A random signal detection method based on FNN

Abstract

The present invention discloses a random signal detection method based on an FNN. The method includes obtaining frequency domain data of a random access signal; determining a random access sequence length based on system configuration information; constructing a parameter configuration of a sequence generation information judgment model for a random access signal based on the FNN based on the obtained access sequence length; importing the obtained frequency domain data of the random access signal into the FNN based random access signal sequence generation information judgment model to obtain the obtained sequence generation information of the random access signal; and constructing a parameter configuration of a timing advance judgment model for a random access signal based on the FNN based on the obtained access sequence length; importing the obtained frequency domain data of the random access signal into the FNN based random access signal timing advance judgment model to obtain the timing advance of the obtained random access signal. The present invention can more quickly obtain relevant information about a received random access signal and reduce signal processing complexity.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a random signal detection method based on FNN. Background Art

[0002] With the continuous development and advancement of technology, 4G has gradually evolved into 5G (5th Generation New Radio), and in the future, will develop into the even more advanced 6G technology. Currently, the application of 5G technology is not very widespread. While some mobile phones can utilize 5G technology to achieve higher download and upload speeds, the uses of 5G technology extend far beyond this. In the future, 5G technology will also be applied to autonomous vehicles and the Internet of Things. The use of 5G technology will require higher computing speeds and more accurate results. Since the European Union announced in 2013 that it would accelerate the development of 5G technology, 3GPP (3rd Generation Partnership Project) began research on 5G NR in September 2015. In April 2016, 3GPP officially launched 5G NR standardization work.

[0003] With the development of fifth-generation mobile communication technology, the International Telecommunication Union-Radio Communications Sector (ITU-R) has identified three major application scenarios for 5G: enhanced mobile broadband (eMBB), ultra-reliable and low-latency communication (URLLC), and massive machine-type communication (mMTC). These three application scenarios also place higher requirements on access latency for user equipment (UE).

[0004] In the 5G communication system, the base station (gNodeB) obtains the access information of the UE through random access, thereby obtaining the corresponding identification feature (ID) of the UE. Physical Random Access Channel (PRACH) detection is the first step of random access. In PRACH detection, the base station estimates the transmission delay between the UE and the base station by receiving the preamble in the signal, and can determine whether the UE has issued a random access request, and determine which UE has issued the random access request, and establish contact with it.

[0005] With the widespread use of Massive MIMO (Massive MIMO) technology, which is used to improve system capacity and coverage, it is now impossible to obtain the optimal beam for the user during the PRACH phase. Therefore, the only way is to traverse all beams to determine the UE corresponding to the access information. This will seriously increase the access delay between the UE and the base station. Summary of the Invention

[0006] The present invention relates to the field of communication technology, and in particular to a random signal detection method based on FNN (Feedforward neural network, fully connected neural network), which is used to quickly establish a connection between a user and a base station in the 5G communication era and quickly find the sequence information corresponding to the preamble code received by the base station.

[0007] The present invention is a random signal detection method based on FNN, comprising the following steps:

[0008] Step 1: Obtain frequency domain data of random access signal;

[0009] Step 2: Determine the random access sequence length based on system configuration information;

[0010] Step 3: Based on the access sequence length obtained in step 2, the frequency domain data of the obtained random access signal is imported into the FNN-based random access signal sequence generation information judgment model to obtain the obtained random access signal sequence generation information;

[0011] Step 4: According to the access sequence length obtained in step 2, the frequency domain data of the obtained random access signal is imported into the timing advance judgment model of the random access signal based on the FNN to obtain the timing advance of the obtained random access signal.

[0012] The FNN-based random access signal sequence generation information judgment model in the above step 3 adopts a fully connected layer neural network.

[0013] The FNN-based random access signal sequence generation information judgment model in the above step 3 includes an input layer, a hidden layer, and an output layer.

[0014] The FNN-based random access signal sequence generation information judgment model in step 3 above includes one input layer, one hidden layer, and one output layer.

[0015] In the above-mentioned step 3, the FNN-based random access signal sequence generation information judgment model has the number of neurons in the input layer equal to the length of the random access sequence in step 2.

[0016] The FNN-based random access signal sequence generation information judgment model in step 3 above includes one hidden layer and 200 neurons.

[0017] The FNN-based random access signal sequence generation information judgment model in step 3 above includes one output layer and 34 neurons. The FNN-based random access signal sequence generation information judgment model in step 3 above includes one output layer and 55 neurons.

[0018] The FNN-based random access signal sequence generation information judgment model in step 3 above includes one input layer, two hidden layers, and one output layer.

[0019] The FNN-based random access signal sequence generation information judgment model in step 3 above includes two hidden layers, the first hidden layer has 1000 neurons, and the second hidden layer has 500 neurons.

[0020] The FNN-based random access signal sequence generation information judgment model in step 3 above is obtained through pre-training. During the training process, parameters are configured for the input layer, hidden layer, and output layer. The hidden layer activation function of the FNN-based random access signal sequence generation information judgment model is the "ReLU" activation function with L2 regularization. The output layer activation function is the "Softmax" activation function with L2 regularization.

[0021] The training of the above-mentioned FNN-based random access signal sequence generation information judgment model uses the Adam optimizer with a learning rate set to 0.0005.

[0022] The timing advance determination model of the random access signal based on the FNN in step 4 above includes one input layer, one hidden layer and one output layer.

[0023] In the timing advance determination model of the random access signal based on FNN in step 4 above, the number of neurons in the input layer is the length of the random access sequence in step 2.

[0024] The timing advance determination model of the random access signal based on FNN in the above step 4 includes one hidden layer and 100 neurons.

[0025] In the timing advance determination model of the random access signal based on FNN in step 4 above, the number of neurons in the output layer is 4.

[0026] The FNN-based random access signal timing advance determination model in step 4 above is obtained through pre-training. During the training process, parameters are configured for the input layer, hidden layer, and output layer. The hidden layer activation function of the FNN-based random access signal timing advance determination model is the "ReLU" activation function with L2 regularization. The output layer activation function is the "Softmax" activation function with L2 regularization.

[0027] The training of the timing advance judgment model of the FNN-based random access signal uses the Adam optimizer with a learning rate set to 0.0001.

[0028] The present invention can ultimately detect the input random access signal, quickly determine the sequence generation information and time advance corresponding to the input signal, and effectively solve the signal processing complexity in the actual system.

[0029] The present invention can generate the loss value and accuracy of the training set and validation set in each epoch of the model.

[0030] The present invention can draw a function curve of the loss value of the validation set of the model as a function of the number of epochs.

[0031] The present invention provides a random access signal detection method based on FNN. Compared with traditional algorithms, this technical solution can detect signals in a shorter time, reduce the complexity of signal processing, and establish a connection between users and base stations more quickly. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0033] Figure 1 The training of the sequence generation information judgment model of the random access signal based on the FNN in the first embodiment of the present invention (PRACH sequence with a length of 139) is shown;

[0034] Figure 2 The training of the sequence generation information judgment model of the random access signal based on the FNN in the second embodiment of the present invention (PRACH sequence with a length of 839) is shown;

[0035] Figure 3 The diagram shows the training of a timing advance determination model for a random access signal based on an FNN in the third embodiment of the present invention.

[0036] Figure 4 It is an implementation flow chart of the present invention. DETAILED DESCRIPTION

[0037] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0038] It should be understood that the embodiments described are only a portion of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0039] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "an", "the" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise.

[0040] The present invention is a random signal detection method based on FNN, comprising the following steps:

[0041] Step 1: Obtain frequency domain data of random access signal;

[0042] Step 2: Determine the random access sequence length based on system configuration information;

[0043] Step 3: Based on the access sequence length obtained in step 2, the frequency domain data of the obtained random access signal is imported into the FNN-based random access signal sequence generation information judgment model to obtain the obtained random access signal sequence generation information;

[0044] Step 4: According to the access sequence length obtained in step 2, the frequency domain data of the obtained random access signal is imported into the timing advance judgment model of the random access signal based on the FNN to obtain the timing advance of the obtained random access signal.

[0045] In 5G systems, the random access signal is defined as the physical random access channel (PRACH), which carries the random access sequence. For ease of description, this specification uses PRACH instead of random access signal and PRACH sequence instead of random access sequence.

[0046] Three embodiments are listed below:

[0047] Example 1:

[0048] For a PRACH sequence of length 139, the acquired PRACH frequency-domain data sequence matrix is ​​first converted and stored in an array. Each sequence is then labeled accordingly. The sequence is then loaded into a network model with one hidden layer for training, generating an FNN-based random access signal sequence generation information judgment model. This model has 139 neurons in the input layer, 200 neurons in the hidden layer, and a "relu" activation function with L2 regularization. The output layer has 34 neurons and a "softmax" activation function with L2 regularization. The validation set accuracy and loss values ​​are plotted over epochs during the training process.

[0049] When a PRACH signal is received, if the PRACH sequence length is determined to be 139 through system configuration, a sequence generation information judgment model of a random access signal based on an FNN is introduced to output a judgment of the corresponding sequence generation information.

[0050] Example 2:

[0051] For a PRACH sequence of length 839, the acquired PRACH frequency domain data sequence matrix is ​​first converted and stored in an array. Each sequence is then labeled accordingly. The sequence is then loaded into a two-hidden-layer network model for training, generating a FNN-based random access signal sequence generation information judgment model. This model has an input layer of 839 neurons; the first hidden layer has 1000 neurons, using the relu activation function with L2 regularization; the second hidden layer has 500 neurons, using the relu activation function with L2 regularization; and the output layer has 55 neurons, using the softmax activation function with L2 regularization. The validation set accuracy and loss values ​​are plotted over epochs during the training process.

[0052] When a PRACH signal is received, if the PRACH sequence length is determined to be 839 through system configuration, a sequence generation information judgment model of a random access signal based on an FNN is introduced, and a judgment result of the corresponding sequence generation information can be output.

[0053] Example 3

[0054] To determine the timing advance of random access signals, the PRACH signal is loaded into a network model with one hidden layer for training, generating an FNN-based random access signal timing advance determination model. This model has an output layer with 139 or 839 neurons; a hidden layer with 100 neurons, a relu activation function with L2 regularization; and an output layer with 4 neurons, a softmax activation function with L2 regularization. The validation set accuracy and loss values ​​are plotted over epochs during the training process.

[0055] When judging the timing advance of the received PRACH signal, the received PRACH signal is introduced into the timing advance judgment model of the random access signal based on the FNN, and a corresponding timing advance judgment result can be output.

[0056] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A random signal detection method based on FNN, characterized by: The following steps are involved: Step 1: Obtain frequency domain data of random access signal; Step 2: Determine the random access sequence length to be 139 or 839 based on system configuration information; Step 3: Based on the access sequence length obtained in step 2, the obtained frequency domain data of the random access signal is imported into the FNN-based random access signal sequence generation information judgment model to obtain the obtained random access signal sequence generation information; Step 4: According to the access sequence length obtained in step 2, the frequency domain data of the obtained random access signal is imported into the timing advance judgment model of the random access signal based on the FNN to obtain the timing advance of the obtained random access signal.

2. The random signal detection method based on FNN according to claim 1, wherein for a random access sequence of length 139, The FNN-based random access signal sequence generation information judgment model in step 3 adopts a fully connected layer neural network; The FNN-based random access signal sequence generation information judgment model in step 3 comprises an input layer, a hidden layer, and an output layer; The input layer has 1 layer; The number of neurons in the input layer is the length of the random access sequence in step 2; The hidden layer has 1 layer; The number of neurons in the hidden layer is 200; The output layer has 1 layer; The number of neurons in the output layer is 34.

3. The random signal detection method based on FNN according to claim 1, wherein for a random access sequence of length 839, The FNN-based random access signal sequence generation information judgment model in step 3 adopts a fully connected layer neural network; The FNN-based random access signal sequence generation information judgment model in step 3 comprises an input layer, a hidden layer, and an output layer; The input layer has 1 layer; The number of neurons in the input layer is the length of the random access sequence in step 2; There are 2 hidden layers; The first hidden layer has 1000 neurons; The second hidden layer has 500 neurons; The output layer has 1 layer; The number of neurons in the output layer is 55.

4. The random signal detection method based on FNN according to claim 1, wherein for a random access sequence of length 139 or 839, the method is characterized in that: The timing advance determination model of the random access signal based on the FNN in step 4 adopts a fully connected layer neural network; The timing advance determination model of the random access signal based on the FNN in step 3 comprises an input layer, a hidden layer and an output layer; The input layer has 1 layer; The number of neurons in the input layer is the length of the random access sequence in step 2; The hidden layer has 1 layer; The first hidden layer has 100 neurons; The output layer has 1 layer; The number of neurons in the output layer is 4.

5. The random signal detection method based on FNN according to claim 1, characterized in that: The timing advance judgment model of the random access signal based on the FNN is obtained through pre-training.

Images