Channel estimation method, apparatus, device, storage medium, and program product

Abstract

The application discloses a channel estimation method, device, equipment, storage medium and program product, relates to the wireless technical field, and the method comprises the following steps: preprocessing channel state information of a frequency domain and a time domain of a communication network to obtain preliminary channel estimation data; optimizing the preliminary channel estimation data by a preset spike neural network adopting a multilayer dynamic neuron structure to generate optimized channel estimation data; extracting time-frequency domain state information of a channel based on joint time-frequency domain data converted by the optimized channel estimation data; and performing channel characteristic analysis on the time-frequency domain state information to generate a channel estimation result. Since the application improves the spike neural network by adopting the multilayer dynamic neuron structure in advance, the channel characteristic analysis is performed on the time-frequency domain state information corresponding to the generated optimized channel estimation data, and a high-precision channel estimation result can be obtained, thereby effectively improving the accuracy and robustness of channel estimation.

Description

TECHNICAL FIELD

[0001] The present application relates to the wireless technical field, and particularly relates to a channel estimation method, device, equipment, storage medium and program product. BACKGROUND

[0002] With the rapid development of wireless communication technology, 5G network as a new generation of mobile communication technology has the characteristics of high bandwidth, low delay and high reliability, and is widely used in various scenes. However, 5G network still faces many technical challenges in practical application, one of which is the channel estimation and correction problem. In 5G network, high frequency band transmission signals will be affected by complex wireless propagation environment, resulting in the channel showing significant frequency selective characteristics. This frequency selectivity will cause channel estimation error, and then affect the accuracy and reliability of data transmission.

[0003] The existing channel estimation methods mainly include Minimum Mean Square Error Estimation (MMSE), Least Squares Estimation (LSE) and Matched Filter methods. These methods are generally difficult to accurately capture the dynamic changes of the channel when processing the frequency selective channel, resulting in insufficient channel estimation accuracy. SUMMARY

[0004] The main purpose of the present application is to provide a channel estimation method, device, equipment, storage medium and program product, which aims to solve the technical problem that the existing channel estimation method is generally difficult to accurately capture the dynamic changes of the channel when processing the frequency selective channel, resulting in insufficient channel estimation accuracy.

[0005] To achieve the above purpose, the present application provides a channel estimation method, which comprises:

[0006] Pretreating the channel state information of the frequency domain and time domain of the communication network to obtain preliminary channel estimation data;

[0007] Optimizing the preliminary channel estimation data through a preset Spike neural network to generate optimized channel estimation data, wherein the Spike neural network adopts a multi-layer dynamic neuron structure;

[0008] Extracting the time-frequency domain state information of the channel based on the joint time-frequency domain data converted by the optimized channel estimation data;

[0009] Performing channel characteristic analysis on the time-frequency domain state information to generate a channel estimation result.

[0010] In an embodiment, the step of optimizing the preliminary channel estimation data by the preset Spike neural network to generate optimized channel estimation data comprises:

[0011] Converting the preliminary channel estimation data into time-frequency dual-domain neuron firing patterns by the preset Spike neural network;

[0012] Determining a channel estimation error according to similarities between the time-frequency dual-domain neuron firing patterns;

[0013] Iteratively converging the channel estimation error based on the neuron synaptic weight of the Spike neural network iterative optimization until the channel estimation error reaches a preset threshold;

[0014] When the channel estimation error reaches the preset threshold, determining optimized channel estimation data based on the neuron synaptic weight and the time-frequency dual-domain neuron firing patterns.

[0015] In an embodiment, the step of extracting time-frequency domain state information of a channel based on the joint time-frequency domain data converted from the optimized channel estimation data comprises:

[0016] Time-frequency joint decomposing the optimized channel estimation data to obtain corresponding joint time-frequency domain data;

[0017] Extracting frequency responses of each subcarrier based on the joint time-frequency domain data;

[0018] Taking the frequency responses of each subcarrier as the time-frequency domain state information of the channel.

[0019] In an embodiment, the step of performing channel characteristic analysis on the time-frequency domain state information to generate a channel estimation result comprises:

[0020] Taking the joint time-frequency domain data as input data to construct a channel model;

[0021] Performing parameter fitting on the channel model through the frequency responses of each subcarrier to obtain channel characteristic parameters and determine a channel gain in the channel characteristic parameters;

[0022] Performing frequency domain interpolation and fitting on the channel gain to generate frequency domain channel estimation data;

[0023] Performing inverse Fourier transform on the frequency domain channel estimation data to obtain a channel estimation result.

[0024] In an embodiment, the step of preprocessing channel state information of a frequency domain and a time domain of a communication network to obtain preliminary channel estimation data comprises:

[0025] Channel state information in frequency domain and time domain is obtained by channel sampling of a communication network through a multi-antenna system;

[0026] The channel state information in frequency domain and time domain is denoised according to an adaptive filtering algorithm to obtain denoised channel state information;

[0027] The denoised channel state information is filtered through a band-pass filter to obtain preliminary channel estimation data.

[0028] In an embodiment, after the step of performing channel characteristic analysis on the time-frequency domain state information to generate channel estimation results, the method further comprises:

[0029] A precoding matrix and an equalization matrix are constructed according to the channel estimation results;

[0030] The original transmitted signal is channel-processed based on the precoding matrix and the equalization matrix to obtain a recovered original transmitted signal;

[0031] Frequency-selective errors in the recovered original transmitted signal are corrected to generate corrected channel state information;

[0032] Data transmission is performed according to the corrected channel state information.

[0033] In addition, to achieve the above object, the present application further provides a channel estimation device, which comprises:

[0034] A preprocessing module is configured to preprocess channel state information in frequency domain and time domain of a communication network to obtain preliminary channel estimation data;

[0035] A data optimization module is configured to optimize the preliminary channel estimation data through a preset Spike neural network to generate optimized channel estimation data, wherein the Spike neural network adopts a multi-layer dynamic neuron structure;

[0036] A time-frequency information module is configured to extract time-frequency domain state information of a channel based on joint time-frequency domain data converted from the optimized channel estimation data;

[0037] A result generation module is configured to perform channel characteristic analysis on the time-frequency domain state information to generate channel estimation results.

[0038] In addition, to achieve the above object, the present application further provides a channel estimation device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the channel estimation method as described above.

[0039] In addition, to achieve the above object, the application further provides a storage medium, which is a computer readable storage medium, and a computer program is stored on the storage medium, and the computer program is executed by a processor to implement the steps of the channel estimation method.

[0040] In addition, to achieve the above object, the application further provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps of the channel estimation method.

[0041] The one or more technical solutions provided by the application have at least the following technical effects: the channel estimation method provided by the application comprises: preprocessing channel state information in a frequency domain and a time domain of a communication network to obtain preliminary channel estimation data; optimizing the preliminary channel estimation data by a preset Spike neural network to generate optimized channel estimation data, wherein the Spike neural network adopts a multi-layer dynamic neuron structure; extracting time-frequency domain state information of a channel based on joint time-frequency domain data converted from the optimized channel estimation data; and performing channel characteristic analysis on the time-frequency domain state information to generate a channel estimation result.

[0042] Since the multi-layer dynamic neuron structure is used to improve the Spike neural network in advance, the optimized channel estimation data can be generated by the improved Spike neural network, thereby effectively improving the precision and robustness of channel estimation. Then, the high-precision channel estimation result can be obtained by performing channel characteristic analysis on the time-frequency domain state information of the channel extracted from the optimized channel estimation data. The existing channel estimation method avoids the difficulty in capturing the dynamic changes of the channel when processing the frequency-selective channel, thereby effectively improving the accuracy of channel estimation. BRIEF DESCRIPTION OF DRAWINGS

[0043] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the application and, together with the specification, serve to explain the principles of the application.

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without creative labor.

[0045] Figure 1 The overall flowchart of estimating and correcting the network channel provided by the application is shown in the following figure:

[0046] Figure 2 The flowchart provided by the channel estimation method embodiment one of the application is shown in the following figure:

[0047] Figure 3 A flowchart of a process for optimizing preliminary channel estimation data by a Spike neural network is provided for Embodiment One of the present application.

[0048] Figure 4 A flowchart is provided for Embodiment Two of the channel estimation method of the present application.

[0049] Figure 5 A module structure diagram of the channel estimation device of the present application is provided for Embodiment One of the present application.

[0050] Figure 6 A device structure diagram of the hardware operating environment involved in the channel estimation method of Embodiment One of the present application is provided.

[0051] The implementation, functional features and advantages of the present application will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0052] It should be understood that the specific embodiments described herein are merely intended to explain the technical solutions of the present application, and are not intended to limit the present application.

[0053] In order to better understand the technical solutions of the present application, the following will be described in detail in conjunction with the accompanying drawings and specific embodiments.

[0054] It should be noted that in the existing channel estimation methods, the minimum mean square error estimation (MMSE), the least square estimation (LSE) and the matched filtering method are generally difficult to accurately capture the dynamic changes of the channel when processing the frequency selective channel, resulting in insufficient channel estimation accuracy. Specifically, the existing technology has the following defects:

[0055] First, the performance of the existing channel estimation method under the frequency selective channel is not ideal. The traditional methods such as MMSE and LSE are difficult to achieve high-precision channel estimation when facing the frequency selective characteristics. This is because these methods mostly assume that the channel is flat in the frequency domain, and it is difficult to effectively handle the complex channel response characteristics caused by frequency selectivity, resulting in large deviation of the channel estimation result.

[0056] Second, the existing channel estimation method lacks adaptability in a variable channel environment. The channel characteristics in the 5G network will change with time and environment, and the traditional channel estimation and correction method is often static and cannot be dynamically adjusted according to the real-time channel state. Therefore, in actual application, the correction effect of these methods is not ideal, which affects the communication quality.

[0057] Third, the existing channel correction method lacks flexibility and adaptability. The traditional channel correction method mainly relies on fixed precoding and equalization algorithm, which is difficult to cope with channel changes in different scenarios. This fixed correction strategy is difficult to achieve effective channel correction in complex wireless environment, especially in high-speed mobile scenarios or severe multipath propagation conditions, thereby affecting the stability and reliability of data transmission.

[0058] To solve the above problems, the present application provides a new channel estimation and correction method, as shown in Figure 1 Figure 1 The overall flowchart of the network channel estimation and correction provided by the present application mainly includes the following processes: first, channel sampling is performed in the 5G network to obtain channel state information. Then, the channel state information is preprocessed to generate preliminary channel estimation data. Then, the channel estimation data is optimized to improve the accuracy of channel estimation. Then, frequency domain data extraction is performed, and according to the extracted data, channel model construction is performed to generate channel estimation results. Then, the correction process is performed, and the channel estimation results are corrected. After correction, an adaptive correction mechanism is introduced to cope with changes in the channel environment. Finally, the corrected data is fed back to the system and applied to the data transmission process. The specific implementation process can be referred to below.

[0059] By using the new channel estimation and correction method provided by the present application, the accuracy of channel estimation and the flexibility and adaptability of correction can be improved, the shortcomings of the prior art can be solved, and the high performance of 5G network in various application scenarios can be ensured, and the communication quality and user experience can be improved.

[0060] It should be noted that the execution subject of the present embodiment can be a computing service device with data preprocessing, channel data optimization, time-frequency domain conversion and channel characteristic analysis functions, such as a personal computer, a server, etc., or an electronic device capable of realizing the above functions, a channel estimation device for executing the channel estimation method of the present application, etc. The present embodiment does not limit this. The channel estimation device is taken as an example to describe the present embodiment and the following embodiments.

[0061] Based on this, the present embodiment provides a channel estimation method, as shown in Figure 2 , Figure 2 The flowchart provided by the first embodiment of the channel estimation method of the present application.

[0062] In the present embodiment, the channel estimation method includes steps S10-S40:

[0063] Step S10: Preprocess the channel state information of the frequency domain and time domain of the communication network to obtain preliminary channel estimation data.

[0064] ​It should be noted that the communication network can be a mobile communication network in wireless communication technology, and the communication network can include a 5G network, a 6G network, etc. The embodiment is not limited in this regard.

[0065] For ease of understanding, the following is described by taking a 5G network as an example. For example, the channel characteristics in the 5G network change with time and environment. The traditional channel estimation and correction method is generally static, and therefore, the real-time channel state of the 5G network can be dynamically estimated by using the channel estimation method of the embodiment.

[0066] It can be understood that the channel state information in the frequency domain and the time domain can be channel data including attenuation, phase shift, and other characteristics at different frequencies and channel characteristics represented in the time dimension. The channel state information in the time domain and the frequency domain records the channel state information in the time domain and the frequency domain of the communication network and is basic data for generating channel estimation of the embodiment.

[0067] It should be understood that the preliminary channel estimation data can be preliminary channel data obtained by preprocessing (for example, denoising, normalization, and filtering) the channel state information in the frequency domain and the time domain.

[0068] In the embodiment, after the channel estimation device samples the channel of the 5G network and obtains the channel state information in the frequency domain and the time domain, the channel state information in the frequency domain and the time domain can be preprocessed, such as denoising, normalization, and filtering, to obtain preliminary channel estimation data.

[0069] In a possible implementation, the step S10 of the embodiment can include the steps of: sampling the channel of the communication network by using a multi-antenna system to obtain channel state information in the frequency domain and the time domain; denoising the channel state information in the frequency domain and the time domain according to an adaptive filtering algorithm to obtain denoised channel state information; and filtering the denoised channel state information by using a band-pass filter to obtain preliminary channel estimation data.

[0070] It should be noted that the multi-antenna system can refer to an array system formed by using multiple transmitting antennas and multiple receiving antennas in a communication system.

[0071] Before channel sampling, the transmitting antennas and the receiving antennas in the multi-antenna system can be configured, the transmitting antennas and the receiving antennas can be caused to work in a predetermined frequency band, and the transmitting antennas and the receiving antennas can be caused to form a proper array arrangement in space. Therefore, the multi-antenna system can be used to sample the channel in the 5G network by using wireless communication technology to obtain channel state information in the frequency domain and the time domain. For example, the specific process is as follows:

[0072] S111. The transmitting antenna transmits a known training sequence signal X(t), where the training sequence signal has specific frequency and time domain characteristics and is used for channel estimation.

[0073] S112. The receiving antenna receives the training sequence signal transmitted through the wireless channel as the received signal Y(t), and records the frequency domain and time domain characteristics of the received signal.

[0074] S113. Calculate the channel impulse response h(t) by comparing the received signal Y(t) with the known training sequence signal X(t).

[0075] S114. Perform a Fourier transform on the channel impulse response h(t) to obtain the channel frequency domain response H(f).

[0076] S115. Record the acquired channel impulse response h(t) and channel frequency response H(f) as the time-domain and frequency-domain state information of the channel, and generate the basic data for channel estimation.

[0077] Understandably, an adaptive filtering algorithm can be an algorithm that automatically adjusts the filter parameters based on the statistical characteristics of the input signal to eliminate noise interference in the channel state information in the frequency and time domains. A bandpass filter can be an electronic filter that allows signals within a specific frequency band (i.e., the passband) in the channel state information in the frequency and time domains to pass through, while attenuating frequency components outside the passband, in order to extract the effective frequency components of the channel.

[0078] The above process can be achieved through an improved deep-heuristic adaptive filtering system, whose core idea originates from the adaptive processing capability of biological nervous systems to external signals. By simulating the signal capture, filtering, and optimization mechanisms of biological systems in complex dynamic environments, the system can dynamically adjust filtering parameters in real time to enhance its processing capability for specific signal characteristics (e.g., frequency domain and time domain information).

[0079] The specific functions of the deep-inspired adaptive filtering system include:

[0080] Transformation of preliminary channel estimation data: The preliminary channel estimation data (including time-domain and frequency-domain signals) collected from the channel is converted into a standardized input signal suitable for subsequent neural network processing.

[0081] Adaptive filtering: Adjusts filter parameters based on dynamic changes in channel conditions (e.g., noise levels, frequency variations) to reduce noise and improve the quality of channel estimation data.

[0082] Bio-inspired mechanism: By simulating the synaptic weight adjustment and memory reinforcement mechanisms in biological neural networks, the system can learn channel characteristics and gradually optimize the filtering effect.

[0083] The deep heuristic adaptive filtering system is different from the traditional filter. The parameters of the traditional filter are usually fixed, while the deep heuristic adaptive filtering system can dynamically adjust the parameters to cope with complex 5G channel environment. By simulating the learning and adjustment of biological mechanisms, the system can not only handle the current channel state, but also remember the past characteristics, providing a long-term optimization strategy.

[0084] Therefore, after sampling the channel state information in the frequency domain and the time domain, the deep heuristic adaptive filtering system can further process the channel state information in the frequency domain and the time domain, including denoising, normalization and filtering process, to generate preliminary channel estimation data. In combination with the above description, the specific process is as follows:

[0085] S121, denoising the collected channel state information in the frequency domain and the time domain, which can use an adaptive filtering algorithm to eliminate noise interference and obtain denoised channel state information.

[0086] S122, normalizing the denoised channel state information to adjust the signal amplitude of the channel state information to a predetermined range.

[0087] S123, filtering the normalized channel state information, using the above band-pass filter to extract the effective frequency components of the channel, and recording the filtered channel state information as preliminary channel estimation data.

[0088] In this embodiment, by configuring the transmitting antennas and receiving antennas in the multi-antenna system to work in a predetermined frequency band and form a proper array arrangement in space, and further transmitting known training sequence signals, the receiving antennas record the frequency domain and time domain characteristics of the received signals, further calculate the channel impulse response, and obtain the frequency domain response of the channel through Fourier transform. Finally, the obtained channel impulse response and frequency domain response are recorded as channel time domain and frequency domain state information, generating basic data for channel estimation. By using the multi-antenna system and Fourier transform combination for the first time, the accuracy of channel estimation can be significantly improved. Subsequently, the channel estimation device also pre-processes the collected channel state information in the frequency domain and the time domain, including using an adaptive filtering algorithm to eliminate noise interference, normalizing the denoised channel state information to adjust the signal amplitude of the channel state information to a predetermined range, and using a band-pass filter to extract the effective frequency components of the channel. By combining the adaptive filtering algorithm with the band-pass filter for the first time, the effect of channel preprocessing can be significantly improved, providing more accurate data for subsequent channel estimation optimization.

[0089] Step S20: optimizing the preliminary channel estimation data through a preset Spike neural network to generate optimized channel estimation data, wherein the Spike neural network adopts a multi-layer dynamic neuron structure.

[0090] It should be noted that the spiking neural network (SNN) can be a relatively new artificial neural network improved by using a multi-layer dynamic neuron structure, and the inspiration comes from the working mode of the biological nervous system. Unlike traditional artificial neural networks, SNN transmits signals in the form of discrete spikes rather than continuous real values. The spiking neural network simulates the communication process between neurons by pulses, and can combine signal information in time and space. The working principle is as follows:

[0091] In the pulse transmission of the spiking neural network, the state of each neuron is represented by membrane potential, and when the input signal accumulates to exceed the threshold, the neuron will fire a pulse and transmit it to the adjacent neuron. At the same time, for the time-frequency domain coding of the spiking neural network, the time and frequency characteristics of the signal can be combined, and the pulse behavior of the neuron excitation can be adjusted through the time difference and frequency distribution, so as to realize the coding and processing of complex signals. And SNN adjusts the connection weight between neurons through synaptic plasticity rules to optimize network performance for dynamic learning.

[0092] Through the above-mentioned spiking neural network, it can be good at processing time domain and frequency domain information at the same time, and is suitable for the dynamic characteristics of 5G channel. Moreover, since SNN only consumes computing resources when neurons fire pulses, it has higher computing efficiency than traditional ANN, so as to accurately model the frequency selectivity characteristics and multipath effects of 5G channel.

[0093] It can be understood that the optimized channel estimation data can be obtained by continuously iterating and improving the communication channel characteristics of the preliminary channel estimation data through the above-mentioned spiking neural network.

[0094] In this embodiment, the channel estimation device can optimize the preliminary channel estimation data through the above-mentioned spiking neural network improved by using a multi-layer dynamic neuron structure, and after continuous iteration and improvement, the optimized channel estimation data can be generated.

[0095] In a possible implementation, the step S20 of the embodiment can include the steps of: converting the preliminary channel estimation data into a time-frequency dual-domain neuron firing mode through a preset spiking neural network; determining a channel estimation error according to the similarity between the time-frequency dual-domain neuron firing modes; iteratively converging the channel estimation error based on the neuron synaptic weight of the spiking neural network iterative optimization until the channel estimation error reaches a preset threshold; and determining the optimized channel estimation data based on the neuron synaptic weight and the time-frequency dual-domain neuron firing mode when the channel estimation error reaches the preset threshold.

[0096] It should be noted that the preset threshold is a threshold for iterative channel estimation error, for example, the error reaches one percent, that is, the iteration process is stopped, and the embodiment is not limited to this.

[0097] For example, in order to facilitate understanding of the process of optimizing the spiking neural network, reference is made to Figure 3 , Figure 3 A flowchart for optimizing the preliminary channel estimation data by the spiking neural network is provided in the embodiment of the present application. The preliminary channel estimation data is optimized by using the improved spiking neural network, so that the accuracy of channel estimation can be optimized. The specific process is as follows:

[0098] S21, the above heuristic adaptive filtering system can obtain input signals suitable for spiking neural network processing after processing: preliminary channel estimation data and the received signal Y(t), the input signal needs to contain time domain and frequency domain information. The spiking neural network receives the processed channel data, and further mines the complex relationship in the data through the joint coding mechanism of the time domain and the frequency domain.

[0099] S22, an improved SNN is constructed, which contains a multi-layer dynamic neuron structure to adapt to the frequency and dynamic change characteristics of the 5G network:

[0100] SNN={L1,L2,…,L n};

[0101] Wherein, SNN={L1,L2,…,L n} represents the spiking neural network, L i represents the i-th layer of neurons, and each layer of neurons is composed of multiple dynamic neurons to adapt to different frequencies and dynamic changes of channel characteristics.

[0102] S23, in the improved spiking neural network, the time domain and frequency domain joint coding method is used to convert the preliminary channel estimation data and the received signal Y(t) into spiking sequences to form a time-frequency dual-domain neuron firing pattern:

[0103]

[0104] Wherein, S i (t,f) represents the firing state of the i-th neuron in the time-frequency domain, which takes 1 or 0, indicating whether the neuron is fired at a certain time-frequency point; θ i is the firing threshold of the i-th neuron; t represents the upper limit of integration, which represents the current time point; represents the preliminary channel estimation data, which represents the response of the channel at time τ; f is the variable of the exponential function, which represents the current frequency point; e -j2πfτis one of the integrand functions of the integral, representing the frequency domain information of the channel. The result of the integral represents the cumulative effect of the channel response at time t and frequency f, and when the result of the integral is greater than the firing threshold θ i , the neuron is fired at the time-frequency point, taking a value of 1, otherwise it is not fired, taking a value of 0.

[0105] S24, calculate the similarity between the firing patterns of the time-frequency dual-domain neurons to calculate the channel estimation error ∈(t):

[0106]

[0107] wherein, and respectively represent the firing patterns of the i-th neuron in the time-frequency domain on the preliminary channel estimation data and the received signal Y(t), and ∈(t) is the overall error of the channel estimation.

[0108] S25, dynamically adjust the synaptic weights W(t,f) of the neurons according to the improved Spike-Timing-Dependent Plasticity rule:

[0109]

[0110] wherein, ΔW ij (t,f) is the synaptic weight adjustment amount, Δt and Δf are the Spike-Timing-Difference in the time domain and the frequency domain respectively, τ + , τ - , are time constants, A + and A - are learning rates.

[0111] S26, use a multi-scale learning mechanism to simultaneously adjust the synaptic weights of the neurons at different time and frequency scales to converge the channel estimation error:

[0112]

[0113] wherein, φ k (t,f) is a multi-scale adjustment factor, and η is a learning rate.

[0114] S27, iteratively update the synaptic weights W(t,f) of the neurons until the channel estimation error ∈(t) converges to a preset threshold:

[0115] while ∈(t) > ∈ threshond : W(t,f) ← W(t,f) + ΔW(t,f);

[0116] wherein, ∈ threshold is a preset error convergence threshold.

[0117] S28, optimizing the synaptic weights W(t, f) in the spiking neural network by repeated iterations, in each iteration, calculating channel estimation data

[0118]

[0119] wherein, is the optimized frequency domain channel estimation data, W ij (t, f) is the synaptic weight, S i (t, f) is the firing state of the i-th neuron in the time-frequency domain. By inverse Fourier transform, the optimized frequency domain channel estimation data is converted into time domain channel estimation data

[0120]

[0121] wherein, is the optimized time domain channel estimation data, denotes the inverse Fourier transform operator.

[0122] The optimized time domain channel estimation data and the frequency domain channel estimation data are recorded as the final optimized channel estimation data.

[0123] It should be noted that, as Figure 3 indicated, in the process of constructing the SNN, S22 includes the following steps simultaneously:

[0124] S221, constructing an improved spiking neural network model, which is composed of multiple layers of dynamic neurons, and the neuron structure of each layer is as follows:

[0125] S222, in each layer of dynamic neurons, the potential of the neuron is updated according to the following formula:

[0126] V i (t) = V i (t-1) + ∑ j W ij (t)S j (t-1) - λV i (t-1);

[0127] wherein, V i (t) represents the membrane potential of the i-th neuron at time t, W ij (t) is the synaptic weight between the i-th neuron and the j-th neuron, S j (t-1) is the output signal of the j-th neuron at time t-1, and λ is the membrane potential decay coefficient.

[0128] S223, when the membrane potential V i (t) exceeds the threshold value θ of the neuron i , the neuron i fires a spike and resets its membrane potential:

[0129]

[0130] where S i (t) represents the output signal of the i-th neuron at time t, θ i is the threshold value of the i-th neuron.

[0131] S224, for each layer of neurons, the inter-layer connection weight W ij (t) is updated using the channel estimation data and the received signal Y(t):

[0132]

[0133] where, represents the input signal received by the i-th neuron at time t.

[0134] S225, in the multi-layer dynamic neuron structure, the inter-layer weight W ij (t) is adjusted to adapt to the frequency and dynamic characteristics of the 5G network.

[0135] S226, further explanations can be made for the network structure layers:

[0136] L1: input layer. The input layer is responsible for receiving the preliminary channel estimation data, formatting it and inputting it into the neural network. This layer contains several input neurons, each corresponding to a channel estimation feature. If the channel state information includes multiple frequency bands and time domain samples, the number of input layer neurons should be consistent with the dimension of the channel estimation data. The data is standardized and normalized in the input layer to ensure that the data range is suitable for the input of the activation function of the subsequent neurons.

[0137] L2: first hidden layer (construction and improvement of spiking neural network). The first hidden layer is composed of several spiking neurons, which is responsible for feature extraction of channel data. The number of neurons in this layer can be determined according to the complexity of the input data, which is twice the number of input layer neurons to ensure sufficient feature extraction capability. Spiking neurons use a biological-inspired temporal coding strategy, which can generate a spike when the signal reaches a certain threshold, thereby preserving the timing information of the signal. After the spike is generated, it will propagate to the next layer for further processing.

[0138] L3: Second hidden layer (time-frequency joint coding layer). This layer performs time-frequency joint coding on the impulse signals passed down from the previous layer. The number of neurons in this layer is usually comparable to that of the first hidden layer, but can be adjusted according to the specific time-frequency decomposition method (such as wavelet transform or Fourier transform). Time-frequency joint coding ensures the fidelity of channel characteristics in both time and frequency domains, providing more accurate data support for subsequent error calculation and optimization.

[0139] L4: Third hidden layer (error calculation and dynamic weight adjustment layer). The third hidden layer calculates the channel estimation error by receiving the time-frequency coded signals from the previous layer. The number of neurons in this layer is twice that of the output neurons of the previous layer to ensure sufficient error detection capability. After error calculation, error optimization is performed through dynamic adjustment of synaptic weights. The update of synaptic weights uses an adaptive method based on Spike Time-Dependent Plasticity Rule (STDP), achieving rapid error convergence.

[0140] L5: Output layer (multi-scale learning mechanism and optimized output). The output layer is the last layer of the neural network, which introduces a multi-scale learning mechanism to achieve more refined optimization data for synaptic weights. The output layer usually contains a small number of neurons, each corresponding to a final optimized channel estimation data output. In this layer, after multiple iterations of optimization and fine-tuning of synaptic weights, high-precision channel estimation data is finally generated.

[0141] In this embodiment, an improved Spike Neural Network (SNN) is proposed to further optimize the accuracy of channel estimation. The model contains multiple layers of dynamic neuron structure to adapt to the frequency and dynamic characteristics of 5G networks, uses a time-frequency joint coding method to convert the preliminary channel estimation data and the received signals into Spike sequences, forms a time-frequency dual-domain neuron firing pattern, evaluates the channel estimation error by calculating the similarity between the dual-domain neuron firing patterns, and dynamically adjusts the neuron synaptic weights according to the improved Spike Time-Dependent Plasticity Rule, iteratively updates until the channel estimation error converges to the preset threshold. By applying Spike Neural Network to channel estimation optimization, the accuracy and robustness of channel estimation are effectively improved.

[0142] Step S30: Extract the time-frequency domain state information of the channel based on the joint time-frequency domain data converted from the optimized channel estimation data.

[0143] It should be noted that the joint time-frequency domain data can be data that combines both time and frequency dimensions. Through joint time-frequency domain data, the frequency components and their changes at different time points can be simultaneously displayed, providing a more comprehensive characterization of the characteristics and dynamic changes of channel signals.

[0144] It can be understood that the time-frequency domain state information can be state characteristic information about the optimized channel estimation data embodied in two dimensions of time and frequency. The time-frequency domain state information can include real and imaginary parts, amplitudes and phases of channel responses, and the like.

[0145] In the embodiment, the channel estimation device can first convert the optimized channel estimation data from the time domain to joint time-frequency domain data, and then extract time-frequency domain state information of the channel by using a suitable time-frequency window function and dynamically adjusting parameters according to channel characteristics of the 5G network, so as to improve the accuracy of channel characteristic analysis.

[0146] In a feasible embodiment, the step S30 can include the steps of: performing joint time-frequency domain decomposition on the optimized channel estimation data to obtain corresponding joint time-frequency domain data; extracting frequency responses of each subcarrier based on the joint time-frequency domain data; and taking the frequency responses of each subcarrier as time-frequency domain state information of the channel.

[0147] It should be noted that the frequency response of the subcarrier can be a response characteristic of the subcarrier to different frequency signals in the channel data, which describes the gain, attenuation, phase shift and the like of the subcarrier at different frequencies.

[0148] For example, after the optimized channel estimation data is generated, the time domain channel data can be converted into frequency domain data by Fourier transform based on the frequency selective characteristic, and the frequency responses of each subcarrier are extracted. The specific process is as follows:

[0149] S31, performing joint time-frequency domain decomposition on the above-mentioned optimized channel estimation data to convert the time domain channel data into joint time-frequency domain data:

[0150]

[0151] wherein H(t,f) represents a time-frequency domain channel response, is the nth time domain channel estimation data, N is the number of decomposition points, and g(t-n) is a time-frequency window function.

[0152] S32, selecting a suitable time-frequency window function g(t-n), and the time-frequency window function is dynamically adjusted according to the channel characteristics of the 5G network:

[0153]

[0154] wherein β is a time attenuation parameter, γ is a frequency modulation parameter, and is dynamically adjusted according to the time and frequency selective characteristics of the channel.

[0155] For the purpose of understanding the process of dynamic adjustment, an example is given as follows:

[0156] The adjustment of the time decay parameter β is as follows: when there is a large multipath effect in the channel environment (i.e. the signal arrives at the receiving end on different paths, resulting in a large time delay), the value of β is increased to increase the decay speed in the time domain, thereby better suppressing the interference caused by multipath propagation. In the case of a stable channel environment and a small time delay, the value of β is appropriately reduced to maintain the stability of the channel for a long time and reduce frequent dynamic adjustment.

[0157] The adjustment of the frequency modulation parameter γ is as follows: when the channel has strong frequency selectivity (i.e. the channel has different gains or attenuations for signals of different frequencies), the value of γ can be increased to make the time-frequency window function more focused on the specific frequency components of the signal, thereby more accurately capturing the channel characteristics in the frequency domain. If the frequency response of the channel is relatively smooth, the value of γ can be reduced to expand the frequency coverage of the time-frequency window, so that the estimated data has a wider frequency response and enhances the stability of the overall channel estimation.

[0158] S33, using the time-frequency domain channel response H(t,f), extract the frequency response of each subcarrier:

[0159] H sub (t,f)=H(t,f)·δ(f-f k );

[0160] wherein H sub (t,f) represents the time-frequency domain response of the subcarrier, δ(f-f k ) is a unit impulse function, and f k is the frequency of the kth subcarrier.

[0161] S34, extract the amplitude and phase of the time-frequency domain response H sub (t,f) of each subcarrier:

[0162]

[0163] wherein |H sub (t,f)| represents the amplitude of the time-frequency domain response, arg(H sub (t,f)) represents the phase of the time-frequency domain response, and represent the real part and the imaginary part of the time-frequency domain channel response, respectively.

[0164] S35, record the extracted time-frequency domain response of each subcarrier as the time-frequency domain state information of the channel.

[0165] In this embodiment, the time-frequency joint decomposition technology is adopted to convert the time domain channel data into joint time-frequency domain data, and the frequency response of each subcarrier is extracted. By selecting a suitable time-frequency window function, the parameters are dynamically adjusted according to the channel characteristics of the 5G network, the frequency response of the subcarrier is accurately extracted, and the amplitude and phase are extracted, which not only improves the accuracy of channel characteristic analysis, but also provides high-precision data support for subsequent channel model construction and channel correction.

[0166] Step S40: performing channel characteristic analysis on the time-frequency domain state information to generate channel estimation results.

[0167] It should be noted that the channel estimation results can be data or information obtained after estimating the communication channel characteristics of the time-frequency domain state information. The obtained results can include the amplitude response, phase response, noise characteristics, etc. of the channel.

[0168] In this embodiment, the channel estimation device can first extract the channel characteristic parameters of the channel in the time-frequency domain state information, and then analyze and transform the channel characteristic parameters to obtain the final channel estimation results.

[0169] In the technical scheme provided in this embodiment, after the channel estimation device performs channel sampling on the 5G network to obtain the channel state information in the frequency domain and the time domain, it can perform preliminary processing such as denoising, normalization and filtering on the channel state information in the frequency domain and the time domain to obtain preliminary channel estimation data. Then the improved Spike neural network with a multi-layer dynamic neuron structure can be used to optimize the preliminary channel estimation data, and after continuous iteration and improvement, optimized channel estimation data can be generated. Then the optimized channel estimation data can be converted from the time domain to joint time-frequency domain data, and then the time-frequency domain state information of the channel can be extracted by using a suitable time-frequency window function and dynamically adjusting the parameters according to the channel characteristics of the 5G network, so as to improve the accuracy of channel characteristic analysis. Finally, the channel characteristic parameters of the channel in the time-frequency domain state information are extracted, and then the channel characteristic parameters are analyzed and transformed to obtain the final channel estimation results. Since the Spike neural network is improved in this embodiment by using a multi-layer dynamic neuron structure, the improved Spike neural network can generate optimized channel estimation data, which effectively improves the accuracy and robustness of channel estimation. Then, by performing channel characteristic analysis on the time-frequency domain state information of the channel extracted from the optimized channel estimation data, high-precision channel estimation results can be obtained. This avoids the difficulty of existing channel estimation methods in capturing the dynamic changes of the channel when processing frequency-selective channels, thereby effectively improving the accuracy of channel estimation. In this way, in a variable channel environment, the transmission accuracy and stability of the signal can be ensured, making the channel correction more flexible and being able to effectively cope with complex channel changes in the 5G network, thereby improving the communication quality and user experience.

[0170] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as the above-mentioned embodiment one can refer to the above description, and the subsequent description will not be repeated. On this basis, please refer to Figure 4 , Figure 4 The flowchart of the second embodiment of the channel estimation method of the present application is provided. The step S40 of the present example further comprises steps S41-S44:

[0171] Step S41: the joint time-frequency domain data is taken as input data to construct a channel model.

[0172] It should be noted that the channel model is a mathematical model for describing the channel characteristics experienced by the signal in the transmission process. By establishing the channel model, the unique channel characteristics of the frequency response of each subcarrier can be accurately reflected.

[0173] Step S42: the channel model is parameter fitted by the frequency response of each subcarrier to obtain channel characteristic parameters, and the channel gain in the channel characteristic parameters is determined.

[0174] It should be noted that the channel characteristic parameter can be a quantitative index for describing the specific characteristics of the communication channel of the frequency response of the subcarrier, such as attenuation coefficient, delay, etc.

[0175] Step S43: the channel gain is frequency domain interpolated and fitted to generate frequency domain channel estimation data.

[0176] Step S44: the frequency domain channel estimation data is inverse Fourier transformed to obtain a channel estimation result.

[0177] Exemplarily, after obtaining the frequency response of the subcarrier, the channel model can be constructed in advance, the channel characteristic analysis is performed by using the above-mentioned frequency response data, and the channel estimation result is generated, and the specific process is as follows:

[0178] Firstly, the channel model is constructed, the time-frequency domain channel response H(t,f) is taken as input data, and a mathematical model containing channel characteristics is established. Among them, the time-frequency domain channel response data H(t,f) is input into the model. The channel model includes the response characteristics of time and frequency. In the process of constructing the model, statistical signal processing methods such as least square method or maximum likelihood estimation are used to fit the input channel response data to obtain a mathematical model reflecting the channel characteristics. The core of this channel model is to dynamically adapt to the change of the channel, and to make a reasonable estimation on the multipath effect, time delay and frequency selectivity characteristics of the signal.

[0179] Then, the above-mentioned frequency domain response data H sub(t,f), parameter fitting is performed on the channel model to extract channel characteristic parameters, including multipath delay, fading coefficient and frequency selective characteristic. Through the sub-band response data H sub (t,f), further parameter fitting is performed on the channel model to accurately extract channel characteristic parameters. The fitting process includes obtaining parameters such as multipath delay, fading coefficient and frequency selective characteristic. The parameters are estimated by gradient optimization algorithm in frequency and time, such as gradient descent method. For the frequency selective characteristic, the frequency response feature extraction technology can be used to parameterize the frequency correlation of the channel characteristic to describe the gain or attenuation of the channel to different frequency signals.

[0180] Then, based on the extracted channel characteristic parameters, the channel gain G(f) is calculated. According to the extracted channel characteristic parameters, the gain G(f) of the channel is calculated, which reflects the amplification or attenuation effect of the channel at different frequencies. In the gain calculation process, interpolation method or smoothing algorithm is used to interpolate the discrete gain data to generate a continuous gain function. Finally, the generated gain function G(f) is used to perform frequency domain fitting on the channel response, so that the channel estimation result has better smoothness and frequency consistency.

[0181] Finally, by performing frequency domain interpolation and fitting on the channel gain G(f), continuous frequency domain channel estimation data is generated. The estimated frequency domain channel estimation data

[0182] In the technical scheme provided in the embodiment, further based on the frequency selective characteristic, the time domain channel data is converted into frequency domain data by Fourier transform, the frequency response of each subcarrier is extracted, the channel model is constructed, the parameter fitting is performed, the channel characteristic parameters are extracted, the channel gain is calculated, the continuous frequency domain channel estimation data is generated by performing frequency domain interpolation and fitting on the channel gain, and the time domain channel estimation data is converted by inverse Fourier transform. Finally, the channel estimation result is generated. By proposing the combination of frequency selective characteristic and Fourier transform, the fine correction of the channel is performed, which can significantly improve the stability and reliability of data transmission.

[0183] Based on the above embodiments of the present application, in the third embodiment of the present application, the same or similar contents as the above embodiments can be referred to the above introduction, and will not be described in detail. On this basis, as shown in Figure 1 , after obtaining the channel estimation result, correction can be performed.

[0184] The steps after step S40 of the example further include the steps of: constructing a precoding matrix and an equalization matrix according to the channel estimation result; performing channel processing on the original transmission signal based on the precoding matrix and the equalization matrix to obtain a recovered original transmission signal; correcting a frequency selective error in the recovered original transmission signal to generate corrected channel state information; and performing data transmission according to the corrected channel state information.

[0185] For example, after obtaining the channel estimation result, the channel can be corrected by using a precoding and equalization algorithm according to the channel estimation result to correct a frequency selective error in the channel, and the specific process is as follows:

[0186] S51, according to the above frequency domain channel estimation data Construct a precoding matrix P:

[0187]

[0188] Where P is the precoding matrix, V opt , Σ opt , are singular value decomposition results of the channel matrix, U H is the conjugate transpose of U.

[0189] S52, use the precoding matrix P to precode the transmission signal to generate a precoded signal X pre :

[0190] X pre = PX;

[0191] Where X pre is the precoded signal, and X is the original transmission signal.

[0192] S53, at the receiving end, construct an equalization matrix W using the estimated frequency domain channel estimation data .

[0193] S54, after receiving the precoded signal Y, use the equalization matrix W to perform equalization processing on the received signal:

[0194] Y eq = WY;

[0195] Where Y eq is the equalization processed signal, and Y is the received signal.

[0196] S55, inverse transform the equalization processed signal Y eq to recover the original transmission signal X rec :

[0197]

[0198] wherein X rec is the recovered signal, U opt 、 is the SVD inverse transform result of the channel matrix.

[0199] S56, the recovered signal X rec is subjected to error detection and correction, and finally the corrected signal data is generated, and the frequency selective error in the channel is corrected.

[0200] S57, an adaptive correction mechanism is introduced, and the correction strategy is dynamically adjusted based on the real-time channel characteristics to cope with the changes in the channel environment.

[0201] S58, the corrected channel state information is fed back to the system and applied to the data transmission process.

[0202] In the technical scheme provided in the embodiment, a dynamic adaptive channel correction mechanism is introduced, the correction strategy is dynamically adjusted according to the real-time channel characteristics, the precoding matrix and the equalization matrix are constructed, the estimated channel response is used for precoding and equalization processing, and the SVD decomposition and inverse Fourier transform algorithms are used. In the variable channel environment, the transmission accuracy and stability of the signal can be ensured, the channel correction is more flexible, the complex channel changes in the 5G network can be effectively coped with, and the communication quality and user experience are improved.

[0203] Exemplarily, in order to help understand the implementation process of the channel estimation method obtained after the above-mentioned embodiments one and two are combined, a specific example is given to illustrate its effect, specifically:

[0204] In the central business district of a certain large city, the deployment of 5G network has been perfected, however, due to the high-rise buildings and complex wireless propagation environment, the channel estimation error caused by the frequency selective channel and the correction effect in the variable channel environment are not ideal. In order to improve the communication quality and user experience of 5G network, a main street in the area is selected for field test and data collection in the embodiment. The test time is set at the peak period of weekdays, from 8:00 to 10:00 in the morning and from 5:00 to 7:00 in the afternoon, so as to ensure the authenticity, reliability and representativeness of the data.

[0205] In the test area, a set of multi-antenna systems is deployed in the embodiment, including 8 transmitting antennas and 8 receiving antennas, working in the high frequency band of 28GHz. First, configure the multi-antenna system to make the transmitting antennas and receiving antennas form appropriate array arrangements in the predetermined frequency band. The transmitting antennas transmit known training sequence signals, which have specific frequency domain and time domain characteristics, for channel estimation. The receiving antennas receive the training sequence signals transmitted through the wireless channel and record the frequency domain and time domain characteristics of the received signals. Then, according to the comparison between the received signals and the known training sequence signals, the channel impulse response is calculated.

[0206] Next, the collected frequency domain and time domain channel state information is denoised, and an adaptive filtering algorithm is used to eliminate noise interference to obtain denoised channel state information. The denoised channel state information is normalized to adjust the signal amplitude to a predetermined range. Then, the normalized channel state information is filtered, and a bandpass filter is used to extract the effective frequency components of the channel. The filtered channel state information is recorded as the preliminary channel estimation data.

[0207] The preliminary channel estimation data and the received signals are input into the improved bio-inspired adaptive filtering system, and the preliminary channel estimation data and the received signals are converted into input signals suitable for Spike neural network processing, which need to contain time domain and frequency domain information. An improved Spike neural network is constructed, which contains a multi-layer dynamic neuron structure to adapt to the frequency and dynamic change characteristics of the 5G network. Through the time-frequency joint coding method, the preliminary channel estimation data and the received signals are converted into Spike sequences to form the time-frequency dual-domain neuron firing pattern. The similarity between the dual-domain neuron firing patterns is calculated to evaluate the channel estimation error. According to the improved Spike time-dependent plasticity rule, the neuron synaptic weights are dynamically adjusted, and a multi-scale learning mechanism is adopted to adjust the neuron synaptic weights at different time and frequency scales simultaneously to converge the channel estimation error. Through repeated iteration, the optimized channel estimation data is generated.

[0208] The preliminary channel estimation data is decomposed into time-frequency joint, and the time domain channel data is converted into joint time-frequency domain data. Select an appropriate time-frequency window function, dynamically adjust the parameters according to the channel characteristics of the 5G network, and extract the frequency response of each subcarrier. The amplitude and phase of each subcarrier time-frequency domain response are extracted and recorded as the channel time-frequency domain state information.

[0209] The extracted frequency response data is used to perform parameter fitting on the channel model to extract channel characteristic parameters, including multipath delay, fading coefficient and frequency selectivity characteristics. Based on the extracted channel characteristic parameters, the channel gain is calculated, the continuous frequency domain channel estimation data is generated by performing frequency domain interpolation and fitting on the channel gain, and the time domain channel estimation data is converted by inverse Fourier transform.

[0210] According to the channel estimation result, a precoding matrix is constructed to precode the transmitted signal to generate a precoded signal. At the receiving end, an equalization matrix is constructed using the estimated channel response, and after receiving the precoded signal, the equalization matrix is used to perform equalization processing on the received signal to recover the original transmitted signal, and error detection and correction are performed to generate corrected signal data.

[0211] In the implementation process, the embodiment is tested for one week in the morning 8-10 and the afternoon 5-7 two peak periods. In the embodiment, 1000 test points are selected, the channel state information of each test point is recorded, and the comparison is made between the signal estimation method of the application and the traditional channel estimation method (minimum mean square error estimation MMSE and least square estimation LSE).

[0212] The test results show that the method of the application is significantly better than the traditional method in terms of channel estimation accuracy, signal transmission stability and data transmission speed. The comparison data of 5G network channel estimation and correction are shown in Table 1 as follows:

[0213] Table 1

[0214] Indicator Method of the present application MMSE method LSE method Channel estimation error 0.05 0.15 0.2 Data transmission speed (Mbps) 950 850 800 Signal transmission stability 99.8％ 97.5％ 95.0％

[0215] In the specific channel estimation process, the channel estimation error of the method of the application is only 0.05, while the errors of the MMSE method and the LSE method are 0.15 and 0.20 respectively. This shows that the method of the application can more accurately capture the channel characteristics when processing frequency selective channels, significantly improving the accuracy of channel estimation. In terms of data transmission speed, the method of the application reaches 950Mbps, while the MMSE method and the LSE method are 850Mbps and 800Mbps respectively. This is mainly due to the dynamic adaptive mechanism in the channel correction process of the application, which can adjust according to the real-time channel characteristics to ensure the efficiency of data transmission. In terms of signal transmission stability, the method of the application reaches 99.8%, which is significantly better than the 97.5% and 95.0% of the MMSE method and the LSE method. This shows that the method of the application can effectively cope with channel changes in complex wireless propagation environment, ensuring the stability and reliability of signal transmission.

[0216] It can be seen from the above test results that the method has obvious advantages in 5G network channel estimation and correction in high frequency band transmission and complex wireless environment, effectively solves many problems in the prior art, and improves the overall performance and user experience of the 5G network.

[0217] The application proposes a channel estimation method based on frequency selective characteristics, which eliminates noise interference through denoising, normalization and filtering processing of channel state information, enhances the accuracy of channel estimation, and obtains denoised channel state information through adaptive filtering algorithm preprocessing of the collected frequency domain and time domain channel state information, and uses a band-pass filter to extract the effective frequency components of the channel and record them as preliminary channel estimation data, effectively improving the accuracy of channel estimation, especially in high frequency band transmission and complex wireless environment.

[0218] The application introduces a dynamic adaptive channel correction mechanism, dynamically adjusts the correction strategy according to the real-time channel characteristics, constructs a precoding matrix and an equalization matrix, uses the estimated channel response for precoding and equalization processing, and uses SVD decomposition and inverse Fourier transform algorithms to ensure the transmission accuracy and stability of the signal in a variable channel environment, making the channel correction more flexible and effectively dealing with complex channel changes in the 5G network, improving the communication quality and user experience.

[0219] The application introduces an improved spike neural network model for processing and optimizing preliminary channel estimation data, dynamically adjusts the synaptic weight of neurons through a multi-layer dynamic neuron structure and a time-frequency dual-domain neuron excitation mode, combines a biologically inspired adaptive filtering system, iteratively updates the channel estimation data, uses a time-frequency joint coding method to convert the channel estimation data and the received signal into a spike sequence, calculates the similarity between the dual-domain neuron excitation modes to evaluate the channel estimation error, and converges the channel estimation error through a multi-scale learning mechanism, significantly improving the accuracy and robustness of channel estimation.

[0220] The application uses a time-frequency joint decomposition technology to convert time domain channel data into joint time-frequency domain data and extract the frequency response of each subcarrier. By selecting a suitable time-frequency window function, dynamically adjusting the parameters according to the channel characteristics of the 5G network, accurately extracting the frequency response of the subcarriers, and extracting the amplitude and phase, not only improves the accuracy of channel characteristic analysis, but also provides high-precision data support for subsequent channel model construction and channel correction.

[0221] It should be noted that the above examples are only for understanding the application and do not constitute a limitation on the channel estimation method of the application. More forms of simple transformation based on this technical concept are within the protection scope of the application.

[0222] The application also provides a channel estimation device, please refer to Figure 5 , Figure 5 Figure 1 is a schematic diagram of a module structure of a channel estimation device according to an embodiment of the application; the channel estimation device comprises:

[0223] A preprocessing module 501 is configured to preprocess channel state information in a frequency domain and a time domain of a communication network to obtain preliminary channel estimation data.

[0224] A data optimization module 502 is configured to optimize the preliminary channel estimation data by using a preset Spike neural network to generate optimized channel estimation data, wherein the Spike neural network adopts a multi-layer dynamic neuron structure.

[0225] A time-frequency information module 503 is configured to extract time-frequency domain state information of a channel based on joint time-frequency domain data converted from the optimized channel estimation data.

[0226] A result generation module 504 is configured to perform channel characteristic analysis on the time-frequency domain state information to generate a channel estimation result.

[0227] As an implementation form, the preprocessing module 501 is further configured to perform channel sampling on the communication network by using a multi-antenna system to obtain channel state information in a frequency domain and a time domain; perform denoising processing on the channel state information in the frequency domain and the time domain according to an adaptive filtering algorithm to obtain denoised channel state information; and perform filtering processing on the denoised channel state information by using a band-pass filter to obtain the preliminary channel estimation data.

[0228] As an implementation form, the data optimization module 502 is further configured to convert the preliminary channel estimation data into a time-frequency dual-domain neuron firing mode by using the preset Spike neural network; determine a channel estimation error according to a similarity between the time-frequency dual-domain neuron firing modes; perform iterative convergence on the channel estimation error based on neuron synaptic weights of the Spike neural network that are iteratively optimized until the channel estimation error reaches a preset threshold; and determine optimized channel estimation data based on the neuron synaptic weights and the time-frequency dual-domain neuron firing mode when the channel estimation error reaches the preset threshold.

[0229] As an implementation form, the time-frequency information module 503 is further configured to perform time-frequency joint decomposition on the optimized channel estimation data to obtain corresponding joint time-frequency domain data; extract frequency responses of each subcarrier based on the joint time-frequency domain data; and use the frequency responses of each subcarrier as time-frequency domain state information of a channel.

[0230] As an implementation form, the result generation module 504 is further configured to construct a channel model by taking the joint time-frequency domain data as input data; perform parameter fitting on the channel model by using the frequency response of each subcarrier to obtain channel characteristic parameters, and determine a channel gain in the channel characteristic parameters; perform frequency domain interpolation and fitting on the channel gain to generate frequency domain channel estimation data; and perform inverse Fourier transform on the frequency domain channel estimation data to obtain the channel estimation result.

[0231] As an implementation form, the channel estimation device further includes a correction module configured to construct a precoding matrix and an equalization matrix according to the channel estimation result; perform channel processing on an original transmission signal based on the precoding matrix and the equalization matrix to obtain a recovered original transmission signal; correct a frequency selective error in the recovered original transmission signal to generate corrected channel state information; and perform data transmission according to the corrected channel state information.

[0232] Other embodiments or specific implementations of the channel estimation device provided in the present application can refer to the above-mentioned method embodiments, and will not be described here.

[0233] The channel estimation device provided in the present application adopts the channel estimation method in the above-mentioned embodiments, and can solve the technical problem that the existing channel estimation method generally cannot accurately capture the dynamic change of the channel when processing the frequency selective channel, resulting in insufficient channel estimation accuracy. Compared with the prior art, the beneficial effects of the channel estimation device provided in the present application are the same as those of the channel estimation method provided in the above-mentioned embodiments, and other technical features in the channel estimation device are the same as those disclosed in the above-mentioned embodiment methods, and will not be described here.

[0234] The present application provides a channel estimation device, which comprises at least one processor and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the channel estimation method in the above-mentioned embodiment one.

[0235] The following refers to Figure 6 , Figure 6A device structure diagram of a hardware operating environment involved in the channel estimation method in the embodiments of the present application is shown, which shows a structure diagram of a channel estimation device suitable for implementing the embodiments of the present application. The channel estimation device in the embodiments of the present application can include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (Personal Digital Assistant), PADs (Portable Application Description), PMPs (Portable Media Player), vehicle-mounted terminals (for example, vehicle-mounted navigation terminals), and the like, and fixed terminals such as digital TVs, desktop computers, and the like. Figure 6 The channel estimation device shown is only an example, and should not bring any limitation to the functions and use range of the embodiments of the present application.

[0236] As shown in Figure 6 The channel estimation device can include a processing apparatus 1001 (for example, a central processor, a graphics processor, and the like) that can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage apparatus 1003 into a random access memory (RAM) 1004. In the RAM 1004, various programs and data required for the operation of the channel estimation device are also stored. The processing apparatus 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input apparatus 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, and the like; an output apparatus 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, and the like; the storage apparatus 1003 including, for example, a magnetic tape, a hard disk, and the like; and a communication apparatus 1009. The communication apparatus 1009 can allow the channel estimation device to perform wireless or wired communication with other devices to exchange data. Although the channel estimation device with various systems is shown in the figure, it should be understood that all the systems shown are not required to be implemented or provided. More or fewer systems can be alternatively implemented or provided.

[0237] The channel estimation device provided by the present application adopts the channel estimation method in the above embodiment, and can solve the technical problem that the existing channel estimation method is generally difficult to accurately capture the dynamic change of the channel when processing the frequency selective channel, resulting in insufficient channel estimation accuracy. Compared with the prior art, the beneficial effects of the channel estimation device provided by the present application are the same as those of the channel estimation method provided by the above embodiment, and other technical features in the channel estimation device are the same as those disclosed in the previous embodiment method, which will not be repeated here.

[0238] The present application provides a computer readable storage medium having computer readable program instructions (i.e. computer programs) stored thereon for executing the channel estimation method in the above embodiment.

[0239] The computer readable storage medium provided by the present application may, for example, be a U disk, but is not limited to an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, system or device, or any combination of the above.

[0240] The readable storage medium provided by the present application is a computer readable storage medium which stores computer readable program instructions (i.e. computer programs) for executing the above channel estimation method. Compared with the prior art, the beneficial effects of the computer readable storage medium provided by the present application are the same as those of the channel estimation method provided by the above embodiment, which will not be repeated here.

[0241] The present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the channel estimation method as described above.

[0242] The computer program product provided by the present application can solve the technical problem that the existing channel estimation method is generally difficult to accurately capture the dynamic change of the channel when processing the frequency selective channel, resulting in insufficient channel estimation accuracy. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as those of the channel estimation method provided by the above embodiment, which will not be repeated here.

[0243] The above only describes some embodiments of the present application, and does not limit the patent scope of the present application, and any equivalent structural transformation made by using the contents of the present application specification and drawings, or direct / indirect application in other related technical fields is included in the patent protection scope of the present application.

Claims

1. A channel estimation method, characterized in that, The method includes: Preprocessing of the channel state information in the frequency and time domains of the communication network yields preliminary channel estimation data; The preliminary channel estimation data is optimized by a preset Spike neural network to generate optimized channel estimation data. The Spike neural network adopts a multi-layer dynamic neuron structure. Based on the joint time-frequency domain data converted from the optimized channel estimation data, the time-frequency domain state information of the channel is extracted; Channel characteristic analysis is performed on the time-frequency domain state information to generate channel estimation results; The step of optimizing the preliminary channel estimation data using a preset Spike neural network to generate optimized channel estimation data includes: The preliminary channel estimation data is converted into a time-frequency dual-domain neuron excitation mode using a pre-defined Spike neural network. The channel estimation error is determined based on the similarity between the excitation patterns of the time-frequency dual-domain neurons. Based on the neuron synaptic weights of the Spike neural network iteratively optimized, the channel estimation error is iteratively converged until the channel estimation error reaches a preset threshold. When the channel estimation error reaches the preset threshold, optimized channel estimation data is determined based on the neuron synaptic weights and the time-frequency dual-domain neuron firing modes.

2. The method as described in claim 1, characterized in that, The step of extracting the time-frequency domain state information of the channel based on the joint time-frequency domain data converted from the optimized channel estimation data includes: The optimized channel estimation data is subjected to joint time-frequency decomposition to obtain the corresponding joint time-frequency domain data; Based on the joint time-frequency domain data, the frequency response of each subcarrier is extracted; The frequency response of each subcarrier is used as the time-frequency domain state information of the channel.

3. The method as described in claim 2, characterized in that, The step of performing channel characteristic analysis on the time-frequency domain state information and generating channel estimation results includes: The joint time-frequency domain data is used as input data to construct a channel model; The channel model is fitted with parameters by the frequency response of each subcarrier to obtain channel characteristic parameters, and the channel gain in the channel characteristic parameters is determined. Frequency domain interpolation and fitting are performed on the channel gain to generate frequency domain channel estimation data; The frequency domain channel estimation data is subjected to inverse Fourier transform to obtain the channel estimation result.

4. The method according to any one of claims 1 to 3, characterized in that, The step of preprocessing the channel state information in the frequency and time domains of the communication network to obtain preliminary channel estimation data includes: Channel sampling of the communication network is performed using a multi-antenna system to obtain channel state information in the frequency and time domains; The channel state information in the frequency domain and time domain is denoised using an adaptive filtering algorithm to obtain denoised channel state information. The denoised channel state information is filtered by a bandpass filter to obtain preliminary channel estimation data.

5. The method according to any one of claims 1 to 3, characterized in that, After the step of performing channel characteristic analysis on the time-frequency domain state information and generating channel estimation results, the method further includes: Based on the channel estimation results, construct the precoding matrix and the equalization matrix; Based on the precoding matrix and the equalization matrix, channel processing is performed on the original transmitted signal to obtain the recovered original transmitted signal; The frequency selectivity error in the recovered original transmitted signal is corrected to generate corrected channel state information; Data transmission is performed based on the corrected channel state information.

6. A channel estimation device, characterized in that, The apparatus performs the channel estimation method as described in claim 1, the apparatus comprising: The preprocessing module is used to preprocess the channel state information in the frequency domain and time domain of the communication network to obtain preliminary channel estimation data. The data optimization module is used to optimize the preliminary channel estimation data through a preset Spike neural network to generate optimized channel estimation data. The Spike neural network adopts a multi-layer dynamic neuron structure. The time-frequency information module is used to extract the time-frequency domain state information of the channel based on the joint time-frequency domain data converted from the optimized channel estimation data; The result generation module is used to perform channel characteristic analysis on the time-frequency domain state information and generate channel estimation results.

7. A channel estimation device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the channel estimation method as described in any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the channel estimation method as described in any one of claims 1 to 5.

9. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the channel estimation method as described in any one of claims 1 to 5.

Images