Method and system for detecting active state of base station access equipment

Abstract

The invention discloses an active state detection method and system for base station access equipment, and the method comprises the steps: constructing and training a preamble detection neural network, and extracting first channel state information from a preamble set and covariance information of a received signal; constructing and training a data detection neural network, and extracting second channel state information from a data signal and a lead code set in the received signal and the first channel state information; the spatial expansion unit converts the second channel state information into an active state estimation result based on a preset conversion threshold value; and the base station summarizes the results of all the spatial expansion units and outputs a final activity state detection result. According to the invention, the accuracy of channel estimation and active state estimation is improved; according to the method, deep learning and covariance detection are combined to overcome the defects of a traditional compressed sensing method, short non-orthogonal preambles are better utilized, the detection performance under short preamble sequence application is improved, and activity detection and channel estimation are carried out under the condition that no complex scheduling process exists.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication networks, and in particular to a method and system for detecting the activity state of a base station access device. Background Art

[0002] With the rapid development of power Internet of Things and wireless communications, the power Internet of Things needs to support the simultaneous access and communication of massive devices, which usually carry very small data packets and require extreme coverage. In the unauthorized communication mode, the device can directly transmit the data signal after the preamble signal without the authorization of the base station, which leads to high collision risk and unpredictable number of potential active users. Traditional compressed sensing methods have performance limitations in the joint detection of user activity and channel status, especially when the number of users exceeds the length of the preamble sequence, it cannot meet the access requirements of a large number of devices, and the performance is severely degraded when the coverage area is large. In addition, the detection performance of the compressed sensing method is significantly reduced when processing short preamble sequences.

[0003] The "data transmission method, device, electronic device and storage medium" disclosed in the Chinese patent literature has a publication number of CN115333709A and a publication date of 2022-11-11. The method includes: determining a first compressed sensing model and a second compressed sensing model based on a target signal sent by at least one active terminal; the target signal includes a preamble and valid data; determining each active terminal and the channel state information corresponding to each active terminal based on the first compressed sensing model; determining the valid data corresponding to each active terminal based on each active terminal, the channel state information corresponding to each active terminal and the second compressed sensing model. This technology uses the target signal sent by the active terminal to include a preamble and valid data, realizes one-step synchronous transmission of the preamble and valid data, and determines the valid data transmitted by each active terminal based on the first compressed sensing model and the second compressed sensing model, and completes the transmission of valid data while completing the transmission of the preamble, so that the transmission delay and signaling overhead are greatly reduced. However, it does not solve the performance limitations of the compressed sensing method in the joint detection of user activity and channel status, cannot meet the access requirements of a large number of devices, and has serious performance degradation when the coverage area is large. Summary of the invention

[0004] The present invention aims to overcome the performance limitations of the prior art when using compressed sensing methods to detect the activity status of access devices, which cannot meet the access needs of a large number of devices and has serious performance degradation when the coverage area is large. A method and system for detecting the activity status of base station access devices are provided.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A method for detecting an activity state of a base station access device, comprising: Constructing and training a preamble detection neural network to extract first channel state information from a preamble set and covariance information of a received signal; Constructing and training a data detection neural network to extract second channel state information from a data signal and a preamble set in a received signal and the first channel state information; The spatial expansion unit converts the second channel state information into an activity state estimation result based on a preset conversion threshold; The base station aggregates the activity state estimation results output by all spatial extension units and outputs a final activity state detection result.

[0006] In the present invention, the spatial expansion unit receives the signal in each area and calculates the sample covariance of the received signal; the processed signal covariance information and the preamble code set are used as input and sent to the preamble code detection neural network (PDNN) to extract the first channel state information; the output first channel state information, the data segment signal of the received signal and the preamble code set are used as the input of the detection neural network (DDMM), and the detection neural network outputs more accurate second channel state information; the second channel state information is processed using a preset conversion threshold and the activity state estimation result is output; finally, the spatial expansion unit transmits the trained model parameters to the base station, and the base station is integrated into the global model to obtain the final global prediction detection result; the present invention improves the accuracy of channel estimation and activity state estimation when the device adopts a short preamble sequence; by combining deep learning and covariance detection to better solve the problems of traditional compressed sensing methods, better utilize short non-orthogonal preamble codes, improve the detection performance under the application of short preamble sequences, and perform user activity detection and channel estimation without a complex scheduling process.

[0007] Preferably, converting the second channel state information into an activity state estimation result based on a preset conversion threshold comprises: The square of the modulus value of the second channel state information is compared with the conversion threshold. If the modulus value is greater than the conversion threshold, the activity state estimation result is an active state, otherwise it is an inactive state.

[0008] Preferably, the preamble detection neural network includes T preamble neural network layers, each preamble neural network layer receives the channel estimation result of the previous preamble neural network layer and the preamble set and covariance information of the received signal for estimation and update, and outputs a new channel estimation result.

[0009] Preferably, each preamble neural network layer in the preamble detection neural network includes a plurality of first hidden layers; performing iterative calculation: The output of each first hidden layer is the output of the previous first hidden layer multiplied by the weight of the current layer, plus the bias vector of the current layer, and then multiplied by the activation function of the current layer; in the first first hidden layer, the input of the corresponding preamble neural network layer is used as the output of the previous first hidden layer.

[0010] Preferably, when training the preamble detection neural network, the square of the Euclidean norm is calculated based on the difference between the channel estimation result and the true result of each preamble neural network layer, and then the weighted sum is performed as the first loss function; the square of the Euclidean norm is calculated based on the difference between the channel estimation result and the true result of the preamble detection neural network, and the sum is used as the second loss function; First, train to reduce the value of the first loss function until convergence, and then train to minimize the second loss function.

[0011] Preferably, the data detection neural network includes a plurality of second hidden layers; A layer number threshold is preset. For the second hidden layer whose number of layers is less than the layer number threshold, its input is the output of the previous second hidden layer and the input of the data detection neural network; for the second hidden layer whose number of layers is greater than or equal to the layer number threshold, its input is the output of the previous second hidden layer.

[0012] Preferably, the data processing process of the second hidden layer includes: for any second hidden layer, its output is, after the input of the second hidden layer is multiplied by the corresponding weight and added with the corresponding bias vector, multiplied by the corresponding activation function.

[0013] Preferably, when training the data detection neural network, the training is performed by minimizing the value of the third loss function; The third loss function is to calculate the square of the Euclidean norm of the difference between the channel estimation result of the data detection neural network and the true result and sum them up.

[0014] Preferably, a global detection model is provided in the base station, and after averaging the activity state estimation results from the spatial expansion units, the weights and bias parameters of the fully connected layer are processed, and finally a batch normalization operation is performed to obtain the final activity state detection result.

[0015] An activity status detection system for a base station access device includes a base station and a plurality of space expansion units within the base station range; The spatial expansion unit is provided with a preamble detection neural network, a data detection neural network and a conversion threshold, and after receiving a signal within its own coverage area, outputs an estimation result of the device activity state in the corresponding area; The base station is provided with a global detection model, which aggregates the activity state estimation results output by all the spatial expansion units and outputs a final activity state detection result.

[0016] The present invention has the following beneficial effects: the base station efficiently utilizes the covariance structure of the signal to detect the activity state through a neural network. The detection using the covariance structure of the signal overcomes the problems of compressed sensing technology that cannot meet the access requirements of a large number of devices and serious performance degradation when the coverage area is large, and improves the performance in small data packet transmission scenarios. At the same time, a deep learning neural network is applied, and the trained deep learning model not only performs well under the channel conditions during training, but also has good generalization ability under different channel statistical conditions; after offline training, it can be deployed in real time at the base station site to extract useful features of the received signal, and is suitable for application in various practical environments; a method combining covariance activity detection and deep learning neural network is adopted to improve the detection efficiency and accuracy of activity state detection in the context of large-scale access, and the base station can maximize the efficiency of activity state and channel estimation when a large number of devices are connected. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the activity status detection method of the base station access device in the present invention.

[0018] Figure 2 Schematic diagram of the activity detection channel estimation neural network in the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0020] Traditional compressed sensing methods have performance limitations in the joint detection of user activities and channel states, especially when the number of users exceeds the length of the preamble sequence. In addition, the performance of the compressed sensing method is significantly reduced when processing short preamble sequences. In order to solve these problems, the present invention proposes a covariance detection method, which has the advantage that it can use the covariance structure of the signal to detect user activities. This method does not rely on orthogonal pilot sequences, so it can work under non-orthogonal pilot sequences, allowing more users to access the network at the same time. In addition, the method improves the detection performance by utilizing the macro-diversity gain provided by a large-scale antenna array, especially when the coverage area is large, it has better performance compared to the co-located deployment architecture. At the same time, in response to the defects of the compressed sensing algorithm, an activity detection channel estimation neural network UAD-CE-NN is designed to improve the detection accuracy.

[0021] like Figure 1 As shown, a method for detecting the activity state of a base station access device includes: Constructing and training a preamble detection neural network to extract first channel state information from a preamble set and covariance information of a received signal; Constructing and training a data detection neural network to extract second channel state information from a data signal and a preamble set in a received signal and the first channel state information; The spatial expansion unit converts the second channel state information into an activity state estimation result based on a preset conversion threshold; The base station aggregates the activity state estimation results output by all spatial extension units and outputs a final activity state detection result.

[0022] It should be noted that in the present invention, the spatial expansion unit receives the signal in each area and calculates the sample covariance of the received signal; the processed signal covariance information and the preamble code set are used as input and sent to the preamble code detection neural network (PDNN) to extract the first channel state information; the output first channel state information, the data segment signal of the received signal and the preamble code set are used as inputs of the detection neural network (DDMM), and the detection neural network outputs more accurate second channel state information; the second channel state information is processed using a preset conversion threshold and the activity state estimation result is output; finally, the spatial expansion unit transmits the trained model parameters to the base station, and the base station is integrated into the global model to obtain the final global prediction detection result; the present invention improves the accuracy of channel estimation and activity state estimation when the device adopts a short preamble sequence; by combining deep learning and covariance detection to better solve the problems of traditional compressed sensing methods, better utilize short non-orthogonal preamble codes, improve the detection performance under the application of short preamble sequences, and perform user activity detection and channel estimation without a complex scheduling process.

[0023] It should be noted that the base station uses the covariance structure of the signal to detect the activity state through a neural network. Detection based on the covariance structure of the signal overcomes the problems of compressed sensing technology that cannot meet the access needs of a large number of devices and has serious performance degradation when the coverage area is large, and improves the performance in small data packet transmission scenarios. At the same time, the deep learning neural network is applied. The trained deep learning model not only performs well under the channel conditions during training, but also has good generalization ability under different channel statistical conditions. After offline training, the activity detection channel estimation neural network UAD-CE-NN can be deployed in real time at the base station site to extract useful features of the received signal, which is suitable for application in various practical environments. The method of combining covariance activity detection with deep learning neural network is adopted to improve the detection efficiency and accuracy of activity state detection in the context of large-scale access. The base station can maximize the efficiency of activity state and channel estimation when a large number of devices are connected.

[0024] As a specific embodiment, converting the second channel state information into an activity state estimation result based on a preset conversion threshold includes: The square of the modulus value of the second channel state information is compared with the conversion threshold. If the modulus value is greater than the conversion threshold, the activity state estimation result is an active state, otherwise it is an inactive state.

[0025] It should be noted that the activity detection channel estimation neural network UAD-CE-NN of the present invention mainly includes a preamble code detection neural network and a data detection neural network located in the spatial extension unit, and a global detection model located in the base station. According to the real-time received signal, the corresponding second channel state information is output through the preamble code detection neural network and the data detection neural network, and then the second channel state information is converted into an activity state estimation result and transmitted to the base station. The results of each spatial extension unit are fused through the global detection model to obtain the final activity state estimation result.

[0026] Specifically, when converting the second channel state information to the activity state estimation result, an appropriate conversion threshold is selected as an offset, and estimation is performed using a shrinkage function with an offset:

[0027] in Estimates of activity states for data detection neural networks, is the conversion threshold. Specifically:

[0028] in Represents the square of its modulus value. When the result is 1, the device is in an active state, and when the result is 0, the device is in an inactive state.

[0029] As a specific example, Figure 2 As shown, the preamble detection neural network includes T preamble neural network layers, each preamble neural network layer receives the channel estimation result of the previous preamble neural network layer and the preamble set and covariance information of the received signal for estimation and update, and outputs a new channel estimation result.

[0030] It should be noted that the preamble detection neural network PDNN is trained in the spatial extension unit SEU to generate training data, which includes the preamble of the signal received in the corresponding area of ​​the spatial extension unit and the covariance information of the signal. The neural network model is optimized and trained according to the defined loss function so that the preamble detection neural network PDNN can learn how to extract the first channel state information from the preamble set and covariance information. The entire preamble detection neural network PDNN is divided into T layers and arranged in sequence. Each preamble detection neural network PDNN layer is actually a specific neural network, which uses the received preamble set and the output of the previous PDNN layer to obtain an updated estimate of the channel. Finally, the result of the output of the last PDNN layer can be obtained as the first channel state information.

[0031] Further, each preamble neural network layer in the preamble detection neural network includes a plurality of first hidden layers; Perform iterative calculations: the output of each first hidden layer is the output of the previous first hidden layer multiplied by the weight of the current layer, plus the bias vector of the current layer, and then multiplied by the activation function of the current layer; in the first first hidden layer, the input of the corresponding preamble neural network layer is used as the output of the previous first hidden layer.

[0032] It should be noted that in a neural network, each input feature is multiplied by an element in the weight vector, and then these products are summed. The role of the weight vector is to determine the linear combination of the input signals when passing through the layer. By adjusting the weights, the neural network can learn the characteristics and patterns of the input data. The bias vector is a constant vector that is added to the linear combination of the weights and the input signal. The role of the bias is to provide an offset for the output of the neuron, so that the output of the neuron does not only depend on the strength of the input signal, but can also be adjusted independently of the input signal to a certain extent, so that it can better fit the data. For active users, their channel state information can be observed in the received signal, while the channel information of inactive users is not available.

[0033] Furthermore, when training the preamble detection neural network, the square of the Euclidean norm is calculated based on the difference between the channel estimation result and the true result of each preamble neural network layer, and then the weighted sum is performed as the first loss function; the square of the Euclidean norm is calculated based on the difference between the channel estimation result and the true result of the preamble detection neural network, and the sum is used as the second loss function; First, train to reduce the value of the first loss function until convergence, and then train to minimize the second loss function.

[0034] It should be noted that the training process of first reducing the first loss function to convergence and then minimizing the second loss function is to improve the estimation accuracy and convergence speed of the neural network; the first loss function evaluates the overall accuracy of the channel estimation of all preamble neural network layers PDNN, and by continuously reducing the result of the first loss function, all neural network layers in the PDNN preamble neural network are collectively trained; when the first training step converges, the second training step aims to fine-tune the channel estimation by minimizing the second loss function.

[0035] As a specific example, Figure 2 As shown, the data detection neural network includes several second hidden layers; A layer number threshold is preset. For the second hidden layer whose number of layers is less than the layer number threshold, its input is the output of the previous second hidden layer and the input of the data detection neural network; for the second hidden layer whose number of layers is greater than or equal to the layer number threshold, its input is the output of the previous second hidden layer.

[0036] It should be noted that the data detection neural network DDNN is constructed and trained after the preamble detection neural network PDNN is set up. The preamble set and data signal of the received signal and the first channel state information output by the preamble detection neural network are used as inputs to train it to learn to extract more accurate second channel state information. For the data detection neural network, a specific layer number threshold is set based on its several second hidden layers to distinguish between two different hidden layer data processing methods, and the selection of input data is expanded in the first layer number threshold second hidden layer to ensure data integrity in the training process and the detection process. In the subsequent second hidden layer, the focus is on training the output results of the previous layer, which can improve the convergence speed.

[0037] Furthermore, the data processing process of the second hidden layer includes: for any second hidden layer, its output is, after the input of the second hidden layer is multiplied by the corresponding weight and added with the corresponding bias vector, it is multiplied by the corresponding activation function.

[0038] It should be noted that the data processing method in the second hidden layer of the data detection neural network is the same as the data processing method in the first hidden layer of the preamble detection neural network, both of which are iterative calculations of several hidden layers arranged in sequence, the difference is that the weights, bias vectors, activation functions and input data calculated in each hidden layer are different. After continuous iterative calculations, the output result of the last second hidden layer is used as the second channel state information output by the data detection neural network.

[0039] Furthermore, when training the data detection neural network, the training is performed by minimizing the value of the third loss function; The third loss function is to calculate the square of the Euclidean norm of the difference between the channel estimation result of the data detection neural network and the true result and sum them up.

[0040] It should be noted that the training of the data detection neural network DDNN is similar to the training of the preamble detection neural network PDNN. The overall accuracy of the channel estimation of the data detection neural network is evaluated by the third loss function, and all neural networks in the data detection neural network DDNN are trained by continuously reducing the result of the third loss function. Because the data detection neural network only includes several second hidden layers, which is different from the preamble neural network including T preamble neural network layers and each layer including several first hidden layers, only one third loss function is needed to train the neural network.

[0041] As a specific embodiment, a global detection model is provided in the base station, which averages the activity state estimation results from the spatial expansion units, processes the weights and bias parameters of the fully connected layer, and finally performs a batch normalization operation to obtain the final activity state detection result.

[0042] It should be noted that the weights and bias parameters of the fully connected layer are trainable parameters in the global detection model. After the preamble neural network PDNN and the data detection neural network DDNN are trained, the second channel state information output by the data detection neural network is used as a training set and input into the global detection model of the base station for training. Finally, the weights and bias parameters that meet the requirements are obtained, and the model configuration in the base station is completed.

[0043] In addition to the activity state detection method of a base station access device, the present invention also discloses an activity state detection system of a base station access device, including a base station and a plurality of space expansion units within the range of the base station; The spatial expansion unit is provided with a preamble detection neural network, a data detection neural network and a conversion threshold, and after receiving the signal within its own coverage area, it outputs the estimated result of the device activity status in the corresponding area; The base station is provided with a global detection model, which aggregates the activity state estimation results output by all spatial extension units and outputs the final activity state detection result.

[0044] It should be noted that the present invention considers configuring several spatial extension units around the base station. The range of access devices covered by the base station includes the sum of the ranges covered by each spatial extension unit. Therefore, the specific connection state is that the base station is connected to each spatial extension unit, and each spatial extension unit is connected to the access devices within its coverage. The spatial extension unit estimates the local device activity state within its coverage, and then each spatial extension unit transmits its own estimation results to the aggregated base station, which processes the data to obtain the final activity state detection result. The method of combining covariance activity detection with deep learning neural networks is adopted to improve the detection efficiency and accuracy of activity state detection in the context of large-scale access; at the same time, the base station can maximize the efficiency of activity state and channel estimation when a large number of devices are connected.

[0045] The following is an example of an industrial park to illustrate the activity status detection method of the base station access device of the present invention. A large number of power Internet of Things devices are deployed in the park, such as smart meters, environmental monitoring sensors, industrial control systems, etc. These devices need to frequently transmit and communicate data across multiple networks. At the same time, the equipment and systems in the park need to perform user activity detection and channel estimation without a complex scheduling process to ensure the efficiency and reliability of communication. The specific parameters in the industrial park are as follows:

[0046] In this embodiment, the activity state detection includes the following steps.

[0047] Step 1: The base station assigns a pilot sequence to each device:

[0048] in represents the pilot sequence of the kth user, represents pi, represents an exponential expression with the natural constant e as the base, Represents the length of the pilot sequence, j represents the imaginary unit, and c is a constant used to control the phase interval in the pilot sequence. The pilot sequence is used as a known reference signal to help the receiving end estimate the characteristics of the wireless channel, including channel gain, phase and other information. The PDNN model is trained in the spatial extension unit SEU to generate training data, which includes the preamble set of the regional received signal and the covariance information of the signal. The deep neural network model is optimized and trained according to the defined loss function, so that the preamble detection neural network PDNN learns how to extract the first channel state information from the preamble set and covariance information.

[0049] The preamble code set is the set of preamble code signals of the received signal; the covariance information of the received signal It can be expressed as:

[0050] Where Y is the received signal, represents the conjugate transpose operation of a vector or matrix, and K is the number of antennas equipped by the base station.

[0051] The entire preamble detection neural network PDNN is divided into T Layers are arranged in order. Each PDNN layer is actually a specific neural network that uses the received leading sequence and the output of the previous PDNN layer to obtain an updated estimate of the channel. Finally, the final estimate of the output of the last PDNN layer can be obtained.

[0052] To clearly illustrate the processing of PDNN, we take the tth layer in PDNN as an example. To ensure that PDNN is free from overfitting, gradient vanishing, or falling into the local optimum, the input of the t+1th PDNN layer is set to

[0053] in represents the estimated channel state information of the kth antenna in the tth PDNN layer, where K is the number of antennas. The specific input is the vector of covariance information, the estimated channel state information output by the previous PDNN layer, and the product of the preamble set S and the estimated channel state information output by the previous PDNN layer. As a special case, the first PDNN layer does not have the previous PDNN layer. ,in represents the complex conjugate transpose, Represents converting a matrix into a column vector. Represents the channel estimation result of the kth antenna in the t-1th PDNN layer.

[0054] The calculation process in the t-th PDNN layer is expressed as:

[0055] in and are the The weights and bias vectors in the first hidden layer are The first hidden layer; Represents the activation function of the first hidden layer.

[0056] The weight vector represents the strength of the connection between the input signal and the neurons in this layer. In a neural network, each input feature is multiplied by an element in the weight vector, and then these products are summed. The role of the weight vector is to determine the linear combination of the input signals when passing through the layer. By adjusting the weights, the network can learn the characteristics and patterns of the input data. The bias vector is a constant vector that is added to the linear combination of the weights and the input signal. The role of the bias is to provide an offset for the output of the neuron, so that the output of the neuron does not only depend on the strength of the input signal, but can also be adjusted independently of the input signal to a certain extent, so that it can better fit the data. The activation function is used to process the output of each layer, so that the network can extract useful features from the received signal and perform nonlinear transformations. By selecting a suitable activation function, PDNN can perform user activity detection and channel estimation more effectively.

[0057] Thus, the channel estimation result of the kth antenna in the tth first hidden layer is obtained. It is essential to select a suitable loss function to configure the PDNN model. In order to improve the estimation accuracy of the neural network and increase the convergence speed, the training of the PDNN is divided into two steps: In the first step, the first loss function is defined as the difference between the channel estimation result and the true result of each preamble neural network layer, and the square of the Euclidean norm is calculated and then weighted summed, which is expressed as

[0058] in is the weight of the mean square error (MSE) of the t-th PDNN layer, Represents the square of the Euclidean norm, which is the sum of the squares of the elements of a vector. represents the channel estimation result of the kth antenna at the tth PDNN layer, represents the real channel state of the kth antenna, is the number of antennas of the base station. The overall accuracy of the channel estimation of all PDNN units is evaluated according to the first loss function. By continuously reducing the first loss function, all neural networks of the PDNN are collectively trained. When the first training step converges, the second training step aims to fine-tune the channel estimation by minimizing the second loss function:

[0059] in Represents the channel estimation result of the kth antenna in the preamble code detection neural network.

[0060] After training is completed, the covariance information of the received signal is input in actual application and preamble set Finally, the first channel state information is outputted by the preamble detection neural network. .

[0061] Step 2: The data detection neural network DDNN is trained after setting up the PDNN. The DDNN uses the received signal and the first channel estimation information output by the PDNN as the training set. The training data detection neural network DDNN learns to extract channel state information. The input of the data detection neural network is defined as:

[0062] in It is the signal of the mth data segment received by the kth antenna (the signal has a total of M data segments), and the square brackets indicate the stacking of all M data segment signals.

[0063] In DDNN, there are several second hidden layers. In the lth second hidden layer, the input is defined as

[0064] Among them is The output of the l-1th second hidden layer, It is a specific hidden layer index in DDNN, corresponding to the layer number threshold, which is used to distinguish between two different data processing methods.

[0065] The output of the lth second hidden layer is calculated as follows:

[0066] in represents the weight matrix of the lth second hidden layer, is the activation function used by the lth second hidden layer, is the bias vector of the lth second hidden layer. The bias is a parameter learned in the network that controls the output of the activation function. Then, and are all regarded as the input of the next hidden layer until the last second hidden layer of DDNN. That is, the output of the last second hidden layer, as the channel estimation result .

[0067] For the training of DDNN model, similar to PDNN, the third loss function of DDNN is defined as:

[0068] The overall accuracy of the channel estimation of DDNN is evaluated according to the third loss function. By continuously reducing the value of the third loss function, all neural networks of DDNN are collectively trained.

[0069] After training, similar to PDNN, the data processing process in the data detection neural network DDNN can be represented by a mapping function:

[0070] in is the data processing process in DDNN, m is the index of the mth antenna receiving signal in the base station, Y={ } is the set of received signals of the preamble segment and M data segments, including the preamble code set in addition to the received signal The first channel state information output by PDNN is used as input. In DDNN, the information of the data segment can be used together with the preamble set to improve the output result of PDNN. Finally, a more accurate second channel state information is output. .

[0071] Step 3: According to the second channel state information obtained by DDNN, select an appropriate conversion threshold to convert the channel state output result into the device activity result. The activity state is obtained by the following shrinkage function with offset estimate:

[0072] in Estimates of activity states for data detection neural networks, is the conversion threshold. Specifically:

[0073] in Represents the square of its modulus value. When the result is 1, the device is in an active state, and when the result is 0, the device is in an inactive state.

[0074] The training phase proposed by PDNN and DDNN is performed offline. After the offline training converges, the entire neural network can be deployed in the spatial extension unit SEU to extract useful features of the received signal.

[0075] Step 4: After the local training of all spatial extension units (SEUs) is completed, the output model parameters are transmitted to the base station, and the base station performs global model training:

[0076] in is the activity state detection result output by the global model, is the activity state estimation result output by the spatial extension unit SEU, BN represents the batch normalization layer, Average represents the average operation of the local estimation results of all SEUs, W and b represent the weight and bias parameters of the fully connected layer, which are the trainable parameters of the global model on the base station. After the training is completed, the global model collects the output results of all spatial extension units SEU and outputs the final activity state detection result.

[0077] The method of the present invention reduces the error rate of activity state detection and improves the accuracy of activity state detection. The following is the detection error probability of different detection algorithms when the signal-to-noise ratio is 20dB (the activity state is detected as the inactivity state):

[0078] And the detection error probability of different detection algorithms when the signal-to-noise ratio is 10dB:

[0079] By comparison, it is found that the error probabilities of the traditional approximate message passing method and the orthogonal matching pursuit method under different signal-to-noise ratio environments are higher than those of the neural network detection method of the present invention.

[0080] The above embodiments are further elaborations and illustrations of the present invention for ease of understanding, and are not limitations of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting the activity status of a base station access device, characterized in that: include: Constructing and training a preamble detection neural network to extract first channel state information from a preamble set and covariance information of a received signal; Constructing and training a data detection neural network to extract second channel state information from a data signal and a preamble set in a received signal and the first channel state information; The spatial expansion unit converts the second channel state information into an activity state estimation result based on a preset conversion threshold; The base station aggregates the activity state estimation results output by all spatial extension units and outputs a final activity state detection result.

2. The method for detecting the activity status of a base station access device according to claim 1, characterized in that: The converting the second channel state information into an activity state estimation result based on a preset conversion threshold comprises: The square of the modulus value of the second channel state information is compared with the conversion threshold. If the modulus value is greater than the conversion threshold, the activity state estimation result is an active state, otherwise it is an inactive state.

3. A method for detecting the activity status of a base station access device according to claim 1 or 2, characterized in that: The preamble detection neural network includes T preamble neural network layers, each preamble neural network layer receives the channel estimation result of the previous preamble neural network layer and the preamble set and covariance information of the received signal for estimation and update, and outputs a new channel estimation result.

4. The method for detecting the activity status of a base station access device according to claim 3, characterized in that: Each preamble neural network layer in the preamble detection neural network includes a plurality of first hidden layers; performing iterative calculation: The output of each first hidden layer is the output of the previous first hidden layer multiplied by the weight of the current layer, plus the bias vector of the current layer, and then multiplied by the activation function of the current layer; in the first first hidden layer, the input of the corresponding preamble neural network layer is used as the output of the previous first hidden layer.

5. The method for detecting the activity status of a base station access device according to claim 3, characterized in that: When training the preamble detection neural network, the square of the Euclidean norm is calculated based on the difference between the channel estimation result and the true result of each preamble neural network layer, and then the weighted sum is performed as the first loss function; the square of the Euclidean norm is calculated based on the difference between the channel estimation result and the true result of the preamble detection neural network, and the sum is used as the second loss function; First, train to reduce the value of the first loss function until convergence, and then train to minimize the second loss function.

6. A method for detecting the activity status of a base station access device according to claim 1 or 2, characterized in that: The data detection neural network includes a plurality of second hidden layers; A layer number threshold is preset. For the second hidden layer whose number of layers is less than the layer number threshold, its input is the output of the previous second hidden layer and the input of the data detection neural network; for the second hidden layer whose number of layers is greater than or equal to the layer number threshold, its input is the output of the previous second hidden layer.

7. A method for detecting the activity status of a base station access device according to claim 6, characterized in that: The data processing process of the second hidden layer includes: for any second hidden layer, its output is, after the input of the second hidden layer is multiplied by the corresponding weight and added with the corresponding bias vector, it is multiplied by the corresponding activation function.

8. The method for detecting the activity status of a base station access device according to claim 6, characterized in that: When training the data detection neural network, the training is performed by minimizing the value of the third loss function; The third loss function is to calculate the square of the Euclidean norm of the difference between the channel estimation result of the data detection neural network and the true result and sum them up.

9. A method for detecting the activity status of a base station access device according to claim 1 or 2 or 4 or 5 or 7 or 8, characterized in that: The base station is provided with a global detection model, which averages the activity state estimation results from the spatial extension units, processes the weights and bias parameters of the fully connected layer, and finally performs a batch normalization operation to obtain the final activity state detection result.

10. An activity status detection system for a base station access device, applicable to the activity status detection method according to any one of claims 1 to 9, characterized in that: It includes a base station and several spatial extension units within the base station; The spatial expansion unit is provided with a preamble detection neural network, a data detection neural network and a conversion threshold, and after receiving a signal within its own coverage area, outputs an estimation result of the device activity state in the corresponding area; The base station is provided with a global detection model, which aggregates the activity state estimation results output by all the spatial expansion units and outputs a final activity state detection result.

Images