A method for OFDM-IM signal detection based on noise reduction network
By preprocessing and data enhancement of OFDM-IM signals, a noise reduction and detection network is built, and a judgment is made in combination with energy distribution, the errors of noise distribution in signal detection and insufficient sample diversity are solved, and the problem of noise distribution in activation carrier position judgment and insufficient sample diversity is improved. Accuracy of signal detection and model adaptability.
Patent Information
- Application Number
- CN202211444691.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-11-18
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2042-11-18
AI Technical Summary
In OFDM-IM signal detection, although the noise reduction signal reduces noise, it may lead to errors in the activation carrier position judgment, and the diversity of signal samples after preprocessing is insufficient, limiting the learning range of the model and being unable to cope with complex noise environments.
By obtaining received signals and sent signals for preprocessing, data enhancement is carried out, noise reduction and detection is built, signal noise reduction and detection is used for signal noise reduction and detection, and judgment is made in combination with energy distribution, and joint judgment methods are designed to improve the model's generalization ability to transform.
It improves the accuracy of signal detection and the generalization ability of the model, enhances the adaptability to complex noise environments, and improves the performance of signal detection.
Smart Images

Figure CN115865593B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of signal detection of wireless communications, and in particular relates to an OFDM-IM signal detection method based on a noise reduction network. Background Art
[0002] Orthogonal frequency division multiplexing technology has become one of the key technologies in the 4G and 5G fields due to its advantages such as high spectrum utilization and good resistance to multipath fading. In recent years, the emergence of index modulation OFDM (OFDM-Index Modulation, OFDM-IM) has become a modulation method to replace or supplement the existing OFDM system. In the OFDM-IM system, the modulation dimension of OFDM is expanded from one to two. OFDM-IM divides all subcarriers into two parts, one is silent subcarriers and the other is activated subcarriers. The modulated constellation point information is placed on the activated subcarriers. Of course, silent and non-silent carriers can also carry additional information. Index modulation OFDM has many advantages over traditional OFDM. First, OFDM-IM modulation can compromise between spectrum efficiency and transmission performance. Second, when sending OFDM-IM signals, it has a lower peak-to-average ratio than OFDM signals. Finally, OFDM-IM signals greatly reduce inter-carrier interference and improve the robustness of the system. In short, OFDM-IM technology has a good application prospect and is worthy of our in-depth exploration.
[0003] In the OFDM-IM system, for signal detection research, scholars use machine learning to detect OFDM-IM signals. At the same time, the development of noise reduction technology has greatly promoted the development of communications.
[0004] Soltani et al. (M. Soltani, V. Pourahmadi, A. Mirzaei and H. Sheikhzadeh, "Deep Learning-Based Channel Estimation," in IEEE Communications Letters, vol. 23, no. 4, pp. 652-655, April 2019, doi: 10.1109 / LCOMM.2019.2898944.) proposed a two-stage cascade network ChannelNet, which converts the pilot signal into a low-resolution image, improves the resolution through a super-resolution algorithm, and passes through a DNCNN network (K. Zhang, W. Zuo, Y. Chen, D. Meng and L. Zhang, "Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising," in IEEE Transactions on Image Processing, vol. 26, no. 7, pp. 3142-3155, July 2017, doi:10.1109 / TIP.2017.2662206.) to achieve channel estimation. This study shows that noise reduction technology has important research value in improving communication performance, especially the development of deep learning noise reduction technology plays an increasingly important role in the research of communication field.
[0005] TVLuong et al. (TVLuong, Y.Ko, NAVien, DHNNguyen, and M.Matthaiou, "Deep learning-based detector for OFDM-IM," IEEE Wireless Commun. Lett., vol.8, no.4, pp.1159–1162, Aug.2019.) proposed a machine learning-based OFDM-IM detector. This paper uses deep learning methods to detect signals in OFDM-IM systems, and can achieve near-optimal bit error rate performance with very low model training time. Although this method can effectively recover information using a low-complexity network, it does not significantly improve bit error rate performance.
[0006] Based on the current research on OFDM-IM signal detection, it is found that there are still some challenges in the process of using noise reduction networks and detection networks for signal detection:
[0007] 1. Although the noise reduction signal has reduced noise, its energy distribution may also lead to erroneous judgment on the location of the activated carrier. Energy distribution plays an auxiliary role in determining the location of the activated subcarrier and is crucial to signal detection;
[0008] 2. In response to problems such as insufficient diversity of signal samples after preprocessing, the limited noise distribution will limit the learning range of the model, making it unable to cope with more complex noise environments. Therefore, enriching the noise of the signal during the model training stage is crucial to improving the generalization ability of the network. Summary of the invention
[0009] In order to solve the above technical problems, the present invention proposes an OFDM-IM signal detection method based on a noise reduction network, comprising:
[0010] S1: Get the received signal Y, the transmitted signal X and the channel information H, pre-process the received signal Y through the channel information H, and obtain the equalized signal in a two-dimensional format
[0011] S2: Data enhancement is performed on the signal on the silent subcarrier of the two-dimensional equalized signal to obtain the enhanced equalized signal
[0012] S3: Build a noise reduction network to enhance the equalization signal And train the denoising network with the transmitted signal X as the label.
[0013] S4: Convert the equalized signal in two-dimensional format Input the trained denoising network to perform data denoising, and obtain the denoised signal and the energy distribution of the denoised signal;
[0014] S5: reconstructing the energy distribution of the denoised signal;
[0015] S6: construct a detection network and train the detection network;
[0016] S7: Use the trained detection network to detect the denoised signal, and determine the location of the activated carrier by reconstructing the energy distribution to obtain the signal estimation value. The estimated value of the signal Make a decision and get the estimated bit stream
[0017] Beneficial effects of the present invention:
[0018] The present invention takes into account that the IQ signal of the denoised signal may change the distribution quadrant of the original signal. In order to better determine the activated carrier position, a joint determination method of the denoised signal and the received signal is designed. In view of the problem that the diversity of the signal samples after preprocessing is insufficient, and considering that data enhancement can enrich the noise of the signal, the generalization ability of the model is enhanced by strengthening the silent subcarriers in the model training stage, thereby further improving the denoising performance. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 A schematic diagram of an OFDM-IM signal detection method based on a noise reduction network according to the present invention;
[0020] Figure 2 It is a schematic diagram of the OFDM-IM signal detection model based on the noise reduction network of the present invention. DETAILED DESCRIPTION
[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0022] An OFDM-IM signal detection method based on a noise reduction network, such as Figure 1 As shown, including:
[0023] S1: Get the received signal number Y, the transmitted signal X and the channel information H, pre-process the received signal Y through the channel information H, and obtain the equalized signal in a two-dimensional format
[0024] S2: Data enhancement is performed on the signal on the silent subcarrier of the two-dimensional equalized signal to obtain the enhanced equalized signal
[0025] S3: Build a noise reduction network to enhance the equalization signal And train the denoising network with the transmitted signal X as the label;
[0026] S4: Convert the equalized signal in two-dimensional format Input the trained denoising network to perform data denoising, and obtain the denoised signal and the energy distribution of the denoised signal;
[0027] S5: reconstructing the energy distribution of the denoised signal;
[0028] S6: construct a detection network and train the detection network;
[0029] S7: Use the trained detection network to detect the denoised signal, and determine the location of the activated carrier by reconstructing the energy distribution to obtain the signal estimation value. The estimated value of the signal Make a decision and get the estimated bit stream
[0030] like Figure 2 As shown in the figure, it is assumed that the channel information H is known at the receiving end, the received OFDM-IM signal is Y, and the transmitted signal is X; for a single-input single-output communication link (SISO), the OFDM-IM system consists of N C subcarriers, where N C The subcarriers are divided into L groups, and the number of subcarriers in each group is N L , then we can know N L =N C / L, for a set of received signals, if only one subcarrier is activated, it can be assumed that Among them, Y M It is the received signal that activates the carrier position.
[0031] In order to simplify the training process, define a constant Θ = {Θ DE ,Θ IM}, where Θ DE is the constant of the denoising network, Θ IM is the constant of the detection network, f IM and f DE They are noise reduction function and detection function respectively. Before training the network, we need to randomly generate data bit stream b, and we can get our transmitted signal X through OFDM-IM modulation.
[0032] Data enhancement is performed on a signal on a silent subcarrier of an equalized signal in a two-dimensional format, including:
[0033]
[0034] in, represents the enhanced equalized signal, n1, Respectively represent the silent carrier position 1, N L The noise distribution, is the real part of the equalized signal, is the imaginary part of the equalized signal.
[0035] We continue to add complex noise to the silent subcarriers to make the symbols of the silent carriers larger, or make the symbols of the activated carriers smaller. During the training phase, the model can learn more complex noise environments, and the model will be more robust and show better performance.
[0036] The denoising network includes: a CNN structure consisting of 20 convolutional layers, the first layer uses 64 filters of size 3x3x1, followed by a Relu activation function, the subsequent 18 convolutional layers all use 64 filters of size 3x3x64, and then batch normalization is performed, the activation function is also the Relu function, and the last layer uses a 3x3x64 filter to reconstruct the output.
[0037] After the enhanced equalization signal And train the denoising network with the transmitted signal X as the label, including:
[0038]
[0039] Among them, Loss1 represents the loss function of the denoising network, M1 represents the training set size of the denoising network, Ψ represents the output of the denoising network, X represents the sending signal, and |||| represents the norm operation.
[0040] The equalized signal in two-dimensional format Input the trained denoising network to perform data denoising, including:
[0041] The signal after noise reduction:
[0042] Y de =f DE (θ DE ; Y / H)
[0043] Among them, Y de represents the signal after noise reduction, f DE represents the noise reduction function, θ DE represents the constant of the noise reduction network, Y represents the received signal, and H represents the channel information;
[0044] Energy distribution of the signal after noise reduction:
[0045]
[0046] Among them, E de Represents the energy distribution of the signal after noise reduction, Y de represents the signal after noise reduction, and They respectively represent the noise reduction signals at the corresponding carrier positions.
[0047] Through the noise reduction network, a noise-reduced signal can be obtained. The noise-reduced signal has a clearer energy distribution, but there is also the possibility of misjudgment, which will misjudge the location of the activated carrier. In order to obtain a more effective energy distribution, the energy distribution of the noise-reduced equalized signal and the energy distribution of the received signal are combined. Sum To determine the activation position.
[0048] Reconstruct the energy distribution of the denoised signal, including:
[0049]
[0050] Among them, E sum represents the energy distribution of the reconstructed signal, Y de represents the signal after noise reduction, Y represents the received signal, γ represents the energy distribution coefficient, SNR represents the signal-to-noise ratio in an actual channel environment.
[0051] The detection network includes: a DNN structure consisting of a 2-layer fully connected network, the first layer has 128 neurons, the activation function is a Tanh function, the second layer has S neurons, the activation function is a Sigmoid function, wherein S represents the number of bits of the output.
[0052] Training the detection network includes:
[0053]
[0054] Among them, Loss2 represents the loss function of the detection network, M2 represents the training set size of the detection network, L represents the output of the detection network, b represents the randomly generated data bit stream, and |||| represents the norm operation.
[0055] The trained detection network is used to detect the denoised signal, and the location of the activated carrier is determined by reconstructing the energy distribution to obtain the signal estimation value. include:
[0056]
[0057] Among them, f IM and f DE Respectively represent the detection function and the noise reduction function, θ IM and θ DE They represent the constants of the detection function and the noise reduction function respectively, Y represents the received signal, and H represents the channel information.
[0058] The estimated value of the signal Make a decision and get the estimated bit stream Includes: Calculation of the estimated signal The Euclidean distance of the sent bits 0 and 1 is used, and the bit value corresponding to the shortest Euclidean distance is used as the final bit stream
[0059] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A method for detecting OFDM-IM signals based on a noise reduction network, characterized in that: include: S1: Get the received signal Y, the transmitted signal X and the channel information H, pre-process the received signal Y through the channel information H, and obtain the equalized signal in a two-dimensional format S2: Data enhancement is performed on the signal on the silent subcarrier of the two-dimensional equalized signal to obtain the enhanced equalized signal Data enhancement is performed on a signal on a silent subcarrier of an equalized signal in a two-dimensional format, including: in, represents the enhanced equalized signal, n1, Respectively represent the silent carrier position 1, N L The noise distribution, is the real part of the equalized signal, is the imaginary part of the equalized signal; S3: Build a noise reduction network to enhance the equalization signal And train the denoising network with the transmitted signal X as the label; The denoising network comprises: a CNN structure consisting of 20 convolutional layers, wherein the first layer uses 64 filters of size 3x3x1, followed by a Relu activation function, and the subsequent 18 convolutional layers all use 64 filters of size 3x3x64, and then batch normalization is performed, and the activation function is also the Relu function, and the last layer uses a 3x3x64 filter to reconstruct the output; S4: Convert the equalized signal in two-dimensional format Input the trained denoising network to perform data denoising, and obtain the denoised signal and the energy distribution of the denoised signal; The equalized signal in two-dimensional format Input the trained denoising network to perform data denoising, including: The signal after noise reduction: Y de =f DE (θ DE ;Y / H) Among them, Y de represents the signal after noise reduction, f DE represents the noise reduction function, θ DE represents the constant of the noise reduction network, Y represents the received signal, and H represents the channel information; Energy distribution of the signal after noise reduction: Among them, E de represents the energy distribution of the signal after noise reduction, and Respectively represent the noise reduction signals at the corresponding carrier positions; S5: reconstructing the energy distribution of the denoised signal; Reconstruct the energy distribution of the denoised signal, including: Among them, E sum represents the energy distribution of the reconstructed signal, γ represents the energy distribution coefficient, SNR represents the signal-to-noise ratio in the actual channel environment; S6: construct a detection network and train the detection network; The detection network comprises: a DNN structure consisting of a 2-layer fully connected network, wherein the first layer has 128 neurons, the activation function is a Tanh function, and the second layer has S neurons, the activation function is a Sigmoid function, wherein S represents the number of bits of the output; S7: Use the trained detection network to detect the denoised signal, and determine the location of the activated carrier by reconstructing the energy distribution to obtain the signal estimation value. The estimated value of the signal Make a decision and get the estimated bit stream The trained detection network is used to detect the denoised signal, and the location of the activated carrier is determined by reconstructing the energy distribution to obtain the signal estimation value. include: Among them, f IM represents the detection function, θ IM A constant representing the detection function.
2. The OFDM-IM signal detection method based on a noise reduction network according to claim 1, characterized in that: After the enhanced equalization signal And train the denoising network with the transmitted signal X as the label, including: Among them, Loss1 represents the loss function of the denoising network, M1 represents the training set size of the denoising network, Ψ represents the output of the denoising network, X represents the sending signal, and |||| represents the norm operation.
3. The OFDM-IM signal detection method based on a noise reduction network according to claim 1, characterized in that: Training the detection network includes: Among them, Loss2 represents the loss function of the detection network, M2 represents the training set size of the detection network, L represents the output of the detection network, b represents the randomly generated data bit stream, and |||| represents the norm operation.
4. The OFDM-IM signal detection method based on a noise reduction network according to claim 1, characterized in that: The estimated value of the signal Make a decision and get the estimated bit stream Includes: Calculation of the estimated signal The Euclidean distance of the sent bits 0 and 1 is used, and the bit value corresponding to the shortest Euclidean distance is used as the final bit stream
