Channel equalizer construction method and apparatus, computer device, readable storage medium and program product

Abstract

The present application relates to a channel equalizer construction method and apparatus, a computer device, a computer readable storage medium and a computer program product. The method comprises: acquiring a symbol sequence sample; using a sliding window to perform sliding window processing on a plurality of symbols according to a symbol receiving sequence, and for a plurality of current target symbols in the sliding window, inputting the plurality of target symbols into a feed-forward equalizer to obtain a feed-forward output; for each feed-forward output after an Lth feed-forward output, acquiring L feed-forward outputs before a current feed-forward output, and inputting the L feed-forward outputs into a feedback equalizer to obtain a feedback output; determining a prediction symbol on the basis of the current feed-forward output and the feedback output; and determining a prediction error on the basis of the prediction symbol and an expected response, and adjusting parameters in the feed-forward equalizer and the feedback equalizer on the basis of the prediction error until the prediction error obtained by calculation is smaller than a preset value.

Description

Channel equalizer construction method and device, computer equipment, readable storage medium and program product

[0001] Cross-reference to Related Applications

[0002] The present application claims priority to the Chinese patent application No. 2024112653676, filed on September 10, 2024, and entitled "Channel equalizer construction method and device, computer equipment, readable storage medium and program product", the contents of which are hereby incorporated by reference in their entirety. TECHNICAL FIELD

[0003] The present application relates to the field of communication technology, in particular to a channel equalizer construction method and device, computer equipment, computer readable storage medium and computer program product. BACKGROUND

[0004] With the continuous rapid development of society, digital communication technology is increasingly important in the information age. People's desire for digital communication technology and communication quality is also increasing. Channel equalization technology, as a technology that can effectively improve the transmission performance of a communication system and reduce the interference caused by various non-ideal factors on the communication system, plays an increasingly important role in digital communication systems. Today, channel equalization technology is widely used in various digital communication systems. SUMMARY

[0005] The present application provides a channel equalizer construction method and device, computer equipment, computer readable storage medium and computer program product capable of reducing the inter-symbol interference of equalized symbols.

[0006] In a first aspect, the present application provides a channel equalizer construction method, comprising:

[0007] Obtaining a symbol sequence sample, the symbol sequence sample comprising a plurality of symbols arranged in a symbol receiving order;

[0008] Using a sliding window to perform sliding window processing on the plurality of symbols in the symbol receiving order, for a plurality of target symbols in the sliding window, inputting the plurality of target symbols into a feedforward equalizer to obtain a feedforward output;

[0009] For each feedforward output after the Lth feedforward output, obtaining L feedforward outputs before the current feedforward output, inputting the L feedforward outputs into a feedback equalizer to obtain a feedback output;

[0010] Determining a predicted symbol based on the current feedforward output and the feedback output;

[0011] Based on the predicted symbol and the expected response, a prediction error is determined, and parameters in the feedforward equalizer and the feedback equalizer are adjusted based on the prediction error until the calculated prediction error is less than a preset value.

[0012] In one of the embodiments, the size of the sliding window is 2K1, and the number of symbols in the symbol sequence sample is N; the sliding window processing of the plurality of symbols in the symbol receiving order using the sliding window comprises:

[0013] The first to the 2K1 symbols are extracted from the symbol sequence sample as the plurality of target symbols in the sliding window at present;

[0014] After obtaining the corresponding feedforward output, the second to the 2K1+1 symbols are extracted from the symbol sequence sample as the plurality of target symbols in the sliding window at present;

[0015] After obtaining the corresponding feedforward output, the sliding window is moved by one symbol as a step and the extraction operation is performed until the last target symbol in the sliding window at present is the Nth symbol in the symbol sequence sample.

[0016] In one of the embodiments, the feedforward equalizer comprises an input layer, a hidden layer and an output layer; the number of neurons in the input layer is the same as the number of the plurality of target symbols, and the number of neurons in the hidden layer is the same as the order of the channel to be fitted by the equalizer; the inputting of the plurality of target symbols into the feedforward equalizer to obtain the feedforward output comprises:

[0017] The plurality of target symbols are respectively input into a plurality of first neurons in the input layer;

[0018] Each first neuron in the ith set inputs the received target symbol into an ith-order neuron in the hidden layer; the ith set is a set composed of a preset number of target symbols before the center symbol and a preset number of target symbols after the center symbol in the plurality of target symbols, and each corresponding first neuron;

[0019] Each order neuron in the hidden layer performs corresponding operation on the received target symbol to obtain an operation result and output the operation result to the output layer;

[0020] The output layer adds the received operation results to obtain the feedforward output.

[0021] In one of the embodiments, the corresponding operation of each order neuron in the hidden layer on the received target symbol comprises:

[0022] For the ith-order neuron, an ith tap coefficient corresponding to each received target symbol is obtained.

[0023] Multiplying the i-th tap coefficient corresponding to each target symbol and the corresponding target symbol, summing each multiplication result, adding the sum result and the bias corresponding to the i-th neuron, taking the i-th power of the addition result, and obtaining the operation result.

[0024] In one of the embodiments, the inputting the L-th feedforward output to the feedback equalizer to obtain a feedback output comprises:

[0025] Obtaining a decision coefficient corresponding to the i-th feedforward output before the current feedforward output;

[0026] Multiplying the decision coefficient corresponding to the i-th feedforward output and the i-th feedforward output to obtain a decision result corresponding to the i-th feedforward output;

[0027] Adding the decision results corresponding to the L-th feedforward output respectively to obtain the feedback output.

[0028] In one of the embodiments, the parameters of the feedforward equalizer comprise tap coefficients, and the parameters of the feedback equalizer comprise decision coefficients; and the adjusting the parameters in the feedforward equalizer and the feedback equalizer based on the prediction error comprises:

[0029] Obtaining a mean square error gradient of the prediction error;

[0030] For the tap coefficients or the decision coefficients, obtaining a current value of a current coefficient, multiplying the current value by a learning rate to obtain a multiplication result, subtracting the multiplication result from the current value to obtain a subtraction result, and updating the current value using the subtraction result.

[0031] In one of the embodiments, the plurality of symbols in the symbol sequence sample are a plurality of symbols received by the receiving end in an actual working process of receiving symbols sent by the sending end through a channel.

[0032] In one of the embodiments, the determining a prediction symbol based on the current feedforward output and the feedback output comprises:

[0033] Adding the current feedforward output and the feedback output to obtain the prediction symbol.

[0034] In one of the embodiments, the determining a prediction symbol based on the current feedforward output and the feedback output comprises:

[0035] Pre-setting a feedforward weight and a feedback weight, multiplying the current feedforward output by the feedforward weight, multiplying the feedback output by the feedback weight, and adding the two multiplication results to obtain the prediction symbol.

[0036] In one of the embodiments, based on the predicted symbol and the expected response, a prediction error is determined, including:

[0037] The prediction error is obtained by subtracting the predicted symbol from the expected response.

[0038] In one of the embodiments, the constructed channel equalizer incorporates Volterra series.

[0039] In one of the embodiments, the method further includes: using the constructed channel equalizer to recover the symbols transmitted through the channel to obtain recovered symbols.

[0040] In a second aspect, the present application further provides a channel equalizer construction device, including:

[0041] An acquisition module is configured to acquire a symbol sequence sample, the symbol sequence sample including a plurality of symbols arranged in a symbol receiving order;

[0042] A sliding window processing module is configured to use a sliding window to perform sliding window processing on the plurality of symbols in the symbol receiving order, and for a plurality of target symbols currently in the sliding window, input the plurality of target symbols into a feedforward equalizer to obtain a feedforward output.

[0043] The acquisition module is further configured to, for each feedforward output after the Lth feedforward output, acquire L feedforward outputs before the current feedforward output, input the L feedforward outputs into a feedback equalizer to obtain a feedback output, and determine a predicted symbol based on the current feedforward output and the feedback output.

[0044] A parameter adjustment module is configured to determine a prediction error based on the predicted symbol and an expected response, and adjust parameters in the feedforward equalizer and the feedback equalizer based on the prediction error until the calculated prediction error is less than a preset value.

[0045] In a third aspect, the present application further provides a computer device, including a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0046] Acquire a symbol sequence sample, the symbol sequence sample including a plurality of symbols arranged in a symbol receiving order;

[0047] Use a sliding window to perform sliding window processing on the plurality of symbols in the symbol receiving order, and for a plurality of target symbols currently in the sliding window, input the plurality of target symbols into a feedforward equalizer to obtain a feedforward output.

[0048] For each feedforward output after the Lth feedforward output, L feedforward outputs before the current feedforward output are obtained, the L feedforward outputs are input to a feedback equalizer to obtain a feedback output;

[0049] Based on the current feedforward output and the feedback output, a predicted symbol is determined;

[0050] Based on the predicted symbol and an expected response, a prediction error is determined, and parameters in the feedforward equalizer and the feedback equalizer are adjusted based on the prediction error until the calculated prediction error is less than a preset value.

[0051] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the following steps:

[0052] Obtaining a symbol sequence sample, the symbol sequence sample comprising a plurality of symbols arranged in a symbol receiving order;

[0053] Using a sliding window to perform sliding window processing on the plurality of symbols in the symbol receiving order, for a plurality of target symbols in the sliding window, inputting the plurality of target symbols to a feedforward equalizer to obtain a feedforward output;

[0054] For each feedforward output after the Lth feedforward output, L feedforward outputs before the current feedforward output are obtained, the L feedforward outputs are input to a feedback equalizer to obtain a feedback output;

[0055] Based on the current feedforward output and the feedback output, a predicted symbol is determined;

[0056] Based on the predicted symbol and an expected response, a prediction error is determined, and parameters in the feedforward equalizer and the feedback equalizer are adjusted based on the prediction error until the calculated prediction error is less than a preset value.

[0057] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, and the computer program is executed by a processor to implement the following steps:

[0058] Obtaining a symbol sequence sample, the symbol sequence sample comprising a plurality of symbols arranged in a symbol receiving order;

[0059] Using a sliding window to perform sliding window processing on the plurality of symbols in the symbol receiving order, for a plurality of target symbols in the sliding window, inputting the plurality of target symbols to a feedforward equalizer to obtain a feedforward output;

[0060] For each feedforward output after the Lth feedforward output, L feedforward outputs before the current feedforward output are obtained, the L feedforward outputs are input to a feedback equalizer to obtain a feedback output;

[0061] Based on the current feedforward output and the feedback output, a prediction symbol is determined;

[0062] Based on the prediction symbol and an expected response, a prediction error is determined, and parameters in the feedforward equalizer and the feedback equalizer are adjusted based on the prediction error until the calculated prediction error is less than a preset value.

[0063] The details of one or more embodiments of the present application are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of the present application will be apparent from the description and drawings, and from the claims. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other related drawings can also be obtained without creative labor.

[0065] Fig. 1 is a diagram of an application environment of a channel equalizer construction method in an embodiment;

[0066] Fig. 2 is a flowchart of a channel equalizer construction method in an embodiment;

[0067] Fig. 3 is a schematic diagram of a channel equalizer construction principle in an embodiment;

[0068] Fig. 4 is a schematic diagram of an actual application of a channel equalizer in an embodiment;

[0069] Fig. 5 is a schematic diagram of a bit error rate comparison of various channel equalizers in an embodiment;

[0070] Fig. 6 is a schematic diagram of a net rate comparison of various channel equalizers in an embodiment;

[0071] Fig. 7 is a structural block diagram of a channel equalizer construction device in an embodiment;

[0072] Fig. 8 is an internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0073] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0074] In the related art, a feedforward equalizer is constructed, and the feedforward equalizer is used to reduce interference. However, the feedforward equalizer cannot eliminate the intersymbol interference of the equalized symbol.

[0075] The channel equalizer construction method provided in the embodiments of the present application can be applied in the application environment shown in FIG. 1. The application environment shown in FIG. 1 includes a sending end and a receiving end, for example, devices that communicate with each other to send and receive symbols in a digital communication system. After the receiving end receives the symbol sent by the sending end through a channel, the symbol can be input to a pre-constructed channel equalizer. The channel equalizer restores the symbol to obtain a restored signal. The channel equalizer is constructed in the following manner: obtaining a symbol sequence sample, the symbol sequence sample including a plurality of symbols arranged in sequence according to a symbol receiving order; using a sliding window to perform sliding window processing on the plurality of symbols according to the symbol receiving order; for a plurality of target symbols in the sliding window, inputting the plurality of target symbols to a feedforward equalizer to obtain a feedforward output; for each feedforward output after the Lth feedforward output, obtaining L feedforward outputs before the current feedforward output, and inputting the L feedforward outputs to a feedback equalizer to obtain a feedback output; determining a predicted symbol based on the current feedforward output and the feedback output; determining a prediction error based on the predicted symbol and an expected response, adjusting parameters in the feedforward equalizer and the feedback equalizer based on the prediction error, until the calculated prediction error is less than a preset value, and regarding the combination of the feedforward equalizer and the feedback equalizer at this time as a trained channel equalizer. Since the feedback equalizer is added in the construction process of the channel equalizer, the restored signal can eliminate the intersymbol interference of the equalized symbol, is closer to the real symbol sent by the sending end, and is more accurate.

[0076] The sending end can be a base station, and the receiving end can be a terminal. This is only one possible scenario, and the sending end and the receiving end can also be corresponding devices in other scenarios, which are not limited in the embodiments of the present application.

[0077] The terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things device can be a smart speaker, a smart television, a smart air conditioner, a smart vehicle device, a projection device, and the like. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, and the like. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, and the like.

[0078] In one exemplary embodiment, as shown in FIG. 2, a channel equalizer construction method is provided, which is described by taking the receiving end in FIG. 1 as an example, and includes the following steps 202 to 210. Among them:

[0079] Step 202, obtaining a symbol sequence sample, wherein the symbol sequence sample includes a plurality of symbols arranged in sequence according to a symbol receiving order.

[0080] Optionally, the symbol sequence sample includes N symbols arranged in sequence according to the symbol receiving order.

[0081] Optionally, the plurality of symbols in the symbol sequence sample can be a plurality of symbols received by the receiving end in actual operation.

[0082] Step 204, using a sliding window to perform sliding window processing on the plurality of symbols according to the symbol receiving order, and for a plurality of target symbols in the sliding window, inputting the plurality of target symbols into a feedforward equalizer to obtain a feedforward output.

[0083] Optionally, the size of the sliding window can be set according to actual conditions, and exemplarily, the size of the sliding window can be 2K1, that is, the sliding window can cover 2K1 symbols.

[0084] Optionally, the sliding window processing means that according to the symbol receiving order, a first symbol group can be covered by using the sliding window, a plurality of symbols in the symbol group are used as a plurality of target symbols in the sliding window, and subsequent processing is performed; then the sliding window is moved by one symbol as a step, that is, a second symbol group is covered by using the sliding window, a plurality of symbols in the symbol group are used as a plurality of target symbols in the sliding window, and subsequent processing is performed; and so on, until the last symbol in the symbol group covered by the sliding window is the last symbol in the symbol sequence sample. The number of symbols in the symbol group is the same as the size of the sliding window. Preferably, the size of the sliding window is fixed.

[0085] Optionally, the feedforward equalizer includes an input layer, a hidden layer and an output layer, and for a plurality of target symbols in the sliding window, the plurality of target symbols are sequentially processed based on the plurality of target symbols through the input layer, the hidden layer and the output layer to obtain the feedforward output. The sliding window can correspondingly calculate a feedforward output at each position.

[0086] Step 206, for each feedforward output after the Lth feedforward output, obtaining L feedforward outputs before the current feedforward output, and inputting the L feedforward outputs into a feedback equalizer to obtain a feedback output.

[0087] Wherein, L is a value set in advance according to actual conditions,

[0088] Optionally, the sliding window can calculate a corresponding feedforward output at each position, and after L feedforward outputs are calculated, for each subsequent feedforward output, L previous feedforward outputs before the current feedforward output can be obtained, and the L feedforward outputs are input to the feedback equalizer to obtain a feedback output.

[0089] At step 208, a predicted symbol is determined based on the current feedforward output and the feedback output.

[0090] Optionally, the predicted symbol can be obtained by adding the current feedforward output and the feedback output. Alternatively, a feedforward weight and a feedback weight can be preset, the current feedforward output is multiplied by the feedforward weight, the feedback output is multiplied by the feedback weight, and the two multiplied results are added to obtain the predicted symbol.

[0091] At step 210, a prediction error is determined based on the predicted symbol and an expected response, and parameters in the feedforward equalizer and the feedback equalizer are adjusted based on the prediction error until the calculated prediction error is less than a preset value.

[0092] The expected response is an undistorted symbol of a middle symbol of the plurality of target symbols corresponding to the current feedforward output.

[0093] The prediction error can be obtained by subtracting the predicted symbol from the expected response. The parameters in the feedforward equalizer and the feedback equalizer are adjusted based on the prediction error until the calculated prediction error is less than a preset value. The combination of the feedforward equalizer and the feedback equalizer at this time is used as a trained channel equalizer.

[0094] As described above, the sliding window can calculate a corresponding feedforward output at each position, and after L feedforward outputs are calculated, for each subsequent feedforward output, L previous feedforward outputs before the current feedforward output can be obtained, the L feedforward outputs are input to the feedback equalizer to obtain a feedback output corresponding to the current feedforward output, and then the predicted symbol, the prediction error, and the parameter adjustment are calculated. That is, after L feedforward outputs are calculated, for each subsequent feedforward output, a parameter adjustment action can be performed once.

[0095] In the above embodiment, a symbol sequence sample is obtained, the symbol sequence sample including a plurality of symbols arranged in a symbol receiving sequence; a sliding window is used to perform sliding window processing on the plurality of symbols in the symbol receiving sequence; for a plurality of target symbols in the sliding window, the plurality of target symbols are input into a feedforward equalizer to obtain a feedforward output; for each feedforward output after the Lth feedforward output, L feedforward outputs before the current feedforward output are obtained, and the L feedforward outputs are input into a feedback equalizer to obtain a feedback output; a predicted symbol is determined based on the current feedforward output and the feedback output; a prediction error is determined based on the predicted symbol and an expected response; and parameters in the feedforward equalizer and the feedback equalizer are adjusted based on the prediction error until the calculated prediction error is less than a preset value. The construction of the feedback equalizer makes the final channel equalizer capable of eliminating intersymbol interference of the equalized symbols and restoring the received symbols to symbols closer to the true symbols sent by the sending end, thus improving the accuracy.

[0096] In some embodiments, the size of the sliding window is 2K1, and the number of symbols in the symbol sequence sample is N; the step of using the sliding window to perform sliding window processing on the plurality of symbols in the symbol receiving sequence includes: extracting the 1st to 2K1th symbols from the symbol sequence sample as the plurality of target symbols in the sliding window; after obtaining the corresponding feedforward output, extracting the 2nd to 2K1+1th symbols from the symbol sequence sample as the plurality of target symbols in the sliding window; and after obtaining the corresponding feedforward output, continuing to move the sliding window by one symbol and perform the extraction operation until the last target symbol in the sliding window is the Nth symbol in the symbol sequence sample.

[0097] The symbol sequence sample includes N symbols arranged in a symbol receiving sequence.

[0098] Optionally, the use process of the sliding window is as follows: the 1st to 2K1th symbols are extracted from the symbol sequence sample as the plurality of target symbols in the sliding window, and the plurality of target symbols are input into the feedforward equalizer to obtain a feedforward output; then, the 2nd to 2K1+1th symbols are extracted from the symbol sequence sample as the plurality of target symbols in the sliding window, and the plurality of target symbols are input into the feedforward equalizer to obtain a feedforward output; and the process is repeated until the last target symbol in the sliding window is the Nth symbol in the symbol sequence sample, the plurality of target symbols in the sliding window are input into the feedforward equalizer to obtain a feedforward output, and the sliding processing is completed.

[0099] The above embodiments describe the sliding processing process, which allows the channel equalizer to complete multiple training iterations, thus improving training efficiency.

[0100] In some embodiments, the feedforward equalizer includes an input layer, a hidden layer, and an output layer; the number of neurons in the input layer is the same as the number of target symbols, and the number of neurons in the hidden layer is the same as the order of the channel that the equalizer needs to fit; the step of inputting the multiple target symbols into the feedforward equalizer to obtain a feedforward output includes: inputting the multiple target symbols into multiple first neurons in the input layer respectively; each first neuron in the i-th set inputs the received target symbols into the i-th order neuron in the hidden layer; the i-th set is a set of first neurons corresponding to a preset number of target symbols before the center symbol and a preset number of target symbols after the center symbol; each order neuron in the hidden layer performs a corresponding operation on the received target symbols, obtains the operation result, and outputs it to the output layer; the output layer adds the received operation results to obtain the feedforward output.

[0101] For example, referring to Figure 3, the channel that the weighing instrument needs to fit is of order 3. The hidden layer includes three neurons: an order 1 neuron, an order 2 neuron, and an order 3 neuron. The multiple target symbols can be input into multiple first neurons in the input layer, with the intermediate symbol of the multiple target symbols being v. k The middle symbol v k The previous K1 symbols were v in sequence. k-1 v k-2 v k-3 ...v k-K1 The middle symbol v k The following K1 symbols are v in sequence. k+1 v k+2 v k+3 ...v k+K1 The first set can be the intermediate symbol v k The first set consists of the first neurons corresponding to the previous K1 symbols and the next K1 symbols. Each first neuron in the first set inputs the received target symbol to its first-order neuron. The second set can be the intermediate symbol v. k The first set consists of the first neurons corresponding to the previous m2 symbols and the next m2 symbols. Each first neuron in the second set inputs the received target symbol to a second-order neuron. The third set can be the intermediate symbol v. kThe first neurons in the third set each correspond to the previous m3 symbols and the following m3 symbols, and each of the first neurons in the third set inputs the received target symbols to the third order neuron. The first order neuron performs corresponding operations on the received target symbols, obtains an operation result, and outputs the operation result to the output layer. Similarly, the second order neuron performs corresponding operations on the received target symbols, obtains an operation result, and outputs the operation result to the output layer. The third order neuron performs corresponding operations on the received target symbols, obtains an operation result, and outputs the operation result to the output layer. The output layer adds the received operation results to obtain the feedforward output y(k).

[0102] In the above embodiment, the number of neurons in the hidden layer is determined by the order of the channel to be fitted by the equalizer, and this configuration simplifies the selection process of the structural hyperparameters in the neural network.

[0103] In some embodiments, the step of performing corresponding operations on the received target symbols by each order of neurons in the hidden layer includes: for the i-th order neuron, obtaining the i-th tap coefficient corresponding to each received target symbol; multiplying the i-th order tap coefficient corresponding to each target symbol and the corresponding target symbol, summing the multiplication results, adding the sum to the bias corresponding to the i-th order neuron, and taking the i-th power of the addition result to obtain the operation result.

[0104] For example, in combination with the feedforward equalizer shown in FIG. 3, the order of the channel to be fitted by the equalizer is 3, and the hidden layer includes three neurons, which are the first order neuron, the second order neuron, and the third order neuron. Each first neuron in the first set inputs the received target symbols to the first order neuron. Each first neuron in the second set inputs the received target symbols to the second order neuron. Each first neuron in the third set inputs the received target symbols to the third order neuron. The first order neuron performs corresponding operations on the received target symbols, obtains an operation result, and outputs the operation result to the output layer. Similarly, the second order neuron performs corresponding operations on the received target symbols, obtains an operation result, and outputs the operation result to the output layer. The third order neuron performs corresponding operations on the received target symbols, obtains an operation result, and outputs the operation result to the output layer. The output layer adds the received operation results to obtain the feedforward output y(k).

[0105] The operation result of the first order neuron can be expressed as follows:

[0106] wherein v (k-i) is the k-i-th target symbol in the plurality of target symbols received by the first order neuron, which is based on the middle symbol v k is the k-i-th target symbol in the plurality of target symbols received by the first order neuron, which is based on the middle symbol v ff1 (i) is v(k-i) The corresponding first-order tap coefficient is b1, and m1 is less than K1.

[0107] The operation result of the second-order neuron can be expressed by the following formula:

[0108] where v (k-i) is the k-i-th target symbol among the multiple target symbols received by the second-order neuron, w k is the reference, and w ff2 (i) is v (k-i) The corresponding second-order tap coefficient is b2, and m2 is less than K1.

[0109] The operation result of the third-order neuron can be expressed by the following formula:

[0110] where v (k-i) is the k-i-th target symbol among the multiple target symbols received by the third-order neuron, w k is the reference, and w ff3 (i) is v (k-i) The corresponding third-order tap coefficient is b3, and m3 is less than m2.

[0111] In the above embodiments, the operation process of the i-th neuron is introduced, and a cross beat frequency term is constructed using a self-defined activation function, thereby avoiding redundant high-order nonlinear terms.

[0112] In some embodiments, the step of inputting the L-th feedforward output into a feedback equalizer to obtain a feedback output comprises: obtaining a decision coefficient corresponding to an i-th feedforward output before the current feedforward output; multiplying the decision coefficient corresponding to the i-th feedforward output and the i-th feedforward output to obtain a decision result corresponding to the i-th feedforward output; and adding the decision results corresponding to the L-th feedforward outputs respectively to obtain the feedback output.

[0113] The feedback output can be expressed by the following formula:

[0114] where r(k-i) is the i-th feedforward output before the current feedforward output, w fb (i) is the decision coefficient corresponding to r(k-i).

[0115] In the above embodiments, the feedback output is obtained, for each feedforward output after the Lth feedforward output, a decision coefficient corresponding to the ith feedforward output before the current feedforward output is obtained; the decision coefficient corresponding to the ith feedforward output is multiplied by the ith feedforward output to obtain a decision result corresponding to the ith feedforward output; and the decision results corresponding to the L feedforward outputs are added to obtain the feedback output. The feedback equalizer is introduced in the process of constructing the channel equalizer, so that the final obtained channel equalizer can eliminate the intersymbol interference of the equalized symbols, restore the received symbols to symbols closer to the real symbols sent by the sending end, and is more accurate.

[0116] In some embodiments, the parameters of the feedforward equalizer include tap coefficients, and the parameters of the feedback equalizer include decision coefficients; and the adjusting the parameters in the feedforward equalizer and the feedback equalizer based on the prediction error includes: obtaining a mean square error gradient of the prediction error; for the tap coefficients or the decision coefficients, obtaining a current value of a current coefficient, multiplying the mean square error gradient by a learning rate to obtain a multiplication result, and updating the current value of the current coefficient using a subtraction result of the current value and the multiplication result.

[0117] Optionally, the predicted symbol can be expressed as:

[0118] wherein y(k) represents the predicted symbol.

[0119] Optionally, the prediction error can be expressed as: e(k) = d(k) - y(k)

[0120] wherein e(n) represents the prediction error, and d(n) represents the expected response.

[0121] Optionally, the mean square error gradient of the prediction error can be expressed as:

[0122] For w ff1 , w ff2 , w ff3 , w fb These coefficients can be updated using the chain rule. Taking w ff2 as an example:

[0123] wherein w ff2 (k,i) is a current value of w ff2 , μΔJ is a multiplication result of a learning rate and a mean square error gradient, and w ff2 (k+1,i) is used to update w ff2 (k,i), so as to update w ff2 . The parameters in the feedforward equalizer include the above wff1 , w ff2 , w ff3 The parameters in the feedback equalizer include w fb .

[0124] In the above embodiments, specific ways of updating the parameters in the feedforward equalizer and the feedback equalizer are provided. The construction of the feedback equalizer makes the final channel equalizer eliminate the intersymbol interference of the equalized symbols, and restore the received symbols to be closer to the real symbols sent from the sending end, thus being more accurate.

[0125] In some embodiments, the channel equalizer provided by the embodiments of the present application adds a Volterra series, and the channel equalizer provided by the embodiments of the present application is also called a Volterra-like neural network decision feedback equalizer (VINN DFE). Referring to FIG. 4, at the transmitter, PAM-8 symbols are generated by a 224GSa / s 3dB bandwidth about 65GHz arbitrary waveform generator (AWG). The analog signal is loaded into a thin film lithium niobate Mach-Zehnder modulator (TFLN MZM) with a 3dB bandwidth of 60GHz and a direct current half-wave voltage of 1.3V. The external cavity laser (ECL) operating wavelength is 1552.5nm, and the optical power is 13dBm. The output optical power of the TFLN-MZM is -5dBm. After 500m standard single-mode fiber (SSMF) transmission, the link loss is compensated by using an erbium-doped fiber amplifier (EDFA), and the optical signal is boosted to 6dBm by gain control. The receiving optical power (ROP) is adjusted by a variable optical attenuator (VOA) before entering a photodetector (PD). Finally, the electrical signal output of the PD is extracted by a 59GHz oscilloscope working at a sampling rate of 256GSa / s. The 10dB bandwidth of the system is 58GHz. FIG. 5 shows the relationship between BER and baud rate in the range of 120 to 130GBaud. The optical signal ROP is 6dBm. Since the system frequency response of the 10dB bandwidth is 58GHz, the BER performance deteriorates rapidly as the baud rate increases. When the complexity is limited to 100 MACs, the Volterra-like neural network feedforward equalizer (VINN) is superior to the Volterra nonlinear equalizer (VNLE), and the Volterra-like neural network decision feedback equalizer (VINN DFE) is superior to the Volterra nonlinear decision feedback equalizer (VDFE). The Volterra-like neural network feedforward equalizer (VINN) and the VDFE have similar performance, showing the nonlinearity of the nonlinear decision feedback compensation part of the feedforward filter. The calculated net rate is shown in FIG. 6. The net rate of the system is calculated according to the code rate corresponding to the hard decision BER. As the baud rate increases, the net rate first increases and then decreases. Compared with the VINN DFE provided by the embodiments of the present application, other equalizers are more affected by bandwidth limitation and nonlinearity, and the net rate decreases as the baud rate increases.

[0126] In some embodiments, the method for constructing a channel equalizer further comprises using the constructed channel equalizer to restore the symbols transmitted through the channel to obtain restored symbols.

[0127] By using the channel equalizer to restore the symbols received at the receiving end, channel distortion can be compensated for, and the original signal can be recovered.

[0128] It should be understood that although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of the steps is not strictly limited in sequence, and the steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of the steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or steps or stages in other steps.

[0129] Based on the same idea, the embodiments of the present application also provide a channel equalizer construction device for implementing the channel equalizer construction method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more channel equalizer construction device embodiments provided below can refer to the limitations of the channel equalizer construction method described above, and will not be repeated here.

[0130] In an exemplary embodiment, as shown in FIG. 7, a channel equalizer construction device is provided, which can be a receiver, comprising:

[0131] The acquisition module 701 is configured to acquire a symbol sequence sample, wherein the symbol sequence sample comprises a plurality of symbols arranged in sequence according to a symbol receiving order.

[0132] The sliding window processing module 702 is configured to perform sliding window processing on the plurality of symbols in the symbol receiving order using a sliding window, and for a plurality of target symbols currently in the sliding window, input the plurality of target symbols into a feedforward equalizer to obtain a feedforward output.

[0133] The acquisition module 701 is further configured to, for each feedforward output after the Lth feedforward output, acquire L feedforward outputs before the current feedforward output, input the L feedforward outputs into a feedback equalizer to obtain a feedback output, and determine a predicted symbol based on the current feedforward output and the feedback output.

[0134] The parameter adjustment module 703 is configured to determine a prediction error based on the predicted symbol and an expected response, and adjust parameters in the feedforward equalizer and the feedback equalizer based on the prediction error until the calculated prediction error is less than a preset value.

[0135] In some embodiments, the size of the sliding window is 2K1, and the number of symbols in the symbol sequence sample is N; the sliding window processing module 702 is configured to extract the first to the 2K1th symbols from the symbol sequence sample as a plurality of target symbols in the sliding window; after obtaining the corresponding feedforward output, extract the second to the 2K1+1th symbols from the symbol sequence sample as a plurality of target symbols in the sliding window; after obtaining the corresponding feedforward output, continue to move the sliding window by one symbol and perform the extraction operation until the last target symbol in the sliding window is the Nth symbol in the symbol sequence sample.

[0136] In some embodiments, the feedforward equalizer includes an input layer, a hidden layer, and an output layer; the number of neurons in the input layer is the same as the number of the plurality of target symbols, and the number of neurons in the hidden layer is the same as the order of the channel to be fitted by the equalizer; the sliding window processing module 702 is configured to input the plurality of target symbols into a plurality of first neurons in the input layer, respectively; each first neuron in the ith set inputs the received target symbol into the ith order neuron in the hidden layer; the ith set is a set of first neurons corresponding to a preset number of target symbols before the center symbol and a preset number of target symbols after the center symbol; each order neuron in the hidden layer performs a corresponding operation on the received target symbol to obtain an operation result and outputs the operation result to the output layer; the output layer adds the received operation results to obtain the feedforward output.

[0137] In some embodiments, the sliding window processing module 702 is configured to, for the ith order neuron, obtain an ith tap coefficient corresponding to each target symbol; multiply the ith order tap coefficient corresponding to each target symbol and the corresponding target symbol, sum the obtained multiplication results, add the sum result and a bias corresponding to the ith order neuron, take the ith power of the addition result to obtain the operation result.

[0138] In some embodiments, the obtaining module 701 is configured to obtain a decision coefficient corresponding to the ith feedforward output before the current feedforward output; multiply the decision coefficient corresponding to the ith feedforward output and the ith feedforward output to obtain a decision result corresponding to the ith feedforward output; add the decision results corresponding to the L feedforward outputs to obtain the feedback output.

[0139] In some embodiments, the parameter adjustment module 703 is configured to obtain a mean square error gradient of the prediction error; for the tap coefficient or the decision coefficient, obtain a current value of a current coefficient, multiply the current value by a learning rate to obtain a multiplication result; update the current value by subtracting the multiplication result from the current value.

[0140] In some embodiments, the plurality of symbols in the sequence of symbol samples are a plurality of symbols received by the receiving end in an actual working process of receiving the symbols transmitted by the transmitting end through the channel.

[0141] In some embodiments, the obtaining module 701 is further configured to add the current feedforward output and the feedback output to obtain the predicted symbol.

[0142] In some embodiments, the obtaining module 701 is further configured to pre-set a feedforward weight and a feedback weight, multiply the current feedforward output by the feedforward weight, multiply the feedback output by the feedback weight, and add the two multiplication results to obtain the predicted symbol.

[0143] In some embodiments, the parameter adjusting module 703 is further configured to subtract the predicted symbol from the expected response to obtain a prediction error.

[0144] In some embodiments, the constructed channel equalizer incorporates a Volterra series.

[0145] In some embodiments, the channel equalizer construction apparatus is further configured to use the constructed channel equalizer to restore the symbols transmitted through the channel to obtain restored symbols.

[0146] The various modules in the above channel equalizer construction apparatus can be implemented in whole or in part by software, hardware, and combinations thereof. The various modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform the operations corresponding to the various modules.

[0147] The features described in the embodiments of the above channel equalizer construction method are applicable to the embodiments of the channel equalizer construction apparatus, and various implementation manners of the embodiments of the channel equalizer construction apparatus can be referred to the related descriptions in the foregoing embodiments of the channel equalizer construction method, which will not be described herein again.

[0148] In an example embodiment, a computer device is provided, which can be a server, and an internal structure diagram of the computer device can be as shown in FIG. 8. The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store data such as parameters in the feedforward equalizer and the feedback equalizer. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with terminals outside through a network connection. The computer program is executed by the processor to implement a channel equalizer construction method.

[0149] Those skilled in the art can understand that the structure shown in FIG. 8 is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. Specifically, the computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0150] In an example embodiment, a computer device is provided, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the following steps:

[0151] obtain a symbol sequence sample, the symbol sequence sample including a plurality of symbols arranged in a symbol receiving order;

[0152] perform sliding window processing on the plurality of symbols in the symbol receiving order using a sliding window. For a plurality of target symbols in the sliding window, input the plurality of target symbols into a feedforward equalizer to obtain a feedforward output;

[0153] For each feedforward output after the Lth feedforward output, obtain L feedforward outputs before the current feedforward output, input the L feedforward outputs into a feedback equalizer to obtain a feedback output;

[0154] determine a predicted symbol based on the current feedforward output and the feedback output;

[0155] Based on the predicted symbol and expected response, a prediction error is determined, and based on the prediction error, parameters in the feedforward equalizer and the feedback equalizer are adjusted until the calculated prediction error is less than a preset value.

[0156] In one embodiment, the size of the sliding window is 2K1, and the number of symbols in the symbol sequence sample is N; when the processor executes the computer program, the following steps are further implemented: the first to the 2K1 symbols in the symbol sequence sample are extracted as the current target symbols in the sliding window; after obtaining the corresponding feedforward output, the second to the 2K1+1 symbols in the symbol sequence sample are extracted as the current target symbols in the sliding window; after obtaining the corresponding feedforward output, the sliding window is moved by one symbol at a time and the extraction operation is performed until the last target symbol in the sliding window is the Nth symbol in the symbol sequence sample.

[0157] In one embodiment, the feedforward equalizer comprises an input layer, a hidden layer and an output layer; the number of neurons in the input layer is the same as the number of target symbols, and the number of neurons in the hidden layer is the same as the order of the channel to be fitted by the equalizer; when the processor executes the computer program, the following steps are further implemented: the target symbols are respectively input into a plurality of first neurons in the input layer; each first neuron in the ith set inputs the received target symbol into the ith order neuron in the hidden layer; the ith set is a set of first neurons corresponding to a preset number of target symbols before the center symbol and a preset number of target symbols after the center symbol; each order neuron in the hidden layer performs a corresponding operation on the received target symbol, obtains an operation result and outputs the operation result to the output layer; the output layer adds the received operation results to obtain the feedforward output.

[0158] In one embodiment, when the processor executes the computer program, the following steps are further implemented: for the ith order neuron, the ith tap coefficient corresponding to each target symbol is obtained; each target symbol is multiplied by the ith order tap coefficient corresponding to the target symbol, the sum of the multiplication results is obtained, the sum result is added to the bias corresponding to the ith order neuron, the ith power of the addition result is obtained, and the operation result is obtained.

[0159] In one embodiment, when the processor executes the computer program, the following steps are further implemented: the decision coefficient corresponding to the ith feedforward output before the current feedforward output is obtained; the decision coefficient corresponding to the ith feedforward output is multiplied by the ith feedforward output to obtain the decision result corresponding to the ith feedforward output; the decision results corresponding to the L times of feedforward outputs are added to obtain the feedback output.

[0160] In one embodiment, the parameters of the feedforward equalizer comprise tap coefficients, and the parameters of the feedback equalizer comprise decision coefficients; and the processor, when executing the computer program, further implements the following steps: obtaining a gradient of mean square error of the prediction error; for the tap coefficients or the decision coefficients, obtaining a current value of a current coefficient, multiplying the gradient of mean square error by a learning rate to obtain a multiplication result, and updating the current value of the current coefficient by subtracting the multiplication result from the current value.

[0161] In one embodiment, the plurality of symbols in the symbol sequence sample are a plurality of symbols received by the receiving end in an actual working process of receiving the symbols transmitted by the transmitting end through the channel.

[0162] In one embodiment, the processor, when executing the computer program, further implements the following steps: adding the current feedforward output and the feedback output to obtain the predicted symbol.

[0163] In one embodiment, the processor, when executing the computer program, further implements the following steps: pre-setting a feedforward weight and a feedback weight, multiplying the current feedforward output by the feedforward weight, multiplying the feedback output by the feedback weight, and adding the two multiplication results to obtain the predicted symbol.

[0164] In one embodiment, the processor, when executing the computer program, further implements the following steps: subtracting the predicted symbol from the expected response to obtain the prediction error.

[0165] In one embodiment, the constructed channel equalizer adds a Volterra series.

[0166] In one embodiment, the processor, when executing the computer program, further implements the following steps: using the constructed channel equalizer to restore the symbols transmitted through the channel to obtain restored symbols.

[0167] The features described in the above embodiments of the channel equalizer construction method are applicable to the embodiments of the computer device, and various implementation manners of the embodiments of the computer device can be referred to the related descriptions in the above embodiments of the channel equalizer construction method, which will not be described herein again.

[0168] In one embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The processor, when executing the computer program, implements the following steps:

[0169] obtaining a symbol sequence sample, the symbol sequence sample comprising a plurality of symbols arranged in sequence according to symbol receiving order;

[0170] performing sliding window processing on the plurality of symbols in the symbol receiving order, for a plurality of target symbols currently in the sliding window, inputting the plurality of target symbols into a feedforward equalizer to obtain a feedforward output;

[0171] for each feedforward output after the Lth feedforward output, obtaining L feedforward outputs before the current feedforward output, inputting the L feedforward outputs into a feedback equalizer to obtain a feedback output;

[0172] based on the current feedforward output and the feedback output, determining a predicted symbol;

[0173] based on the predicted symbol and an expected response, determining a prediction error, and adjusting parameters in the feedforward equalizer and the feedback equalizer based on the prediction error until the calculated prediction error is less than a preset value.

[0174] In one embodiment, the size of the sliding window is 2K1, and the number of symbols in the symbol sequence sample is N; when the computer program is executed by the processor, the following steps are further implemented: extracting the first symbol to the 2K1th symbol from the symbol sequence sample as the plurality of target symbols currently in the sliding window; after obtaining the corresponding feedforward output, extracting the second symbol to the 2K1+1th symbol from the symbol sequence sample as the plurality of target symbols currently in the sliding window; after obtaining the corresponding feedforward output, continuing to move the sliding window by one symbol and perform the extraction operation until the last target symbol currently in the sliding window is the Nth symbol in the symbol sequence sample.

[0175] In one embodiment, the feedforward equalizer includes an input layer, a hidden layer, and an output layer; the number of neurons in the input layer is the same as the number of the plurality of target symbols, and the number of neurons in the hidden layer is the same as the order of the channel to be fitted by the equalizer; when the computer program is executed by the processor, the following steps are further implemented: inputting the plurality of target symbols into a plurality of first neurons in the input layer respectively; each first neuron in the ith set inputs the received target symbol into an ith-order neuron in the hidden layer; the ith set is a set of a preset number of target symbols before a center symbol and a preset number of target symbols after the center symbol in the plurality of target symbols, and each corresponding first neuron; each order of neuron in the hidden layer performs corresponding operation on the received target symbol to obtain an operation result and output the operation result to the output layer; the output layer adds the received operation results to obtain the feedforward output.

[0176] In one embodiment, the computer program, when executed on the processor, further implements the following steps: obtaining the i-th tap coefficient corresponding to each target symbol received; multiplying the i-th tap coefficient corresponding to each target symbol with the corresponding target symbol, summing the multiplication results, adding the sum result with the bias corresponding to the i-th neuron, and taking the i-th power of the addition result to obtain an operation result.

[0177] In one embodiment, the computer program, when executed on the processor, further implements the following steps: obtaining the decision coefficient corresponding to the i-th feedforward output before the current feedforward output; multiplying the decision coefficient corresponding to the i-th feedforward output with the i-th feedforward output to obtain a decision result corresponding to the i-th feedforward output; and adding the decision results corresponding to the L feedforward outputs to obtain the feedback output.

[0178] In one embodiment, the parameters of the feedforward equalizer include tap coefficients, and the parameters of the feedback equalizer include decision coefficients; the computer program, when executed on the processor, further implements the following steps: obtaining the gradient of the mean square error of the prediction error; obtaining the current value of the current coefficient for the tap coefficients or the decision coefficients, multiplying the gradient of the mean square error of the prediction error with a learning rate to obtain a multiplication result; and updating the current value by subtracting the multiplication result from the current value.

[0179] In one embodiment, the plurality of symbols in the symbol sequence sample are received by the receiving end in the actual working process of receiving the symbols transmitted by the sending end through the channel.

[0180] In one embodiment, the computer program, when executed on the processor, further implements the following steps: adding the current feedforward output and the feedback output to obtain the predicted symbol.

[0181] In one embodiment, the computer program, when executed on the processor, further implements the following steps: pre-setting a feedforward weight and a feedback weight, multiplying the current feedforward output with the feedforward weight, multiplying the feedback output with the feedback weight, and adding the two multiplication results to obtain the predicted symbol.

[0182] In one embodiment, the computer program, when executed on the processor, further implements the following steps: subtracting the predicted symbol from the expected response to obtain the prediction error.

[0183] In one embodiment, the constructed channel equalizer adds a Volterra series.

[0184] In one embodiment, the computer program, when executed on the processor, further implements the following steps: using the constructed channel equalizer to restore the symbols transmitted through the channel to obtain restored symbols.

[0185] The features described in the embodiments of the channel equalizer construction method apply to the embodiments of the computer readable storage medium, and various implementations of the embodiments of the computer readable storage medium can refer to the related descriptions in the foregoing embodiments of the channel equalizer construction method, which will not be described here again.

[0186] In one embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the following steps:

[0187] Obtaining a symbol sequence sample, the symbol sequence sample comprising a plurality of symbols arranged in a symbol receiving order;

[0188] Using a sliding window to perform sliding window processing on the plurality of symbols in the symbol receiving order, for a plurality of target symbols currently in the sliding window, inputting the plurality of target symbols into a feedforward equalizer to obtain a feedforward output;

[0189] For each feedforward output after the Lth feedforward output, obtaining L feedforward outputs before the current feedforward output, inputting the L feedforward outputs into a feedback equalizer to obtain a feedback output;

[0190] Based on the current feedforward output and the feedback output, determining a predicted symbol;

[0191] Based on the predicted symbol and an expected response, determining a prediction error, and adjusting parameters in the feedforward equalizer and the feedback equalizer based on the prediction error until the calculated prediction error is less than a preset value.

[0192] In one embodiment, the size of the sliding window is 2K1, and the number of symbols in the symbol sequence sample is N; when the computer program is executed by the processor, the following steps are further implemented: extracting the 1st to the 2K1th symbols from the symbol sequence sample as the plurality of target symbols currently in the sliding window; after obtaining the corresponding feedforward output, extracting the 2nd to the 2K1+1th symbols from the symbol sequence sample as the plurality of target symbols currently in the sliding window; after obtaining the corresponding feedforward output, continuing to move the sliding window by one symbol as a step and perform the extraction operation until the last target symbol currently in the sliding window is the Nth symbol in the symbol sequence sample.

[0193] In one embodiment, the feedforward equalizer comprises an input layer, a hidden layer and an output layer; the number of neurons in the input layer is the same as the number of target symbols, and the number of neurons in the hidden layer is the same as the order of the channel to be fitted by the equalizer; the computer program, when executed by the processor, further implements the following steps: inputting the plurality of target symbols into a plurality of first neurons in the input layer respectively; inputting the received target symbols into the i-th order neuron in the hidden layer by each first neuron in the i-th set; the i-th set is a set of first neurons corresponding to a preset number of target symbols before the center symbol and a preset number of target symbols after the center symbol; each order neuron in the hidden layer performs corresponding operation on the received target symbols, obtains an operation result and outputs the operation result to the output layer; and the output layer adds the received operation results to obtain the feedforward output.

[0194] In one embodiment, the computer program, when executed by the processor, further implements the following steps: for the i-th order neuron, obtaining the i-th tap coefficient corresponding to each target symbol; multiplying the i-th order tap coefficient corresponding to each target symbol and the corresponding target symbol, summing the obtained multiplication results, adding the sum result and the bias corresponding to the i-th order neuron, taking the i-th power of the addition result to obtain the operation result.

[0195] In one embodiment, the computer program, when executed by the processor, further implements the following steps: obtaining the decision coefficient corresponding to the i-th feedforward output before the current feedforward output; multiplying the decision coefficient corresponding to the i-th feedforward output and the i-th feedforward output to obtain the decision result corresponding to the i-th feedforward output; adding the decision results corresponding to the L times of feedforward outputs to obtain the feedback output.

[0196] In one embodiment, the parameters of the feedforward equalizer comprise tap coefficients, and the parameters of the feedback equalizer comprise decision coefficients; the computer program, when executed by the processor, further implements the following steps: obtaining the gradient of the mean square error of the prediction error; for the tap coefficients or the decision coefficients, obtaining the current value of the current coefficient, multiplying the learning rate by the gradient of the mean square error to obtain a multiplication result; updating the current value by using the current value minus the multiplication result.

[0197] In one embodiment, the plurality of symbols in the symbol sequence sample are a plurality of symbols received by the receiving end in the actual working process of receiving the symbols sent by the sending end through the channel.

[0198] In one embodiment, the computer program, when executed by the processor, further implements the following steps: adding the current feedforward output and the feedback output to obtain the predicted symbol.

[0199] In one embodiment, the computer program, when executed by the processor, further implements the following steps: presetting a feedforward weight and a feedback weight, multiplying the current feedforward output by the feedforward weight, multiplying the feedback output by the feedback weight, and adding the two multiplied results to obtain the prediction symbol.

[0200] In one embodiment, the computer program, when executed by the processor, further implements the following steps: subtracting the prediction symbol from the expected response to obtain a prediction error.

[0201] In one embodiment, the constructed channel equalizer is a Volterra series.

[0202] In one embodiment, the computer program, when executed by the processor, further implements the following steps: using the constructed channel equalizer to restore the symbol transmitted through the channel to obtain a restored symbol.

[0203] The features described in the above embodiments of the channel equalizer construction method apply to the embodiments of the computer program product, and various implementation manners of the embodiments of the computer program product can be referred to the related descriptions in the above embodiments of the channel equalizer construction method, which will not be described herein again.

[0204] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0205] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0206] The above-described embodiments are merely illustrative of several embodiments of the present application, and the description is relatively specific and detailed, but should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A method of constructing a channel equalizer, characterized by, The method comprises: obtaining a symbol sequence sample comprising a plurality of symbols arranged in a symbol receiving order; performing sliding window processing on the plurality of symbols in the symbol receiving order using a sliding window, for a plurality of target symbols currently in the sliding window, inputting the plurality of target symbols into a feedforward equalizer to obtain a feedforward output; for each feedforward output after the Lth feedforward output, obtaining L feedforward outputs before the current feedforward output, inputting the L feedforward outputs into a feedback equalizer to obtain a feedback output; based on the current feedforward output and the feedback output, determining a predicted symbol; based on the predicted symbol and an expected response, determining a prediction error, and adjusting parameters in the feedforward equalizer and the feedback equalizer based on the prediction error until the calculated prediction error is less than a preset value.

2. The method of claim 1, wherein, The size of the sliding window is 2K1, and the number of symbols in the symbol sequence sample is N; the sliding window processing on the plurality of symbols in the symbol receiving order using the sliding window comprises: extracting the first to the 2K1th symbols from the symbol sequence sample as the plurality of target symbols currently in the sliding window; after obtaining the corresponding feedforward output, extracting the second to the 2K1+1th symbols from the symbol sequence sample as the plurality of target symbols currently in the sliding window; after obtaining the corresponding feedforward output, continue to move the sliding window by one symbol as a step and perform the extraction operation until the last target symbol currently in the sliding window is the Nth symbol in the symbol sequence sample.

3. The method of claim 1, wherein, The feedforward equalizer comprises an input layer, a hidden layer, and an output layer; the number of neurons in the input layer is the same as the number of the plurality of target symbols, and the number of neurons in the hidden layer is the same as the order of the channel to be fitted by the equalizer; the inputting of the plurality of target symbols into the feedforward equalizer to obtain the feedforward output comprises: inputting the plurality of target symbols into a plurality of first neurons in the input layer respectively; each first neuron in the ith set inputs the received target symbol into the ith order neuron in the hidden layer; the ith set is a set composed of a preset number of target symbols before the center symbol and a preset number of target symbols after the center symbol in the plurality of target symbols, and each corresponding first neuron; each order neuron in the hidden layer performs corresponding operation on the received target symbol to obtain an operation result and output the operation result to the output layer; the output layer adds the received operation results to obtain the feedforward output.

4. The method of claim 3, wherein, the corresponding operation of each order neuron in the hidden layer on the received target symbol comprises: for the ith order neuron, obtaining an ith tap coefficient corresponding to each received target symbol; multiplying the ith order tap coefficient corresponding to each target symbol and the corresponding target symbol, summing the obtained multiplication results, adding the sum result and a bias corresponding to the ith order neuron, taking the ith power of the addition result to obtain the operation result.

5. The method of claim 1, wherein, the inputting of the L feedforward outputs into the feedback equalizer to obtain the feedback output comprises: acquire a decision coefficient corresponding to the i-th feedforward output before acquiring the current feedforward output; multiply the decision coefficient corresponding to the i-th feedforward output and the i-th feedforward output to obtain a decision result corresponding to the i-th feedforward output; add the decision results corresponding to the L feedforward outputs respectively to obtain the feedback output.

6. The method according to any one of claims 1 to 5, characterized in that, The parameters of the feedforward equalizer include tap coefficients, and the parameters of the feedback equalizer include decision coefficients; and the adjusting the parameters in the feedforward equalizer and the feedback equalizer based on the prediction error includes: calculating a mean square error gradient of the prediction error; for the tap coefficients or the decision coefficients, acquiring a current value of a current coefficient, multiplying the current value by a learning rate to obtain a multiplication result, subtracting the multiplication result from the current value to obtain a subtraction result, and updating the current value using the subtraction result.

7. The method of claim 1, wherein, The plurality of symbols in the symbol sequence sample are a plurality of symbols received by the receiving end in an actual working process of receiving the symbols sent by the sending end through the channel.

8. The method of claim 1, wherein, determining a prediction symbol based on the current feedforward output and the feedback output includes: adding the current feedforward output and the feedback output to obtain the prediction symbol.

9. The method of claim 1, wherein, determining a prediction symbol based on the current feedforward output and the feedback output includes: pre-setting a feedforward weight and a feedback weight, multiplying the current feedforward output by the feedforward weight, multiplying the feedback output by the feedback weight, and adding the two multiplication results to obtain the prediction symbol.

10. The method of claim 1, wherein, determining a prediction error based on the prediction symbol and an expected response includes: subtracting the prediction symbol from the expected response to obtain the prediction error.

11. The method of any one of claims 1-10, wherein, The constructed channel equalizer adds a Volterra series.

12. The method of any one of claims 1-11, further comprising: restoring the symbols sent through the channel using the constructed channel equalizer to obtain restored symbols.

13. A channel equalizer construction apparatus, comprising: an acquisition module configured to acquire a symbol sequence sample, the symbol sequence sample including a plurality of symbols arranged in a symbol receiving order; a sliding window processing module configured to perform sliding window processing on the plurality of symbols in the symbol receiving order using a sliding window, and for a plurality of target symbols currently in the sliding window, input the plurality of target symbols into a feedforward equalizer to obtain a feedforward output; the acquisition module is further configured to, for each feedforward output after an L-th feedforward output, acquire L feedforward outputs before a current feedforward output, input the L feedforward outputs into a feedback equalizer to obtain a feedback output, and determine a prediction symbol based on the current feedforward output and the feedback output; a parameter adjustment module configured to determine a prediction error based on the prediction symbol and an expected response, and adjust parameters in the feedforward equalizer and the feedback equalizer based on the prediction error until the calculated prediction error is less than a preset value.

14. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor implements the steps of the method of any one of claims 1 to 12 when executing the computer program.

15. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 12.

16. A computer program product comprising a computer program, characterized in that, The computer program, which is executed by a processor, implements the steps of the method according to any one of claims 1 to 12.

Images