Channel equalization model training method, fading signal equalization processing method and device
By adopting hybrid attention mechanism and deep separable convolution techniques in wireless communication, a channel equalization model is built, which solves the problems of long training time and low efficiency of traditional channel equalization schemes, and efficient signal reconstruction and channel equalization are achieved.
Patent Information
- Application Number
- CN202411065038.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-08-05
AI Technical Summary
Traditional channel equalization schemes require a long training time in wireless communication, reduce transmission efficiency, and are difficult to effectively deal with multipath effect and Doppler frequency deviation, resulting in reduced signal modulation recognition accuracy and high demodulation bit error rate.
A channel equalization model training method based on a hybrid attention mechanism is proposed. Using deep separation convolution and time residual expansion convolution techniques, signal characteristics are extracted through dual branch encoder and multi-head self-attention mechanism to construct channel equalization Model, improve model training speed and signal transmission efficiency.
It realizes efficient signal reconstruction, improves channel equalization efficiency, reduces model complexity, improves signal transmission accuracy and efficiency, and enhances the anti-interference ability of faded channels.
Smart Images

Figure CN119071116A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of communication and computer technology, and in particular to a channel equalization model training method, and an equalization processing method and device for fading signals. Background Art
[0002] In wireless communication systems, transmission signals are inevitably affected by fading channels, and there are multipath effects and Doppler frequency deviation phenomena, which in turn lead to transmission signal deterioration, reduced signal modulation recognition accuracy at the receiving end, and a large demodulation bit error rate.
[0003] The global communication system faces many challenges such as obtaining confidential information from inter-symbol interference (ISI). Generally, channel equalization is used to compensate for the damage caused by the fading channel to create conditions for subsequent modulation identification, signal demodulation, signal interception or interference. Therefore, channel equalization has always been a research hotspot in wireless communications.
[0004] Traditional channel equalization schemes mainly rely on time domain or frequency domain channel equalization, which requires sending a large number of pilot signals to test the channel status and train the filter parameters. This not only requires a long training time, but also reduces the transmission efficiency. Summary of the invention
[0005] The purpose of the embodiments of this specification is to provide a channel equalization model training method, an equalization processing method and device for fading signals, so as to efficiently realize signal reconstruction, achieve the purpose of channel equalization, and increase the model training speed and improve the subsequent signal transmission efficiency.
[0006] To solve the above technical problems, the embodiments of this specification are implemented as follows:
[0007] In the first aspect, a channel equalization model training method is proposed, comprising:
[0008] Determine an original signal, and simulate the original signal to generate a sample data set; wherein the sample data in the sample data set includes: a noise-free signal used as label data and a noise signal used as a training sample, wherein the noise signal is a modulated signal carrying the influence of Rayleigh fading;
[0009] An initial model based on a hybrid attention mechanism is selected; the initial model includes: a dual-branch encoder and a single-branch decoder, wherein the encoding block trunk in the dual-branch encoder is used to extract detail spatial features, and the time residual dilated convolution RTDC auxiliary branch is used to construct temporal feature dependencies;
[0010] Input the training sample data in the sample data set into the initial model, extract all-round feature information of the noise signal using a dual-branch encoder, and perform dimension upgrade on the feature information using a decoder to output a reconstructed signal;
[0011] Based on the SmoothL1Loss loss value of the reconstructed signal and the noise-free signal in the training sample data, the initial model is repeatedly iterated, and the learning rate is adjusted to determine the model parameters to obtain a trained channel equalization model.
[0012] Optionally, determining an original signal and simulating the original signal to generate a sample data set specifically includes:
[0013] Generate a binary bit stream randomly and simulate the original signal to be transmitted;
[0014] Based on the original signal, multiple different noise-free signals are generated by simulating modulation according to multiple preset modulation modes, and the noise-free signals are preprocessed, wherein the preprocessing operation at least includes: waveform adjustment and power normalization processing;
[0015] Simulating a Rayleigh fading channel, simulating a noise-free signal after a preprocessing operation, a fading signal transmitted through the Rayleigh fading channel, and simulating a Gaussian white noise channel, adding noise power to the fading signal to obtain a noise signal;
[0016] A variety of different noise-free signals and their corresponding noise signals are summarized as a sample data set.
[0017] Optionally, the initial model is an autoencoder; the autoencoder based on the hybrid attention mechanism includes: a dual-branch encoder, a multi-head self-attention mechanism MHSA fusion feature module, and a single-branch decoder; wherein the dual-branch encoder includes: a coding block trunk and an RTDC auxiliary branch, the coding block trunk includes 4 coding blocks; the single-branch decoder includes 3 stacked decoding blocks;
[0018] The dual-branch encoder is composed of a first convolutional layer and four encoding blocks connected in sequence, and a residual connection of a CBAM attention mechanism is set between the first convolutional layer and the third encoding block, and a residual connection is set in the first layer and the third layer inside the encoding block and after the last scaling operation; the single-branch decoder is composed of three decoding blocks and a transposed convolutional layer connected in series, a residual connection is set between the first decoding block and the second decoding block, and a residual connection is set in the first layer and the third layer inside the decoding block and after the last scaling operation.
[0019] Optionally, the encoding block includes: a deep separable convolution DSC, and a 3-layer DSC cascade compression-excitation (SE) attention mechanism, wherein each layer of DSC is followed by a batch normalization layer and an activation function layer, and the SE module is composed of a pooling layer, a first convolution layer, a LeakyReLU activation function, a second convolution layer, and a Sigmod activation function;
[0020] On the trunk side of the coding block, when a dual-branch encoder is used to extract all-round feature information of the noise signal, the following steps are specifically performed:
[0021] The first layer DSC is used to expand the channel dimension of the feature information of the noise signal input to the coding block to obtain the corresponding feature map;
[0022] The second layer DSC is used to downsample the feature map by controlling the step size and padding;
[0023] The third layer DSC is used to further adjust the number of channels of the input feature map;
[0024] The input feature map is compressed using the pooling layer and the first convolutional layer to compress the feature dimension;
[0025] The activation function and the second convolutional layer are used to stimulate the compressed features, transform the feature dimensions, and assign weights to obtain the feature weight size.
[0026] Optionally, the RTDC auxiliary branch is composed of 3 layers of DSC cascaded with 4 layers of time dilation convolution and CBAM;
[0027] On the auxiliary branch side of the RTDC, the dual-branch encoder is used to extract all-round feature information of the noise signal, specifically including:
[0028] The feature information of the input noise signal is respectively expanded through channels, reorganized in dimensions, and compressed in feature map channels and spatial dimensions through DSC with kernel sizes of 3×3, 1×2, and 3×3. The feature dimensions are matched once based on 2×4 DSC and the residual operation is performed with the compressed output.
[0029] The time series features obtained by residual operation are convolved N times in sequence: using a kernel size of 1×3 and a dilation rate of 2 N-1 The time dilation convolution TCN performs padding convolution and downsampling, where N is a positive integer;
[0030] The CBAM mechanism is used to assign feature weights to the extracted time series feature information.
[0031] Optionally, the formula of the CBAM is:
[0032]
[0033] Among them, the calculation formula of the channel attention module is:
[0034]
[0035] The calculation formula of spatial attention is:
[0036]
[0037] Among them, W0 and W1 are the weight values calculated in MLP, σ is the Sigmoid function, and F C max and F C avg They represent the two-branch parallel processing operations in the channel attention operation process, namely the maximum pooling and average pooling operations; the output M of the channel attention module is C (F) is multiplied with the input feature map F to obtain the channel-optimized feature F′; the feature map F′ is used as the input feature map of the spatial attention module, and the spatial attention module performs spatial weight distribution of feature information to obtain the feature map F″, ignoring the interference of the fading channel in the spatial dimension and emphasizing the original information of the modulated signal; f 7×7 Represents a convolution operation with a convolution kernel size of 7×7.
[0038] Optionally, the multi-head self-attention mechanism MHSA formula is as follows:
[0039] Q=FW q
[0040] K=FW k
[0041] V=FW ν
[0042]
[0043] The input feature F first passes through the LayerNorm layer (LN), and then multiplied by different weight matrices to obtain the query (Q), key (K), and value (V), which are then fed into the MHSA mechanism. Subsequently, by measuring the correlation between Q and K, the self-attention matrix mapping Attn is calculated to encapsulate the global information. Where W q , W k , W v is the weight matrix of the corresponding matrices of Q, K and V. k is the length of F. The MHSA mechanism performs transformations on Q, K, and V through h different linear transformations to compute different outputs. It is important to note that the computations for each head are performed in parallel.
[0044] The multi-head attention mechanism includes two parallel attention heads; each attention head has an independent projection weight matrix;
[0045] After extracting features based on the coding block trunk road side and the RTDC auxiliary branch side, the following feature fusion operations are also included:
[0046] For each attention head, multiply the query, key, and numeric vector with the corresponding weight matrix to obtain the query, key, and numeric vector corresponding to each attention head;
[0047] Perform attention calculation on each attention head to obtain the attention weights and weighted numerical vectors of multiple attention heads;
[0048] The weighted numerical vectors of multiple attention heads are connected or concatenated to obtain the global feature information of the multi-head attention output.
[0049] In the second aspect, a method for equalizing a fading signal is proposed, comprising:
[0050] Acquire test data to be processed, wherein the test data is a modulated signal carrying Rayleigh fading;
[0051] Inputting the test data into a channel equalization processing model for processing to obtain an equalized prediction signal;
[0052] The signal equalization processing model is trained based on the channel equalization model training method described in the first aspect.
[0053] In a third aspect, a channel equalization model training device is proposed, comprising:
[0054] A simulation module, used to determine an original signal, and simulate the original signal to generate a sample data set; wherein the sample data in the sample data set includes: a noise-free signal used as label data and a noise signal used as a training sample, wherein the noise signal is a modulated signal carrying the influence of Rayleigh fading;
[0055] A selection module is used to select an initial model based on a hybrid attention mechanism; the initial model includes: a dual-branch encoder and a single-branch decoder, wherein the encoding block trunk in the dual-branch encoder is used to extract detail spatial features, and the time residual dilation convolution RTDC auxiliary branch is used to construct temporal feature dependencies;
[0056] A processing module, used for inputting the training sample data in the sample data set into the initial model, extracting all-round feature information of the noise signal by using a dual-branch encoder, and upgrading the feature information by using a decoder, and outputting a reconstructed signal;
[0057] The training module is used to repeatedly iterate the initial model and adjust the learning rate to determine the model parameters based on the SmoothL1Loss loss value of the reconstructed signal and the noise-free signal in the training sample data, so as to obtain a trained channel equalization model.
[0058] In a fourth aspect, a device for equalizing a fading signal is proposed, comprising:
[0059] An acquisition module, used for acquiring test data to be processed, wherein the test data is a modulated signal carrying Rayleigh fading;
[0060] A prediction module, used for inputting the test data into a channel equalization processing model for processing to obtain a predicted signal after equalization;
[0061] The signal equalization processing model is trained based on the channel equalization model training method described in the first aspect.
[0062] In a fifth aspect, an electronic device is provided, comprising:
[0063] Processor; and
[0064] A memory arranged to store computer executable instructions, which, when executed, cause the processor to execute the steps of the channel equalization model training method described in the first aspect, and / or the steps of the equalization processing method for fading signals described in the second aspect.
[0065] In the sixth aspect, a computer-readable storage medium is proposed, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple applications, the electronic device executes the steps of the channel equalization model training method described in the first aspect, and / or executes the steps of the equalization processing method of the fading signal described in the second aspect.
[0066] It can be seen from the technical solutions provided in the above embodiments of this specification that deep separable convolution, causal time dilation convolution, transposed convolution, etc. are selected to build the balanced autoencoder DSTReA-HA as an equalization model. In DSTReA-HA, the encoder adopts a dual-branch parallel structure to extract all-round features of the fading signal. The backbone network has a coding block composed of 4 stacked deep separable convolution cascade SE mechanisms and an additional CBAM jump connection branch, which deeply extracts the spatial and channel characteristics of the fading signal; the auxiliary branch is composed of 3 layers of DSC cascaded 4 layers of TCN and CBAM, which mainly extracts the dimension of the timing dependency of the fading signal. Finally, the dual-branch features are fused through a multi-head self-attention mechanism to ensure the complete transmission of the modulated signal feature information. The decoder of DSTReA-HA is relatively simple, mainly composed of a decoding block of 3 transposed convolution cascade SE modules, which completes the upsampling of the feature information from the MHSA mechanism, gradually restores the original signal dimension, and finally outputs it through a 3×3 convolution layer. The deep separable convolution introduced in this application can greatly reduce the complexity of the model while ensuring the performance of the model. The constructed encoding block decoding block fusion SE mechanism can deeply extract the channel attention features of the fading signal. The cascaded RTDC branch makes up for the shortcomings of insufficient timing information extraction of the backbone network. MHSA promotes the fusion and importance distribution of encoder extracted features. Finally, in the training stage, the SmoothL1Loss loss function is called to calculate the gap between the output layer prediction value and the true value of the label to adjust the network model parameters, ensuring the accuracy of the model reconstruction signal, improving the training accuracy of the model, and the accuracy of subsequent equalization processing. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0068] Figure 1 This is a schematic diagram of the steps of a channel equalization model training method provided for Example 1 of the present application.
[0069] Figure 2a It is a schematic diagram of the steps of simulating and generating a sample data set provided by an embodiment of this specification.
[0070] Figure 2b It is a flowchart of generating sample data by simulation provided by an embodiment of this specification.
[0071] Figure 3It is a schematic diagram of the steps of an equalization processing method for a fading signal provided in an embodiment of the present application.
[0072] Figure 4 It is a schematic diagram of the equalization processing flow of the fading signal provided by this application.
[0073] Figure 5 It is a structural diagram of the autoencoder based on the hybrid attention mechanism provided in this application.
[0074] Figure 6 It is a structural diagram of the CBAM attention layer provided in this application.
[0075] Figure 7 It is a schematic diagram of the structure of the residual time dilation convolution provided in this application.
[0076] Figure 8 It is a structural diagram of the multi-head self-attention mechanism MHSA provided in this application.
[0077] Fig. 9 It is a structural diagram of a channel equalization model training device provided in an embodiment of the present application.
[0078] Fig.10 It is a structural schematic diagram of an equalization processing device for fading signals provided in an embodiment of the present application.
[0079] Fig.11 It is a schematic diagram of the structure of an electronic device provided by an embodiment of this specification. DETAILED DESCRIPTION
[0080] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.
[0081] With the vigorous development of artificial intelligence in recent years, machine learning has become a hot topic and has been widely studied and applied. As a development and branch of machine learning, deep learning has achieved a lot of results in image processing, natural language processing and other fields with its convolutional neural network and recurrent neural network. In addition, in some unsupervised learning tasks, deep learning has also achieved results that surpass traditional methods, such as image restoration, style transfer, image enhancement, super-resolution, etc. At present, relevant scholars have used deep learning technology in the field of communications to solve some problems such as channel estimation, modulation recognition and resource allocation, and have achieved good results and performance. Therefore, intelligent and autonomous deep learning technology can bring a new solution to channel equalization technology.
[0082] The core idea of channel equalization in wireless communication systems is to mine the characteristic information of modulated signals under severe Rayleigh fading to reconstruct the signal and restore the original waveform of the signal. Deep learning can form more abstract high-level representation attribute categories or features by combining low-level features to discover the distributed feature representation of data. At present, the equalization of complex fading channels requires multi-dimensional parameter influences, or different networks are required to achieve different training effects. Therefore, a time-residual autoencoder with a hybrid attention mechanism is proposed to solve the equalization problem under Rayleigh fading channels.
[0083] The inventive concept of the present application is: an equalizing autoencoder of a depthwise separable temporal residual equalizer with hybrid attention mechanisms (DSTREA-HA) is designed based on a deep separable convolution (DSC) cascade compression-excitation (SE) encoding block and a transposed convolution cascade compression-excitation (SE) decoding block as the main part, and a temporal residual dilated convolution (RTDC) branch as the auxiliary part, and a multi-head self-attention mechanism (MHSA) is called to help the feature fusion of the equalizing network model, so that the received IQ signal is fully extracted from the time domain information and frequency domain features, and then the original signal is reconstructed to achieve the purpose of equalization.
[0084] This application mainly includes two parts: 1. A method for training a channel equalization model; 2. A method for equalizing a test signal based on the trained channel equalization model.
[0085] Among them, the main contents of the training method include: determining the original signal, and simulating the original signal to generate a sample data set; selecting an initial model based on a hybrid attention mechanism; inputting the training sample data in the sample data set into the initial model, using a dual-branch encoder to extract all-round feature information of the noise signal, and using a decoder to upgrade the feature information and output a reconstructed signal; based on the SmoothL1Loss loss value of the reconstructed signal and the noise-free signal in the training sample data, repeatedly iterating the initial model, adjusting the learning rate to determine the model parameters, and obtaining a trained channel equalization model.
[0086] The main contents of the equalization processing method include: obtaining test data to be processed, wherein the test data is a modulated signal carrying Rayleigh fading; inputting the test data into a channel equalization processing model for processing to obtain a predicted signal after equalization; wherein the signal equalization processing model is trained based on the above-mentioned channel equalization model training method.
[0087] Through the technical solution of the present application, signal reconstruction can be efficiently realized, the purpose of channel equalization can be achieved, and the model training speed can be increased, thereby improving the subsequent signal transmission efficiency.
[0088] The scheme involved in this application is described in detail below through specific embodiments.
[0089] Embodiment 1
[0090] Figure 1 The figure is a schematic diagram of the steps of a channel equalization model training method provided in the first embodiment of the present application. The execution subject of the method can be a software module or a hardware device, such as a cloud host, a computer, a tablet computer, a mobile terminal device, a smart wearable device, etc. The channel equalization model training method may include the following steps:
[0091] Step 102: determine the original signal, and simulate the original signal to generate a sample data set; wherein the sample data in the sample data set includes: a noise-free signal used as label data and a noise signal used as a training sample, and the noise signal is a modulated signal carrying the influence of Rayleigh fading.
[0092] Optionally, refer to Figure 2a As shown, the present application may specifically include the following steps when determining the original signal and simulating the original signal to generate a sample data set:
[0093] Step 202: Randomly generate a binary bit stream and simulate the original signal to be transmitted.
[0094] Specifically, combined Figure 2bAs shown, a binary bit stream is randomly generated to simulate the original data to be transmitted, and 8 different noise-free modulation signals are simulated according to 8 different modulation modes; it should be understood that the 8 types here are just for example. In different business scenarios, different types and numbers of modulation modes can be set according to different business needs, and there is no limitation here. Among them, the simulation tool can be MATLAB2020b, and the 8 modulation modes can be: bpsk, qpsk, 8psk, gmsk, cpfsk, oqpsk, 16qam, 64qam. Call functions in MATLAB such as comm.BPSKModulator, comm.CPFSKModulator, etc. And set the oversampling rate to 8 and the sampling frequency to 1024kHz.
[0095] Step 204: Based on the original signal, simulate modulation according to a preset plurality of different modulation modes to generate a corresponding plurality of different noise-free signals, and perform a preprocessing operation on the noise-free signals, wherein the preprocessing operation at least includes: waveform adjustment and power normalization processing.
[0096] In specific implementation, a square root raised cosine filter can be used to adjust the waveform of the modulated signal, and the edge data generated can be "cut off" to ensure the smoothness of the sample waveform; in order to facilitate the addition of fading and Gaussian white noise, the filtered signal needs to be power normalized to ensure the consistency of the sample power;
[0097] The filter formula is as follows:
[0098]
[0099] Where a is the roll-off coefficient in the root raised cosine filter, t is the sampling time, and T is the symbol period. The modulated signal is processed by the filter g, that is, the corresponding 0101... bit stream is modulated onto the corresponding waveform before it can be transmitted by radio frequency.
[0100] The power normalization formula is as follows:
[0101]
[0102] In the formula, N represents the length of the signal; x(t) represents the input modulated signal that needs to be power normalized, and x'(t) represents the modulated signal after the power normalization.
[0103] Step 206: Simulate a Rayleigh fading channel, simulate a noise-free signal after preprocessing, a fading signal transmitted through the Rayleigh fading channel, and simulate a Gaussian white noise channel to add noise power to the fading signal to obtain a noise signal.
[0104] In the implementation case of the present application, a fading channel is first created. According to the comm.RayleighChannel tool provided by MATLAB, a six-path fading channel can be simulated and generated, in which the path delay of each fading channel is [0200800120023003700]*1e-8 (seconds) and the average fading path gain is [0-0.9-4.9-8-7.8-23.9] (dB).
[0105] At the same time, using the MATLAB environment, a Gaussian white noise signal is simulated, and a noise power with a signal-to-noise ratio of -12dB to 8dB is added to the fading signal with an interval of 2dB. Thus, the required noise signal is obtained.
[0106] Step 208: Aggregate the multiple different noise-free signals and their corresponding noise signals into a sample data set.
[0107] Finally, the real and imaginary parts of the generated IQ dataset are separated to create a sample dataset of size 2×1024. This sample dataset is used to pass into the subsequent initial model composed of the autoencoder for training.
[0108] It should be understood that in this application, in addition to using Figure 2a In addition to generating sample data by the method steps shown, sample data that meets the above requirements can also be generated by other existing methods, which will not be described in detail in this application.
[0109] Step 104: Select an initial model based on a hybrid attention mechanism; the initial model includes: a dual-branch encoder and a single-branch decoder, wherein the encoding block main branch in the dual-branch encoder is used to extract detailed spatial features, and the time residual dilated convolution RTDC auxiliary branch is used to construct temporal feature dependencies.
[0110] Optionally, the initial model is an autoencoder; the autoencoder based on the hybrid attention mechanism includes: a dual-branch encoder, a multi-head self-attention mechanism MHSA fusion feature module, and a single-branch decoder; wherein the dual-branch encoder includes: a coding block main path and an RTDC auxiliary branch, and the coding block main path includes 4 coding blocks; the single-branch decoder includes 3 stacked decoding blocks; the dual-branch encoder is composed of the first convolution layer and 4 coding blocks connected in sequence, and a residual connection of the CBAM attention mechanism is set between the first convolution layer and the third coding block, and a residual connection is set in the first layer and the third layer inside the coding block and after the last scaling operation; the single-branch decoder is composed of 3 decoding blocks and a transposed convolution layer in series, a residual connection is set between the first decoding block and the second decoding block, and a residual connection is set in the first layer and the third layer inside the decoding block and after the last scaling operation.
[0111] Step 106: input the training sample data in the sample data set into the initial model, use a dual-branch encoder to extract all-round feature information of the noise signal, and use a decoder to upgrade the feature information to output a reconstructed signal.
[0112] Optionally, the coding block includes: a deep separable convolution DSC, and a 3-layer DSC cascade compression-excitation (SE) attention mechanism, wherein each layer of DSC is followed by a batch normalization layer and an activation function layer, and the SE module is composed of a pooling layer, a first convolution layer, a LeakyReLU activation function, a second convolution layer, and a Sigmod activation function; on the main road side of the coding block, when a dual-branch encoder is used to extract all-round feature information of the noise signal, it specifically includes:
[0113] The first layer of DSC is used to expand the channel dimension of the feature information of the noise signal input to the coding block to obtain the corresponding feature map; the second layer of DSC is used to downsample the feature map by controlling the step size and padding; the third layer of DSC is used to further adjust the number of channels of the input feature map; the pooling layer and the first convolution layer are used to compress the input feature map to compress the feature dimension; the activation function and the second convolution layer are used to excite the compressed features, transform the feature dimension, and assign weights to obtain the feature weight size.
[0114] In the present application, the ordinary convolution inside the encoding block is replaced by the deep separable convolution DSC, and the decoding block is composed of three layers of DSC cascade SE attention mechanism connected thereto, where each layer of DSC is followed by a batch normalization layer BatchNormalization and an activation function LeakyReLU layer, and the SE module is composed of pooling, the first convolution layer, the LeakyReLU activation function, the second convolution layer, and the Sigmod activation function. The specific formula of the deep separable convolution function Dsc is as follows:
[0115]
[0116] Among them, K represents the size of D k ×D k ×m deep convolution kernel, F is the feature map input to Dsc, when i and j∈(0, D k ), the feature map F is subjected to the corresponding convolution summation operation. In this configuration, each filter mth in K is applied to the corresponding channel mth in the input feature map F, thereby generating the mth channel in the filtered output feature map DSC.
[0117] The first layer DSC is used to expand the channel dimension of the feature information input to the coding block; the second layer DSC is used to downsample the input feature map by controlling the step size and padding; the third layer DSC is used to further adjust the number of channels of the input feature map; the pooling layer and the third convolutional layer are used to compress the input feature map to compress the feature dimension to D k ×1×1; use the activation function and the fourth convolutional layer to stimulate the compressed features and transform the dimension to D k / r×1×1 in the weighted value to obtain the feature weight size D k ×1×1.
[0118] Optionally, the RTDC auxiliary branch is composed of 3 layers of DSC cascaded 4 layers of time dilation convolution and CBAM; on the RTDC auxiliary branch side, when a dual-branch encoder is used to extract all-round feature information of the noise signal, it specifically includes:
[0119] The feature information of the input noise signal is respectively expanded and reorganized through DSC with kernel size of 3×3, 1×2, and 3×3, and the feature map channel and spatial dimension are compressed. The feature dimension is matched once based on 2×4 DSC and the residual operation is performed with the compressed output. The temporal features obtained by the residual operation are sequentially convolved N times: using a kernel size of 1×3 and a dilation rate of 2 N-1 The time dilation convolution TCN is padded and downsampled, where N is a positive integer; the CBAM mechanism is used to assign feature weights to the extracted time series feature information.
[0120] In the present application, the RTDC branch introduced by the coding block auxiliary branch performs timing feature extraction. The formula is as follows:
[0121]
[0122] In the formula, the input sequence of TCN is represented by xi, and the convolution kernel of TCN is represented by k. Where d represents the expansion rate, f d Represents the convolution kernel of the dilated causal convolution. The items x in the input sequence t-di represents the value of the modulating signal at time t-di, illustrating that the dilated convolution operation includes earlier time instances of the modulating signal samples.
[0123] When performing specific extraction processing, 3×3 depth-separable convolution can be used to expand the channel of the input feature; 1×2 depth-separable convolution can be used to reorganize the length and width dimensions of the feature map; 3×3 depth-separable convolution can be used to further compress the channel and spatial dimensions of the feature map; 2×4 depth-separable convolution can be used to match the input feature dimension once and perform residual operation with the output of the previous step; 1×3 time dilation convolution with a dilation rate of 1 and a convolution operation with an interval of 1 for TCN noise reduction can be used; 1×3 time dilation convolution with a dilation rate of 2 and a convolution operation with an interval of 2 for TCN noise reduction can be used, which is doubled compared to the dilation dimension in the previous step, and the receptive field of the model is doubled accordingly, which can extract time features of different dimensions. The dilation rates of 4 and 8 have the same effect as above; the CBAM mechanism is used to allocate important feature weights to the extracted temporal feature information.
[0124] The above multi-level dense residual connections solve the problem of network degradation and ensure the smooth transmission of original information.
[0125] Optionally, the formula of the CBAM is:
[0126]
[0127] Among them, the calculation formula of the channel attention module is:
[0128]
[0129] The calculation formula of spatial attention is:
[0130]
[0131] Among them, W0 and W1 are the weight values calculated in MLP, σ is the Sigmoid function, and F C max and F C avg They represent the two-branch parallel processing operations in the channel attention operation process, namely the maximum pooling and average pooling operations. First, we take the output M of the channel attention module C (F) is multiplied with the input feature map F to obtain the channel-optimized feature F′. Secondly, the feature map F′ is used as the input feature map of the spatial attention module, and the spatial attention module performs spatial weight distribution of feature information to obtain the feature map F″, ignoring the interference of the fading channel in the spatial dimension and emphasizing the original information of the modulated signal; f 7×7 Represents a convolution operation with a convolution kernel size of 7×7.
[0132] Optionally, the multi-head self-attention mechanism MHSA formula is as follows:
[0133] Q=FWq
[0134] K=FW k
[0135] V=FW ν
[0136]
[0137] The input feature F first passes through the LayerNorm layer (LN), and then multiplied by different weight matrices to obtain the query (Q), key (K), and value (V), which are then fed into the MHSA mechanism. Subsequently, by measuring the correlation between Q and K, the self-attention matrix mapping Attn is calculated to encapsulate the global information. Where W q , W k , W v is the weight matrix of the corresponding matrices of Q, K and V. k is the length of F. The MHSA mechanism performs transformations on Q, K, and V through h different linear transformations to compute different outputs. It is important to note that the computations for each head are performed in parallel.
[0138] The multi-head attention mechanism includes two parallel attention heads; each attention head has an independent projection weight matrix;
[0139] After extracting features based on the main road side of the coding block and the auxiliary branch side of the RTDC, the following feature fusion operations are also included: for each attention head, the query, key, and numerical vector are multiplied with the corresponding weight matrix to obtain the query, key, and numerical vector corresponding to each attention head; attention calculation is performed on each attention head to obtain the attention weights and weighted numerical vectors of multiple attention heads; the weighted numerical vectors of multiple attention heads are connected or spliced to obtain the global feature information output by the multi-head attention.
[0140] Step 108: Based on the SmoothL1Loss loss value of the reconstructed signal and the noise-free signal in the training sample data, the initial model is repeatedly iterated, and the learning rate is adjusted to determine the model parameters to obtain a trained channel equalization model.
[0141] In the present application scheme, an autoencoder method based on a hybrid attention mechanism is constructed, and a training set is input for iteration, and the learning rate is adjusted for network training. The autoencoder is composed of an encoder and a decoder, and the convolutional layer is first connected for signal input and output. The encoder is used to extract all aspects of the fading signal features, and the decoder is used to upgrade the dimension of the feature information. The signal reconstruction is completed and finally output by the output layer. The SmoothL1Loss function is used to determine the model parameters by outputting the loss value of the non-fading signal corresponding to the data set.
[0142] In this case, you need to first select the balanced autoencoder as the training model, and use the data set samples generated in step 1 to package the training samples with fading noise into a fixed-size (batchsize×1×2×1024) tensor. The tensor size is expanded through the first convolution layer (input layer) of the autoencoder, and finally the tensor size is restored through the last convolution layer (output layer). The output tensor data uses the loss function SmoothL1Loss to determine the gap with the label data, and then the Adam optimizer is used to determine the next round of model training direction. Repeat this step until the loss value drops to a certain range, that is, the predicted value approaches the label value, achieving a balanced effect. The SmoothL1Loss function formula SLL() is as follows:
[0143]
[0144] Where N represents the length of the input signal, and the variable z i is defined as follows:
[0145]
[0146] Where S i is the true value, S i ″ is the predicted value of the fading signal obtained by the equalizer.
[0147] Reference Figure 3 As shown, it is a schematic diagram of the steps of an equalization processing method for fading signals provided in an embodiment of the present application. The execution subject of the method can be a software module or a hardware device, such as a cloud host, a computer, a tablet computer, a mobile terminal device, a smart wearable device, etc.; the channel equalization model training method can include the following steps:
[0148] Step 302: Acquire test data to be processed, wherein the test data is a modulated signal carrying Rayleigh fading.
[0149] The acquisition of the test data may refer to step 102, which will not be described in detail here.
[0150] Step 304: Input the test data into the channel equalization processing model for processing to obtain a predicted signal after equalization; wherein the signal equalization processing model is trained based on the channel equalization model training method described in steps 102 to 108.
[0151] When implementing it, you can load Figure 1The model training method shown saves the model parameters, and inputs the modulated signal to be equalized in the test set (test data) into the trained equalized autoencoder, automatically outputs the original signal prediction value after the modulated signal is equalized, so as to filter out the influence of Rayleigh fading and achieve the purpose of equalization. After the optimal model parameter set path is determined by using the training samples and the verification samples, for the test samples, the data is directly operated through the determined path set to directly obtain the signal after network equalization.
[0152] Figure 4 As shown, the equalization processing flow of the fading signal provided by the present application includes: a model training process and a prediction process.
[0153] The model training process may include:
[0154] -Simulation to generate sample data sets
[0155] The data set samples of modulated signals with Rayleigh fading are generated by MATLAB software simulation, and the samples contain the original noise-free signal as label data and the signal with fading effect as processing training samples. In addition to the influence of Rayleigh fading, an additional additive white Gaussian noise (AWGN) channel is added, the signal-to-noise ratio is -12dB-8dB, the interval is 2dB, and each type of modulated signal has 2000 data under each signal-to-noise ratio condition, which is divided into training set, validation set and test set in a ratio of 8:1:1. The real and imaginary parts of the complex signal data are separated and constructed as an input tensor of size 2×1024.
[0156] -Training set input autoencoder for training
[0157] Reference Figure 5 As shown in the figure, the autoencoder based on the hybrid attention mechanism mainly includes 4 encoding blocks of the backbone network, the auxiliary branch residual time dilation convolution RTDC, the fusion feature module multi-head self-attention mechanism MHSA and the decoder - three stacked decoding blocks. In the training phase, the signal with fading noise is input into this model and the model parameters close to the label are obtained through iteration. In the batch data equalization phase, the parameters are loaded into the model to calculate the equalization result.
[0158] The main branch of the encoding block includes four encoding blocks stacked in sequence. A residual connection with a CBAM attention mechanism is added between the first and third encoding blocks. The encoding block consists of three layers of depth-separable convolution DSC, and its complexity is roughly one-ninth of the standard convolution CNN. The first and third layers of DSC kernel sizes are both 3×3, and the second layer of DSC kernel will be adjusted according to demand to complete the dimensionality reduction of the feature map. Figure 5Considering the space reasons, the batch standard layer BatchNorm and activation function LeakyReLU included in each DSC are omitted and not drawn. The compression excitation mechanism SE is cascaded with it. The output of the third layer DSC is passed to the SE module to calculate the weight of the spatial features, and then the weight distribution is completed with the output of the third layer DSC through the scaling function, which can resist the influence of the fading channel to a certain extent. Finally, the residual operation is performed with the initial input to ensure the integrity of the original information and avoid network degradation problems.
[0159] The formula of the coding block can be expressed as follows:
[0160]
[0161] Among them, k1, k2, and k3 represent the convolution kernel sizes of the first three layers of depth-wise separable convolution DSC layers in the coding block, Add() is a point-by-point addition function, and Scale() represents the scaling point convolution operation. C Represents the output feature map after three layers of DSC convolution; X C is the signal feature map input to the encoding block. C V C The weight matrix with channel features after residual compression-excitation attention mechanism (SE) is expressed as follows:
[0162] s c =F ex (g c , W) = σ(W2δ(W1g c ))
[0163] Scale(V c ,s c )=s c V c
[0164] g c It represents the feature map after global average pooling, and its dimension is number of channels × 1 × 1. The LeakyReLU function with slope parameter α = 0.001 is denoted as δ, and the sigmoid activation function is denoted as σ; W1 and W2 are the weight values after SE module excitation and compression.
[0165] In this example, the 32 input layers use 3×3 convolution kernels to transform the input tensor dimension to batchsize×32×2×1024, and then the first three encoding blocks transform the feature map dimension to obtain a feature map of batchsize×128×2×512. In order to match the dimension of the residual operation, the feature map obtained by the input layer is processed by a layer of DSC to reduce the dimension and then passed to CBAM for channel attention and spatial attention extraction. The fourth encoding block reduces the feature dimension to batchsize×128×1×256.
[0166] Reference Figure 6 As shown in Figure 1, the CBAM attention layer consists of a channel attention module and a spatial attention module in series. The input data is F, which is a tensor of shape 128×2×512, where 128 is the number of channels, 2 is the height, and 512 is the width. The entire attention process can be summarized as:
[0167]
[0168] CBAM obtains a one-dimensional channel attention feature map F' and a two-dimensional spatial attention feature map F" through the following steps.
[0169] The first is the channel attention module: the input data F is passed through a global average pooling layer to compress the feature map of each channel into a scalar value. The pooled feature map is sent to two fully connected layers, which are used to learn the weights and biases of the channels respectively. For each channel, its corresponding weight and bias are applied to the original feature map to obtain the weighted feature map. Finally, the weighted feature map is passed through a Sigmoid activation function to limit its value between 0 and 1. This feature map is called a one-dimensional channel attention feature map F'.
[0170] Next is the spatial attention module: the one-dimensional channel attention feature map F' is passed through a global maximum pooling layer to compress the feature map of each channel into a scalar value. The pooled feature map is fed into two convolutional layers, which are used to learn the weights and biases of spatial attention respectively. The weights and biases of each channel are applied to the original feature map to obtain the weighted feature map. Finally, the weighted feature map is passed through a Sigmoid activation function to limit its value between 0 and 1. This feature map is called the two-dimensional spatial attention feature map F".
[0171] Reference Figure 7 As shown in the figure, the residual time dilation convolution consists of three layers of DSC cascaded with four layers of time dilation convolution and CBAM, which is used to extract the temporal dependencies in the fading signal. The input features are sequentially compressed by three layers of DSC with kernel sizes of 3×3, 1×2, and 3×3, and a 2×4 DSC residual connection is used to prevent gradient vanishing and gradient explosion.
[0172] The kernel size of the four-layer time dilation convolution TCN is 1×3, and the dilation rates are [1, 2, 4, 8]. The input data is a time series represented by X = [x1, x2, x3, ..., x n ], where x i is the input of time step i. For each dilation rate, TCN uses a convolutional layer to process the input data. Assuming the current dilation rate is d, the input data is padded according to the current dilation rate d. The purpose of padding is to keep the output length after convolution consistent with the input length. Padding can use zero padding or periodic padding. Convolution operation is performed on the padded input data. The activation function LeakyReLU is applied to the result of the convolution operation. After that, the pooling operation and stride are used to reduce the length of the convolution output to reduce the computational cost and increase the receptive field of the model. Repeat the above steps, convolve and downsample the data using different dilation rates, and multiple convolutional layers can be used to increase the depth and complexity of the model. The features after the output of the 4-layer TCN are sent to CBAM for further feature extraction.
[0173] The specific formula of TCN is as follows:
[0174]
[0175] Where d is the expansion rate, f d Represents the convolution kernel of the dilated causal convolution. The items x in the input sequence t-di Represents the value of the modulated signal at time t-di.
[0176] Reference Figure 8 As shown in Figure 2, in this example, the features extracted by the two parallel branches need to be further integrated. Therefore, the multi-head self-attention mechanism MHSA is introduced to complete this task. The structure of MHSA is shown in Figure 2. Figure 8 shown.
[0177] The input sequence X is of size batchsize×128×1×256, which contains Rayleigh fading feature signals. The input sequence X is linearly transformed and mapped to the query (Q), key (K), and value (V) spaces of multiple attention heads. This can be done by respectively transforming X with the weight matrix W Q , W K and W V The dimensions of these weight matrices are consistent with X. Then, the mapped query (Q), key (K), and value (V) are divided into multiple attention heads. In this example, considering performance and complexity, h = 2 attention heads are designed. Then for each attention head i (i ranges from 1 to 4), the query (Q i ), key (K i) and value (V i ). Scaled dot-product attention operation: For each attention head i, a scaled dot-product attention operation is performed.
[0178] The steps MHSA takes include:
[0179] ① Calculate the attention weight: By taking the query (Q i ) and key (K i ) performs a dot product operation, scales the result, and then applies the softmax function to obtain the attention weight matrix (attention matrix) A i .
[0180] ② Weighted sum: The attention weight matrix A i and value (V i ) performs matrix multiplication to obtain the output matrix O of attention head i i .
[0181] ③Multi-head cascade: The output matrix O of each attention head i Connected in series, we get the final output matrix O of the multi-head self-attention mechanism.
[0182] ④ Linear transformation: Perform a linear transformation on the output matrix O of the multi-head self-attention mechanism and map it back to the original feature space. This can be done by comparing it with the weight matrix W O Multiply to achieve.
[0183] ⑤Output: The output matrix after linear transformation is obtained, which represents the signal characteristics after being processed by the multi-head self-attention mechanism.
[0184] Through the above processing, MHSA can simultaneously focus on different positions and features in the input sequence, and use attention weights to capture the global feature information of Rayleigh fading signals. Each attention head can learn different attention patterns to extract useful feature representations and perform equalization on the signal to improve the quality of the received signal. The formula is as follows:
[0185] Q=FW q
[0186] K=FW k
[0187] V=FW ν
[0188]
[0189] The input feature F first passes through the LayerNorm layer (LN), and then multiplied by different weight matrices to obtain the query (Q), key (K), and value (V), which are then fed into the MHSA mechanism. Subsequently, by measuring the correlation between Q and K, the self-attention matrix mapping Attn is calculated to encapsulate the global information. Where W q , W k , W v is the weight matrix of the corresponding matrices of Q, K and V. k is the length of F. The MHSA mechanism transforms Q, K, and V through h different linear transformations to calculate different outputs. It is important to note that the calculation of each head is performed in parallel. The multi-head attention mechanism contains 2 parallel attention heads; each attention head has an independent projection weight matrix.
[0190] Finally, the characteristic information of the fading signal after MHSA fusion is passed through three decoding blocks to reconstruct the signal and restore the waveform. The structure of the decoding block is referenced Figure 5 Each decoding block is connected in sequence by the third convolution layer, transposed convolution, and the fourth convolution layer. The post-cascade SE mechanism extracts the spatial feature weights, and then performs residual operations on the residual connection branches formed between the third and fourth convolutions and with the fifth convolution after the scaling operation. Transposed convolution is introduced in the decoding block to complete the upsampling process. Transposed convolution can be regarded as the inverse operation of the convolution layer. It can effectively filter out the influence of fading noise by increasing the dimension through the extracted feature information.
[0191] ---Training to obtain a balanced autoencoder
[0192] The prediction process includes:
[0193] --Determine the test data set
[0194] The test data set can be obtained from the pre-allocated test set in the above sample data set. After the optimal parameter set path of the model is determined using the training samples and the validation samples, for the test samples, the data is directly operated through the determined path set to directly obtain the signal after network equalization.
[0195] -- Input balanced autoencoder for testing
[0196] Load the model parameters saved in the previous step, input the modulated signal to be equalized in the test set into the trained equalized autoencoder, and automatically output the predicted value of the original signal after the modulated signal is equalized to filter out the influence of Rayleigh fading and achieve the purpose of equalization.
[0197] -- Input reconstruction signal
[0198] Through the above technical solutions, the hybrid attention mechanism autoencoder method has certain advantages over the current mainstream methods: the generalization ability of this method is relatively strong, and the equalization performance of the residual time autoencoder based on the hybrid attention mechanism under 8 types of modulation signals is better than other noise reduction and equalization models in terms of improving signal-to-noise ratio, reducing error vector amplitude, and demodulation bit error rate. Based on the hybrid attention mechanism autoencoder method, the feature map of the one-dimensional IQ modulated signal data can be directly extracted without secondary changes, and compared with the image type data, it has smaller complexity and fewer floating point operations (FLOPs), better generalization, and better deployment in practical applications. Moreover, taking advantage of the fact that CNN is good at extracting detailed feature information and the RNN type network can establish time dependencies, an autoencoder that can fully extract spatial and temporal features is designed. On this basis, the ordinary CNN is replaced with a deep separable convolution DSC, which greatly reduces the model parameters and improves the training speed while maintaining the performance. At the same time, the residual time dilation convolution branch is introduced. By using causal time dilation convolution with a dilation rate of [1, 2, 4, 8], the receptive field of the model can be expanded while reducing the amount of calculation, and the temporal feature information of the modulated signal under the influence of the fading channel can be captured, thereby improving the anti-fading characteristics of the model. The CBAM and MHSA attention mechanisms are introduced in the encoder to realize the weight allocation of important features and weaken the interference of the fading channel. In addition, the multi-head parallel calculation method can improve the efficiency of feature extraction.
[0199] Embodiment 2
[0200] Fig. 9The schematic diagram of the structure of a channel equalization model training device provided in an embodiment of the present application is as follows: The device 900 comprises: a simulation module 901 , a selection module 902 , a processing module 903 , and a training module 904 . Among them, the simulation module 901 is used to determine the original signal and simulate the original signal to generate a sample data set; wherein the sample data in the sample data set includes: a noise-free signal used as label data and a noise signal used as a training sample, and the noise signal is a modulated signal carrying the influence of Rayleigh fading; the selection module 902 is used to select an initial model based on a hybrid attention mechanism; the initial model includes: a dual-branch encoder and a single-branch decoder, wherein the encoding block trunk in the dual-branch encoder is used to extract detail spatial features, and the time residual dilation convolution RTDC auxiliary branch is used to construct a temporal feature dependency relationship; the processing module 903 is used to input the training sample data in the sample data set into the initial model, use the dual-branch encoder to extract all-round feature information of the noise signal, and use the decoder to upgrade the feature information, and output a reconstructed signal; the training module 904 is used to repeatedly iterate the initial model based on the SmoothL1Loss loss value of the reconstructed signal and the noise-free signal in the training sample data, and adjust the learning rate to determine the model parameters to obtain a trained channel equalization model.
[0201] Optionally, as an embodiment, when the simulation module 901 determines the original signal and simulates the original signal to generate a sample data set, it is specifically used to randomly generate a binary bit stream and simulate the original signal to be transmitted; based on the original signal, simulate modulation according to a preset plurality of different modulation modes to generate a corresponding plurality of different noise-free signals, and perform preprocessing operations on the noise-free signal, wherein the preprocessing operations at least include: waveform adjustment and power normalization processing; simulate a Rayleigh fading channel, and simulate the noise-free signal after the preprocessing operation, the fading signal transmitted through the Rayleigh fading channel, and simulate a Gaussian white noise channel to add noise power to the fading signal to obtain a noise signal; and summarize the plurality of different noise-free signals and their corresponding noise signals into a sample data set.
[0202] The types of signals in the sample dataset are as follows:
[0203] (1) BPSK (Binary Phase Shift Keying): BPSK is a basic digital modulation technique used to modulate a digital bit stream into a sinusoidal signal. It uses two phase states to represent two binary values (usually 0 and 1), with the phase of one state being 180 degrees different from the phase of the other state.
[0204] (2) QPSK (Quadrature Phase Shift Keying): QPSK is a common digital modulation technique used to modulate two bits simultaneously into a sine wave signal. It uses four different phase states (0 degrees, 90 degrees, 180 degrees, and 270 degrees) to represent each bit combination.
[0205] (3) 8PSK (8-Phase ShiftKeying): 8PSK is a digital modulation technique used to modulate three bits simultaneously into a sine wave signal. It uses eight different phase states to represent each bit combination.
[0206] (4) OQPSK (Offset Quadrature Phase Shift Keying) is a digital modulation technique that is an improved version of QPSK. In OQPSK, the phase difference between adjacent symbols is 90 degrees. Unlike QPSK, its phase change occurs at the edge of the symbol instead of the center of the symbol.
[0207] (5) GMSK (Gaussian Minimum Shift Keying): GMSK is a frequency modulation technique used to convert digital signals into continuous sinusoidal signals. It uses a Gaussian filter to achieve continuous frequency changes to reduce spectrum spread and interference.
[0208] (6) CPFSK (Continuous Phase Frequency Shift Keying): CPFSK is a phase-continuous frequency modulation technique used to convert digital signals into analog signals. It represents different digital bits by keying at different frequencies while maintaining phase continuity.
[0209] (7) 16QAM (16-Quadrature Amplitude Modulation): 16QAM is an amplitude modulation technique that modulates four bits simultaneously into a sine wave signal. It uses four different amplitude levels and four different phase states to represent each bit combination.
[0210] (8) 64QAM (64-Quadrature Amplitude Modulation): 64QAM is an amplitude modulation technique that modulates six bits simultaneously into a sine wave signal. It uses six different amplitude levels and four different phase states to represent each bit combination.
[0211] The number of samples of each type of modulation signal at each signal-to-noise ratio is 2000, and they are divided into training set, validation set and test set in a ratio of 8:1:1.
[0212] In a specific implementation of the embodiments of the present specification, the initial model is an autoencoder; the autoencoder based on the hybrid attention mechanism includes: a dual-branch encoder, a multi-head self-attention mechanism MHSA fusion feature module, and a single-branch decoder; wherein the dual-branch encoder includes: a coding block main path and an RTDC auxiliary branch, and the coding block main path includes 4 coding blocks; the single-branch decoder includes 3 stacked decoding blocks; the dual-branch encoder is composed of a first convolutional layer and 4 coding blocks connected in sequence, and a residual connection of the CBAM attention mechanism is set between the first convolutional layer and the third coding block, and a residual connection is set in the first layer and the third layer inside the coding block and after the last scaling operation; the single-branch decoder is composed of 3 decoding blocks and a transposed convolutional layer in series, a residual connection is set between the first decoding block and the second decoding block, and a residual connection is set in the first layer and the third layer inside the decoding block and after the last scaling operation.
[0213] In another specific implementation of the embodiments of the present specification, the coding block includes: a deep separable convolution DSC, and a 3-layer DSC cascade SE attention mechanism, wherein each layer of DSC is followed by a batch normalization layer and an activation function layer, and the compression-excitation SE module is composed of a pooling layer, a first convolutional layer, a LeakyReLU activation function, a second convolutional layer and a Sigmod activation function; on the main road side of the coding block, when the processing module 903 uses a dual-branch encoder to extract all-round feature information of the noise signal, it is specifically used to use the first layer of DSC to expand the channel dimension of the feature information of the noise signal input to the coding block to obtain the corresponding feature map; use the second layer of DSC to downsample the feature map by controlling the step size and padding; use the third layer of DSC to further adjust the number of channels of the input feature map; use the pooling layer and the first convolutional layer to compress the input feature map to compress the feature dimension; use the activation function and the second convolutional layer to excite the compressed features, transform the feature dimension, and assign weights to obtain the feature weight size.
[0214] In another specific implementation of the embodiments of the present specification, the RTDC auxiliary branch is composed of 3 layers of DSC cascaded 4 layers of time dilation convolution and CBAM; on the RTDC auxiliary branch side, when the processing module 903 uses a dual-branch encoder to extract all-round feature information of the noise signal, it is specifically used to input the feature information of the noise signal, and respectively perform channel expansion, dimension reorganization, and compression of the feature map channel and spatial dimension through DSC with kernel sizes of 3×3, 1×2, and 3×3, and perform residual operation on the feature dimension based on 2×4 DSC and the compressed output; perform N convolutions on the time series features obtained by the residual operation: use a kernel size of 1×3 and a dilation rate of 2 N-1 The time dilation convolution TCN is padded and downsampled, where N is a positive integer; the CBAM mechanism is used to assign feature weights to the extracted time series feature information.
[0215] In another specific implementation of the embodiment of this specification, the formula of CBAM is:
[0216]
[0217] Among them, the calculation formula of the channel attention module is:
[0218]
[0219] The calculation formula of spatial attention is:
[0220]
[0221] Among them, W0 and W1 are the weight values calculated in MLP, σ is the Sigmoid function, and the output is multiplied by the input feature map F to obtain the channel-optimized feature F' as the input feature map of the spatial attention module; f 7×7 Represents a convolution operation with a convolution kernel size of 7×7.
[0222] In another specific implementation of the embodiment of this specification, the multi-head self-attention mechanism MHSA formula is as follows:
[0223] Q=FW q
[0224] K=FW k
[0225] V=FW ν
[0226]
[0227] The multi-head attention mechanism includes two parallel attention heads; each attention head has an independent projection weight matrix;
[0228] After extracting features based on the main road side of the coding block and the auxiliary branch side of the RTDC, the processing module 903 is also used to perform the following feature fusion operations: for each attention head, the query, key, and numerical vector are multiplied with the corresponding weight matrix to obtain the query, key, and numerical vector corresponding to each attention head; attention calculation is performed on each attention head to obtain the attention weights and weighted numerical vectors of multiple attention heads; the weighted numerical vectors of multiple attention heads are connected or spliced to obtain the global feature information output by the multi-head attention.
[0229] It should be understood that the channel equalization model training device of the embodiment of this specification can also execute Figure 1 The method shown in FIG. 1 and the channel equalization model training device are implemented in Figure 1 The functions of the illustrated embodiment will not be described in detail here.
[0230] Fig.10 A schematic diagram of the structure of a device for equalizing fading signals provided in an embodiment of the present application. The device 1000 includes: an acquisition module 1001 and a prediction module 1002; the acquisition module 1001 is used to acquire test data to be processed, wherein the test data is a modulated signal carrying Rayleigh fading; the prediction module 1002 is used to input the test data into a channel equalization processing model for processing to obtain a predicted signal after equalization; wherein the signal equalization processing model is trained based on the channel equalization model training method.
[0231] Embodiment 3
[0232] An embodiment of the present application provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is loaded and executed by a processor to implement the channel equalization model training method described above, and / or the equalization processing method of the fading signal described above.
[0233] An embodiment of the present application provides an electronic device, which includes a processor and a memory, wherein the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the channel equalization model training method as described above, and / or the equalization processing method of the fading signal.
[0234] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the relevant laws and regulations and do not violate public order and good morals.
[0235] Fig.11A schematic block diagram of an example electronic device 1100 that can be used to implement an embodiment of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.
[0236] like Fig.11 As shown, the electronic device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. In the RAM 1103, various programs and data required for the operation of the electronic device 1100 can also be stored. The computing unit 1101, the ROM 1102, and the RAM 1103 are connected to each other via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.
[0237] Multiple components in the electronic device 1100 are connected to the I / O interface 1105, including: an input unit 1106, such as a keyboard, a mouse, etc.; an output unit 1107, such as various types of displays, speakers, etc.; a storage unit 1108, such as a disk, an optical disk, etc.; and a communication unit 1109, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1109 allows the electronic device 1100 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0238] The computing unit 1101 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1101 performs the various methods and processes described above, such as the channel equalization model training method, or the equalization processing method of the fading signal. For example, in some embodiments, the channel equalization model training method or the equalization processing method of the fading signal may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 1108. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 1100 via the ROM 1102 and / or the communication unit 1109. When the computer program is loaded into the RAM 1103 and executed by the computing unit 1101, the channel equalization model training method described above or one or more steps of the equalization processing method of the fading signal may be executed. Alternatively, in other embodiments, the computing unit 1101 may be configured to execute the channel equalization model training method or the equalization processing method of the fading signal in any other appropriate manner (for example, by means of firmware).
[0239] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0240] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
[0241] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0242] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0243] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.
[0244] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.
[0245] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in this application can be achieved, and this document is not limited here.
[0246] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.
Claims
1. A channel equalization model training method, characterized in that: include: Determine an original signal, and simulate the original signal to generate a sample data set; wherein the sample data in the sample data set includes: a noise-free signal used as label data and a noise signal used as a training sample, wherein the noise signal is a modulated signal carrying the influence of Rayleigh fading; An initial model based on a hybrid attention mechanism is selected; the initial model includes: a dual-branch encoder and a single-branch decoder, wherein the encoding block trunk in the dual-branch encoder is used to extract detail spatial features, and the time residual dilated convolution RTDC auxiliary branch is used to construct temporal feature dependencies; Input the training sample data in the sample data set into the initial model, extract all-round feature information of the noise signal using a dual-branch encoder, and perform dimension upgrade on the feature information using a decoder to output a reconstructed signal; Based on the SmoothL1Loss loss value of the reconstructed signal and the noise-free signal in the training sample data, the initial model is repeatedly iterated, and the learning rate is adjusted to determine the model parameters to obtain a trained channel equalization model.
2. The method according to claim 1, characterized in that Determining an original signal and simulating the original signal to generate a sample data set specifically includes: Generate a binary bit stream randomly and simulate the original signal to be transmitted; Based on the original signal, multiple different noise-free signals are generated by simulating modulation according to multiple preset modulation modes, and the noise-free signals are preprocessed, wherein the preprocessing operation at least includes: waveform adjustment and power normalization processing; Simulating a Rayleigh fading channel, simulating a noise-free signal after a preprocessing operation, a fading signal transmitted through the Rayleigh fading channel, and simulating a Gaussian white noise channel, adding noise power to the fading signal to obtain a noise signal; A variety of different noise-free signals and their corresponding noise signals are summarized as a sample data set.
3. The method according to claim 1, characterized in that The initial model is an autoencoder; the autoencoder based on the hybrid attention mechanism includes: a dual-branch encoder, a multi-head self-attention mechanism MHSA fusion feature module, and a single-branch decoder; wherein the dual-branch encoder includes: a coding block trunk and an RTDC auxiliary branch, the coding block trunk includes 4 coding blocks; the single-branch decoder includes 3 stacked decoding blocks; The dual-branch encoder is composed of a first convolutional layer and four encoding blocks connected in sequence, and a residual connection of a CBAM attention mechanism is set between the first convolutional layer and the third encoding block, and a residual connection is set in the first layer and the third layer inside the encoding block and after the last scaling operation; the single-branch decoder is composed of three decoding blocks and a transposed convolutional layer connected in series, a residual connection is set between the first decoding block and the second decoding block, and a residual connection is set in the first layer and the third layer inside the decoding block and after the last scaling operation.
4. The method according to claim 3, characterized in that The encoding block contains: a deep separable convolution DSC, and a 3-layer DSC cascade SE attention mechanism, wherein each layer of DSC is followed by a batch normalization layer and an activation function layer, and the compression-excitation SE module is composed of a pooling layer, a first convolution layer, a LeakyReLU activation function, a second convolution layer, and a Sigmod activation function; On the trunk side of the coding block, when a dual-branch encoder is used to extract all-round feature information of the noise signal, the following steps are specifically performed: The first layer DSC is used to expand the channel dimension of the feature information of the noise signal input to the coding block to obtain the corresponding feature map; The second layer DSC is used to downsample the feature map by controlling the step size and padding; The third layer DSC is used to further adjust the number of channels of the input feature map; The input feature map is compressed using the pooling layer and the first convolutional layer to compress the feature dimension; The activation function and the second convolutional layer are used to stimulate the compressed features, transform the feature dimensions, and assign weights to obtain the feature weight size.
5. The method according to claim 3, characterized in that The RTDC auxiliary branch is composed of 3 layers of DSC cascaded with 4 layers of time dilation convolution and CBAM; On the auxiliary branch side of the RTDC, the dual-branch encoder is used to extract all-round feature information of the noise signal, specifically including: The feature information of the input noise signal is respectively expanded through channels, reorganized in dimensions, and compressed in feature map channels and spatial dimensions through DSC with kernel sizes of 3×3, 1×2, and 3×3. The feature dimensions are matched once based on 2×4 DSC and the residual operation is performed with the compressed output. The time series features obtained by residual operation are convolved N times in sequence: using a kernel size of 1×3 and a dilation rate of 2 N-1 The time dilation convolution TCN performs padding convolution and downsampling, where N is a positive integer; The CBAM mechanism is used to assign feature weights to the extracted time series feature information.
6. The method according to claim 5, characterized in that The formula of the CBAM is: Among them, the calculation formula of the channel attention module is: The calculation formula of spatial attention is: Among them, W0 and W1 are the weight values calculated in MLP, ~ is the Sigmoid function, the output is multiplied by the input feature map F to obtain the channel-optimized feature F′, which is used as the input feature map of the spatial attention module, and the output F is the feature map F″ with spatial channel attention after CBAM processing; 7×7 Represents a convolution operation with a convolution kernel size of 7×7.
7. The method according to claim 4 or 5, characterized in that The multi-head self-attention mechanism MHSA formula is as follows: Q=FW q K=FW k V=FW v Among them, F represents the feature map input to MHSA, Q, K, and V represent the weight matrices to obtain the query matrix, key matrix, and value matrix respectively; W Q , W K , W V is the weight of the corresponding matrix; Attn is the operation process of the multi-head attention mechanism, D K is the length of F; The multi-head attention mechanism includes two parallel attention heads; each attention head has an independent projection weight matrix; After extracting features based on the coding block trunk road side and the RTDC auxiliary branch side, the following feature fusion operations are also included: For each attention head, multiply the query, key, and numeric vector with the corresponding weight matrix to obtain the query, key, and numeric vector corresponding to each attention head; Perform attention calculation on each attention head to obtain the attention weights and weighted numerical vectors of multiple attention heads; The weighted numerical vectors of multiple attention heads are connected or concatenated to obtain the global feature information of the multi-head attention output.
8. A method for equalizing a fading signal, characterized in that: include: Acquire test data to be processed, wherein the test data is a modulated signal carrying Rayleigh fading; Inputting the test data into a channel equalization processing model for processing to obtain an equalized prediction signal; Wherein, the signal equalization processing model is trained based on the channel equalization model training method described in any one of claims 1-7.
9. A channel equalization model training device, characterized in that: include: A simulation module, used to determine an original signal, and simulate the original signal to generate a sample data set; wherein the sample data in the sample data set includes: a noise-free signal used as label data and a noise signal used as a training sample, wherein the noise signal is a modulated signal carrying the influence of Rayleigh fading; A selection module is used to select an initial model based on a hybrid attention mechanism; the initial model includes: a dual-branch encoder and a single-branch decoder, wherein the encoding block trunk in the dual-branch encoder is used to extract detail spatial features, and the time residual dilation convolution RTDC auxiliary branch is used to construct temporal feature dependencies; A processing module, used for inputting the training sample data in the sample data set into the initial model, extracting all-round feature information of the noise signal by using a dual-branch encoder, and upgrading the feature information by using a decoder, and outputting a reconstructed signal; The training module is used to repeatedly iterate the initial model and adjust the learning rate to determine the model parameters based on the SmoothL1Loss loss value of the reconstructed signal and the noise-free signal in the training sample data, so as to obtain a trained channel equalization model.
10. An equalization processing device for fading signals, characterized in that: include: An acquisition module, used for acquiring test data to be processed, wherein the test data is a modulated signal carrying Rayleigh fading; A prediction module, used for inputting the test data into a channel equalization processing model for processing to obtain a predicted signal after equalization; Wherein, the signal equalization processing model is trained based on the channel equalization model training method described in any one of claims 1-7.
11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, which, when executed by an electronic device including multiple application programs, enable the electronic device to perform the steps of the channel equalization model training method described in any one of claims 1-7, and / or perform the steps of equalization processing of fading signals described in claim 8.
12. A computer program product, characterized in that The computer-readable storage medium stores one or more programs, which, when executed by an electronic device including multiple application programs, enable the electronic device to perform the steps of the channel equalization model training method described in any one of claims 1-7, and / or perform the steps of the equalization processing method for fading signals described in claim 8.
Citation Information
Patent Citations
Training method of channel estimation model, channel estimation method, device and equipment
CN114781455A
Optical fiber communication channel equalization method based on unsupervised learning neural network
CN117040978A
Multi-feature fusion echo cancellation method and system based on self-attention transform network
WO2023044961A1
Model training and scene recognition method and apparatus, device, and medium
WO2023056889A1
Cited By
High-speed train prediction cascade control method and device based on PPC and MPC
CN121448475A
A Predictive Cascade Control Method and Device for High-Speed Trains Based on PPC and MPC
CN121448475B
