propagation of the signal to be modulated and each aggregated principal component feature vector of backward propagation of the signal to be modulated in the set of aggregated principal component feature vectors of backward propagation of the signal to be modulated to obtain a set of cosine similarities; and forming the best matching pair with the i-th aggregated principal component feature vector of forward propagation of the signal to be modulated corresponding to the maximum value in the set of cosine similarities.

[0050] More specifically, the process of inputting each best matching pair in the set of the best matching pairs of the forward propagation aggregated principal component feature vector of the signal to be modulated and the backward propagation aggregated principal component feature vector of the signal to be modulated into the semantic fine-grained joint analysis module to obtain a set of bidirectional propagation aggregated principal component joint feature vectors of the signal to be modulated includes: calculating the point-by-point addition, point-by-point subtraction and point-by-point multiplication of the forward propagation aggregated principal component feature vector of the signal to be modulated and the backward propagation aggregated principal component feature vector of the signal to be modulated in the best matching pair to obtain a first matching fusion result, a second matching fusion result and a third matching fusion result; after cascading the first matching fusion result, the second matching fusion result and the third matching fusion result into a multi-scale fusion feature vector of the forward propagation aggregated principal component of the signal to be modulated and the backward propagation aggregated principal component of the signal to be modulated, one-dimensional convolution processing and maximum value pooling processing are performed on them to obtain the bidirectional propagation aggregated principal component joint feature vector of the signal to be modulated.

[0051] In summary, in the above embodiment, the forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated are input into a significant fusion network based on feature principal component query matching to obtain the significant fusion representation vector of the bidirectional propagation corrected signal to be modulated, including: fusing the forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated using the following principal component query matching formula to obtain the significant fusion representation vector of the bidirectional propagation corrected signal to be modulated, wherein the principal component query matching formula is:

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065] in, represents the forward propagation aggregation representation vector of the signal to be modulated, and represent the mean and standard deviation of the forward propagation aggregate representation vector of the signal to be modulated, respectively, represents the normalized forward propagation aggregation representation vector of the modulated signal, represents the backward propagation aggregation representation vector of the signal to be modulated, and represent the mean and standard deviation of the backward propagation aggregate representation vector of the modulated signal, respectively, represents the normalized modulated signal back-propagation aggregation representation vector, The value of is the length of the forward propagation aggregation representation vector of the signal to be modulated, The value of is the length of the back propagation aggregation representation vector of the modulated signal, represents the covariance matrix of the aggregated samples of the forward propagation of the modulated signal, represents the covariance matrix of the aggregated samples of the modulated signal back propagated, represents a matrix consisting of a set of aggregated principal component eigenvectors of the forward propagation of the signal to be modulated obtained by performing eigenvalue decomposition on the aggregated sample covariance matrix of the forward propagation of the signal to be modulated, represents a diagonal matrix consisting of a set of aggregated principal component eigenvalues ​​of the forward propagation of the signal to be modulated obtained by performing eigenvalue decomposition on the aggregated sample covariance matrix of the forward propagation of the signal to be modulated, , , and respectively represent the first, second, and third principal component eigenvectors of the forward propagation aggregate of the modulated signal. and The modulated signal is forward propagated to aggregate the principal component eigenvectors, , and respectively represent the first, second and third in the set of the aggregated principal component eigenvalues ​​of the forward propagation of the modulated signal The modulated signal is forward propagated to aggregate the principal component eigenvalues, The value of is the number of principal component eigenvalues ​​of the forward propagation aggregation of the signal to be modulated, represents a matrix consisting of a set of principal component eigenvectors of the back-propagated aggregated signal to be modulated obtained by performing eigenvalue decomposition on the back-propagated aggregated sample covariance matrix of the signal to be modulated, represents a diagonal matrix consisting of a set of eigenvalues ​​of the back-propagated aggregated principal components of the signal to be modulated obtained by performing eigenvalue decomposition on the back-propagated aggregated sample covariance matrix of the signal to be modulated, , , , and respectively represent the first, second, and third principal component eigenvectors of the backward propagation aggregate of the modulated signal. , and The modulated signal is back-propagated to aggregate the principal component eigenvectors, , and respectively represent the first, second and third in the set of the aggregated principal component eigenvalues ​​of the backward propagation of the modulated signal The modulated signal is back-propagated to aggregate the principal component eigenvalues, is the diag function, which is used to construct a diagonal matrix. represents the inner product, represents the magnitude of the eigenvector, Indicates the index corresponding to the maximum value. Indicates that The index of the aggregated principal component eigenvector of the modulated signal forward propagated to form the best matching pair, Indicates cascade operation, represents the maximum pooling operation, represents a one-dimensional convolution operation, represents dot product, Indicates point addition, Indicates point reduction, Indicates The joint eigenvector of the principal components of the modulated signal is propagated bidirectionally, The bidirectional propagation correction modulated signal is significantly fused into a representation vector.

[0066] In particular, in S5, the significant fusion representation vector of the bidirectional propagation correction signal to be modulated is input into a decoder-based phase corrector to obtain a phase-corrected modulated signal. That is, the waveform features in the significant fusion representation vector of the bidirectional propagation correction signal to be modulated are decoded by the decoder, and the global time domain variation pattern of the signal waveform after phase correction is learned therefrom, and it is restored to the original signal form through a layer-by-layer back propagation algorithm, thereby obtaining a phase-corrected modulated signal.

[0067] Preferably, considering that the forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated respectively represent the forward time propagation aggregate features and the backward time propagation aggregate features of the optimized semantic features of the waveform of the signal segment to be modulated, when performing significant fusion based on feature principal component query matching, there will be a signal waveform semantic fusion offset caused by the imbalance of feature query matching significance under the difference in feature propagation time direction, so as to further improve the signal waveform semantic fusion regression comprehensibility of the bidirectional propagation corrected signal to be modulated significant fusion representation vector, thereby improving the accuracy of the phase corrected modulated signal obtained by the phase corrector based on the decoder input.

[0068] Therefore, in a preferred embodiment, when the significant fusion representation vector of the two-way propagation correction signal to be modulated is input into a decoder-based phase corrector to obtain a phase correction modulated signal, the significant fusion representation vector of the two-way propagation correction signal to be modulated is optimized, including the steps of:

[0069] Subtract the zero norm of the bidirectional propagation correction signal to be modulated significant fusion representation vector from the length of the bidirectional propagation correction signal to be modulated significant fusion representation vector to obtain a bidirectional propagation correction signal to be modulated significant fusion isolated representation value;

[0070] Calculate the logarithm of the sum of the square of the significant fusion isolated representation value of the two-way propagation correction signal to be modulated and the significant fusion isolated representation value of the two-way propagation correction signal to be modulated with base 2 to obtain a significant fusion information order value of the two-way propagation correction signal to be modulated;

[0071] Calculate a power function with each eigenvalue of the bidirectional propagation correction signal to be modulated significant fusion representation vector as the base, the difference of the bidirectional propagation correction signal to be modulated significant fusion isolated representation value minus one as the exponent, and perform point multiplication with the bidirectional propagation correction signal to be modulated significant fusion information order value to obtain the bidirectional propagation correction signal to be modulated significant fusion leading vector;

[0072] After performing a dot multiplication on the difference between the bidirectional propagation correction signal to be modulated significant fusion representation vector and the bidirectional propagation correction signal to be modulated significant fusion isolated representation value minus one, the dot multiplication is then performed on the inverse of the bidirectional propagation correction signal to be modulated significant fusion isolated representation value to obtain a bidirectional propagation correction signal to be modulated significant fusion field constraint vector;

[0073] Calculate an exponential function with a natural constant as the base and each eigenvalue of the bidirectional propagation correction signal to be modulated significant fusion field constraint vector as the exponent to obtain a bidirectional propagation correction signal to be modulated significant fusion field bias vector; and

[0074] The bidirectional propagation correction signal to be modulated significant fusion pilot vector and the bidirectional propagation correction signal to be modulated significant fusion field bias vector are dot-added to obtain an optimized bidirectional propagation correction signal to be modulated significant fusion representation vector.

[0075] Among them, the bidirectional propagation correction modulated signal significant fusion representation vector, for example, is recorded as The optimization process is expressed as:

[0076]

[0077]

[0078] in, is the significant fusion representation vector of the bidirectional propagation correction modulated signal, represents a power function with each eigenvalue of the bidirectional propagation correction signal to be modulated significant fusion representation vector as the base and the difference of the bidirectional propagation correction signal to be modulated significant fusion isolated representation value minus one as the exponent, is the length of the bidirectional propagation correction modulated signal significant fusion representation vector, is the zero norm of the significant fusion representation vector of the bidirectional propagation correction modulated signal, is the significant fusion isolated representation value of the bidirectional propagation correction modulated signal, represents dot product, Indicates point addition, represents the logarithm operation with base 2, () indicates exponential operation, It is the optimized bidirectional propagation correction modulated signal significant fusion representation vector.

[0079] Therefore, the high-dimensional feature manifold of the significant fusion representation vector of the bidirectional propagation correction signal to be modulated is represented by using the eigenvalues ​​of the feature set as the vector field of the aggregation dimension, and the superposition value of the vector field of the significant fusion representation vector of the bidirectional propagation correction signal to be modulated at the isolated zero position is used as the order information to fix the local position of the eigenvalue of its feature set, and a bias for the reversibility of the feature regression distribution field of the significant fusion representation vector of the bidirectional propagation correction signal to be modulated is added as a reward to achieve the mapping target tracking of the regression distribution of the significant fusion representation vector of the bidirectional propagation correction signal to be modulated for the eigenvalue position, so that the feature set of the significant fusion representation vector of the bidirectional propagation correction signal to be modulated perceives the mapping migration to the aggregate distribution, thereby improving the accuracy of the phase-corrected modulated signal obtained by inputting the phase corrector based on the decoder of the significant fusion representation vector of the bidirectional propagation correction signal to be modulated by improving the signal waveform semantic fusion regression comprehensibility of the significant fusion representation vector of the bidirectional propagation correction signal to be modulated.

[0080] In summary, the phase correction method according to the embodiment of the present application is explained, which uses artificial intelligence technology based on deep learning to perform signal analysis on each local signal segment of the modulated signal to capture the waveform semantic features of each local signal segment, and performs forward and backward information propagation aggregation on the waveform semantic features of each local signal segment, and then based on the matching fusion information between the forward propagation aggregation features and the backward propagation aggregation features of the signal, realizes accurate phase correction of the modulated signal. In this way, the performance and adaptability of the communication system, as well as the accuracy and reliability of signal transmission can be improved through software-level optimization without relying on traditional hardware adjustment technology.

[0081] Furthermore, a phase correction system is also provided.

[0082] Figure 4 FIG. 1 is a block diagram of a phase correction system according to an embodiment of the present application. Figure 4As shown, according to the phase correction system 300 of the embodiment of the present application, it includes: a signal acquisition module 310, which is used to acquire the signal to be modulated; a waveform semantic feature extraction module 320, which is used to perform waveform semantic feature extraction based on local scale on the signal to be modulated to obtain a sequence of waveform semantic feature vectors of the signal segments to be modulated; a feature optimization module 330, which is used to input the sequence of waveform semantic feature vectors of the signal segments to be modulated into a feature optimization module based on sequence intrinsic correlation analysis to obtain a sequence of waveform optimized semantic feature vectors of the signal segments to be modulated; a two-way propagation correction module 340, which is used to perform two-way propagation correction on the sequence of waveform optimized semantic feature vectors of the signal segments to be modulated to obtain a significant fusion representation vector of the two-way propagation corrected signal to be modulated; a phase correction modulated signal generation module 350, which is used to input the significant fusion representation vector of the two-way propagation corrected signal to be modulated into a decoder-based phase corrector to obtain a phase correction modulated signal.

[0083] As described above, the phase correction system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a phase correction algorithm. In a possible implementation, the phase correction system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the phase correction system 300 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the phase correction system 300 can also be one of the many hardware modules of the wireless terminal.

[0084] Alternatively, in another example, the phase correction system 300 and the wireless terminal may also be separate devices, and the phase correction system 300 may be connected to the wireless terminal via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0085] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A phase correction method, characterized in that: include: Obtaining a signal to be modulated; Extracting waveform semantic features of the signal to be modulated based on local scale to obtain a sequence of waveform semantic feature vectors of segments of the signal to be modulated; Inputting the sequence of semantic feature vectors of the waveform of the signal segment to be modulated into a feature optimization module based on sequence endogenous correlation analysis to obtain a sequence of optimized semantic feature vectors of the waveform of the signal segment to be modulated; Performing bidirectional propagation correction on the sequence of waveform-optimized semantic feature vectors of the signal to be modulated to obtain a bidirectional propagation corrected signal to be modulated significant fusion representation vector; Inputting the significant fusion representation vector of the bidirectional propagation correction signal to be modulated into a decoder-based phase corrector to obtain a phase correction modulated signal; Extracting waveform semantic features of the signal to be modulated based on local scale to obtain a sequence of waveform semantic feature vectors of segments of the signal to be modulated, including: Dividing the signal to be modulated into signal segments to obtain a sequence of signal segments to be modulated; Inputting each of the signal segments to be modulated in the sequence of the signal segments to be modulated into a signal waveform feature extractor based on a 1D-CNN model to obtain a sequence of waveform semantic feature vectors of the signal segments to be modulated; Inputting the sequence of waveform semantic feature vectors of the signal segment to be modulated into a feature optimization module based on sequence endogenous correlation analysis to obtain a sequence of waveform optimized semantic feature vectors of the signal segment to be modulated, including: Calculating a semantic association score vector between any two semantic feature vectors of the waveform of the signal segment to be modulated in the sequence of semantic feature vectors of the waveform of the signal segment to be modulated to obtain a sequence of semantic association score vectors of the waveform of the signal segment to be modulated; Calculating the mean vector of the sequence of semantic association score vectors of the waveform of the signal segment to be modulated to obtain a global representation vector of the sequence endogenous correlation; Based on the global representation vector of the endogenous correlation of the sequence, calculating the correlation optimization factor of each of the waveform semantic feature vectors of the signal segment to be modulated in the sequence of waveform semantic feature vectors of the signal segment to be modulated to obtain a sequence of correlation optimization factors; The sequence of the associated optimization factors is input into Activation function to obtain a sequence of associated optimization weight factors; Using each associated optimization weight factor in the sequence of associated optimization weight factors as a weight, weighting each signal segment waveform semantic feature vector to be modulated in the sequence of signal segment waveform semantic feature vectors to be modulated to obtain a sequence of signal segment waveform optimization semantic feature vectors to be modulated; Calculating the semantic association score vector between any two of the to-be-modulated signal segment waveform semantic feature vectors in the sequence of the to-be-modulated signal segment waveform semantic feature vectors to obtain a sequence of to-be-modulated signal segment waveform semantic association score vectors, including: Any two waveform semantic feature vectors of the signal segment to be modulated in the sequence of waveform semantic feature vectors of the signal segment to be modulated are cascaded and multiplied by a weight coefficient matrix, and then dotted with a bias vector to obtain a waveform semantic association score vector of the signal segment to be modulated; Based on the sequence endogenous correlation global representation vector, calculating the correlation optimization factor of each to-be-modulated signal segment waveform semantic feature vector in the sequence of to-be-modulated signal segment waveform semantic feature vectors to obtain a sequence of correlation optimization factors, including: The semantic feature vector of the waveform of the signal segment to be modulated and the global representation vector of the sequence endogenous correlation are multiplied by different weight coefficient vectors respectively and then added to obtain a semantic correlation coefficient; The semantic association coefficient is added with a bias parameter and then passed through a sigmoid activation function to obtain the association optimization factor; Performing bidirectional propagation correction on the sequence of waveform-optimized semantic feature vectors of the signal segment to be modulated to obtain a bidirectional propagation corrected significant fusion representation vector of the signal to be modulated, including: Inputting the sequence of waveform-optimized semantic feature vectors of the signal to be modulated into a forward corrector based on a forward LSTM model to obtain a forward propagation aggregate representation vector of the signal to be modulated; Inputting the sequence of waveform-optimized semantic feature vectors of the signal to be modulated into a backward corrector based on a backward LSTM model to obtain a backward propagation aggregate representation vector of the signal to be modulated; The forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated are input into a significant fusion network based on feature principal component query matching to obtain the significant fusion representation vector of the bidirectional propagation corrected signal to be modulated.

2. The phase correction method according to claim 1, characterized in that: Inputting the forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated into a significant fusion network based on feature principal component query matching to obtain the significant fusion representation vector of the bidirectional propagation corrected signal to be modulated, including: Standardizing the forward propagation aggregate representation vector of the signal to be modulated and the backward propagation aggregate representation vector of the signal to be modulated to obtain a standardized forward propagation aggregate representation vector of the signal to be modulated and a standardized backward propagation aggregate representation vector of the signal to be modulated; Respectively calculating the sample covariance matrix of the standardized forward propagation aggregate representation vector of the signal to be modulated and the standardized backward propagation aggregate representation vector of the signal to be modulated to obtain the forward propagation aggregate sample covariance matrix of the signal to be modulated and the backward propagation aggregate sample covariance matrix of the signal to be modulated; Performing eigenvector extraction based on matrix decomposition on the forward propagation aggregated sample covariance matrix of the signal to be modulated and the backward propagation aggregated sample covariance matrix of the signal to be modulated to obtain a set of aggregated principal component eigenvectors of the forward propagation aggregated signal to be modulated and a set of aggregated principal component eigenvectors of the backward propagation aggregated signal to be modulated; Inputting the set of the aggregated principal component feature vectors of the forward propagation of the signal to be modulated and the set of the aggregated principal component feature vectors of the backward propagation of the signal to be modulated into a maximum approximate query matching network to obtain a set of best matching pairs of the aggregated principal component feature vectors of the forward propagation of the signal to be modulated and the aggregated principal component feature vectors of the backward propagation of the signal to be modulated; Input each best matching pair in the set of best matching pairs of the forward propagation aggregated principal component feature vector of the signal to be modulated and the backward propagation aggregated principal component feature vector of the signal to be modulated into a semantic fine-grained joint analysis module to obtain a set of bidirectional propagation aggregated principal component joint feature vectors of the signal to be modulated; The set of the bidirectional propagation aggregated principal component joint eigenvectors of the signal to be modulated is cascaded to obtain a significant fusion representation vector of the bidirectional propagation corrected signal to be modulated.

3. The phase correction method according to claim 2, characterized in that: Inputting the set of the aggregated principal component feature vectors of the forward propagation of the signal to be modulated and the set of the aggregated principal component feature vectors of the backward propagation of the signal to be modulated into a maximum approximate query matching network to obtain a set of best matching pairs of the aggregated principal component feature vectors of the forward propagation of the signal to be modulated and the aggregated principal component feature vectors of the backward propagation of the signal to be modulated, comprising: Calculating the cosine similarity between the i-th forward propagation aggregated principal component eigenvector of the signal to be modulated in the set of forward propagation aggregated principal component eigenvectors of the signal to be modulated and each backward propagation aggregated principal component eigenvector of the signal to be modulated in the set of backward propagation aggregated principal component eigenvectors of the signal to be modulated to obtain a set of cosine similarities; The backward propagation aggregated principal component feature vector of the signal to be modulated corresponding to the maximum value in the set of cosine similarities and the forward propagation aggregated principal component feature vector of the i-th signal to be modulated form the best matching pair.

4. The phase correction method according to claim 3, characterized in that: Inputting each best matching pair in the set of best matching pairs of the forward propagation aggregated principal component feature vector of the signal to be modulated and the backward propagation aggregated principal component feature vector of the signal to be modulated into a semantic fine-grained joint analysis module to obtain a set of bidirectional propagation aggregated principal component joint feature vectors of the signal to be modulated, including: Calculate the position point addition, position point subtraction and position point multiplication between the forward propagation aggregated principal component feature vector of the to-be-modulated signal and the backward propagation aggregated principal component feature vector of the to-be-modulated signal in the best matching pair to obtain a first matching fusion result, a second matching fusion result and a third matching fusion result; After the first matching fusion result, the second matching fusion result and the third matching fusion result are cascaded into a multi-scale fusion feature vector of the forward propagation aggregation of the signal to be modulated and the backward propagation aggregation of the signal to be modulated, one-dimensional convolution processing and maximum pooling processing are performed on them to obtain the joint feature vector of the bidirectional propagation aggregation principal components of the signal to be modulated.

5. A phase correction system, used to perform the phase correction method according to claim 1, characterized in that: include: A signal acquisition module, used to acquire a signal to be modulated; A waveform semantic feature extraction module, used for extracting waveform semantic features of the signal to be modulated based on a local scale to obtain a sequence of waveform semantic feature vectors of segments of the signal to be modulated; A feature optimization module, used for inputting the sequence of semantic feature vectors of the waveform of the signal segment to be modulated into a feature optimization module based on sequence endogenous correlation analysis to obtain a sequence of optimized semantic feature vectors of the waveform of the signal segment to be modulated; A two-way propagation correction module, used for performing two-way propagation correction on the sequence of waveform-optimized semantic feature vectors of the signal segment to be modulated to obtain a significant fusion representation vector of the signal to be modulated that is corrected by two-way propagation; The phase correction modulation signal generation module is used to input the significant fusion representation vector of the two-way propagation correction modulated signal into the decoder-based phase corrector to obtain a phase correction modulation signal.

Images