ResNet multi-target signal DOA estimation algorithm based on beam forming

Abstract

The invention discloses a ResNet multi-target signal DOA estimation algorithm based on beam forming, and the method comprises the steps: employing a detector array, employing an LCMV beam forming algorithm, carrying out the scanning of a space, carrying out the phase adjustment in the scanning process, carrying out the signal enhancement of a signal from a scanning direction, and inhibiting the interference signals from other directions. A multi-target DOA estimation problem is converted into a dichotomy problem by using a residual network ResNet, and an efficient channel attention mechanism enhancement model is added in the residual network ResNet to extract signal features, so that the accuracy and robustness of target DOA estimation are improved. The method has the technical characteristics of high detection accuracy and precision and the like.

Description

Technical Field

[0001] The present invention relates to an estimation algorithm, and more specifically, to a ResNet multi-target signal DOA estimation algorithm based on beamforming, which belongs to the field of sound source localization. Background Art

[0002] DOA estimation is a sound source localization technology. It means that by analyzing the sound signal, according to the time difference or signal strength difference of the sound wave reaching each sensor, the direction in space from which the sound source comes can be calculated. This technology is widely used in speech recognition, audio signal processing, wireless communication and other fields. However, when the number of sound sources increases, DOA estimation is more susceptible to interference in complex environments, and it will be difficult to accurately distinguish multiple sound sources in approaching directions, especially when the signal strength is similar or overlapped, which has a great impact on the accuracy of the estimation algorithm. Traditional DOA estimation methods based on physical models may face limitations in resolution when facing multi-target estimation, and it is difficult to cope with complex signal environments. In addition, these methods and algorithms have robustness issues in the face of noise and non-ideal conditions. Summary of the invention

[0003] In order to solve the above-mentioned problems of the prior art, the present invention provides a ResNet multi-target signal DOA estimation algorithm based on beamforming, which has the technical characteristics of reducing interference between targets and environmental noise interference, and effectively distinguishing the source angle of each target under multi-target interference.

[0004] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0005] The present invention discloses a ResNet multi-target signal DOA estimation algorithm based on beamforming. The method comprises: using a detector array, applying an LCMV beamforming algorithm, scanning a space, performing phase adjustment during the scanning process, enhancing signals from the scanning direction, and suppressing interference signals from other directions, then using a residual network ResNet to convert the multi-target DOA estimation problem into a binary classification problem, and adding an efficient channel attention mechanism to the residual network ResNet to enhance the extraction of signal features by the model, thereby improving the accuracy and robustness of target DOA estimation.

[0006] Preferably, the method comprises the following steps:

[0007] Step S1: The circular array receives the drone signal, calculates the distance from the sound source to each array and samples accordingly;

[0008] Step S2: using the LCLV beamforming algorithm to adjust the phase of the sampled signal;

[0009] Step S3: splicing the phase-adjusted array output with its weighted sum signal;

[0010] Step S4: Use the ResNet-ECA network to extract features from the data and perform training to obtain a multi-target DOA estimation model;

[0011] Step S5: Use the trained model to infer the newly adjusted and spliced ​​data;

[0012] Step S6: The network determines whether there is a target at each scanning angle according to the multi-target DOA estimation model.

[0013] Preferably, the step S1 specifically includes: using an array composed of omnidirectional sensors to receive surrounding sound signals, all detectors of the array sample the signals, and recording the signals sampled and received by the circular array at time t.

[0014] Preferably, steps S2 to S3 are specifically as follows: for the signal received by each detector in the array at time t, the LCMV beamforming algorithm is used to scan in all directions. For the scan in each azimuth, the signal received by each detector is first multiplied by a suitable weighting vector, phase adjusted, and then added. Each signal after phase adjustment and the added signal are combined into a set of data for determining the existence of the signal in the scanning direction.

[0015] Preferably, step S4 is specifically as follows: during the network training process, the phase-adjusted signal data set in each direction is made into a complete network training set, and three labels are added when naming, namely, the array scanning angle, the target incident angle and the prior judgment result (whether there is a target). The above training set is used as the input of the ResNet-ECA signal recognition network and input into the ResNet network. After the training is completed, a multi-target DOA estimation model is obtained.

[0016] Preferably, in step S5, during the network inference process, the detector array is placed in the required environment, and steps S1 and S2 are repeatedly executed, and the signal received by the array and processed by the LCMV beamforming algorithm is used as input data and input into the multi-target DOA estimation model in step S4, and then the model output is used to determine whether there is a target at each scanning angle.

[0017] Preferably, the input signal of the ResNet network (ResNet-ECA signal recognition network) includes the phase-adjusted signals of each detector and the sum of all adjusted signals;

[0018] Phase adjustment means that the signal received by the array is multiplied by a weighted vector according to the array scanning angle through the LCMV beamforming technology;

[0019] The sum of the phase-adjusted signals enables the phase-adjusted signals within the scanning angle to be superimposed in phase, thereby enhancing the signal, while the noise and interference direction signals within the non-scanning angle are not superimposed in phase and are suppressed.

[0020] Preferably, the output of the ResNet-ECA signal recognition network is to determine whether a target exists in each scanning direction, rather than the arrangement of the scanning angle and the target incident angle. The network needs to record the number of targets detected in all scanning directions.

[0021] Preferably, the input data set of the ResNet-ECA signal recognition network is a two-dimensional matrix, the composition of which is generated by the following steps:

[0022] The signal output by each detector is a one-dimensional signal, which is a sampling of the signal over a period of time. In a single scanning angle calculation, it is still a one-dimensional signal after phase adjustment by the beamforming algorithm. The sum of the adjusted signals of each detector is also a one-dimensional signal. These signals together constitute a two-dimensional data matrix.

[0023] Preferably, the input two-dimensional data matrix of the ResNet-ECA signal recognition network, the data set can be a two-dimensional data matrix obtained from a single scanning angle and / or a giant two-dimensional data matrix composed of all two-dimensional data matrices corresponding to all scanning angles;

[0024] When the input of the ResNet network is only a two-dimensional matrix obtained from a scanning angle, the output of the ResNet network is the judgment whether there is a target signal at this scanning angle; when the input of the ResNet network is a two-dimensional data matrix obtained from omnidirectional scanning angles, the output of the ResNet network (ResNet-ECA signal recognition network) is the ResNet network's judgment of the number of targets existing in all directions and the location of each target.

[0025] Preferably, the phase adjustment signal is obtained by multiplying the original signal collected by the detector by the weighted vector, which is a digital signal. The weighted vector can be appropriately adjusted according to the accuracy requirements of the scanning angle or the specific environment, without retraining the RecNet-ECA network. At the same time, because the weighted vector is a digital signal calculated during the system design process, omnidirectional scanning can be completed in parallel without being substituted into the RecNet-ECA network one by one, saving calculation and judgment time.

[0026] Preferably, based on the beamforming ResNet network, the output of the ResNet network (ResNet-ECA signal recognition network) is 1 or 0, that is, it only determines whether there is a target in the scanning direction. When the input of the ResNet-ECA signal recognition network is a data matrix corresponding to the omnidirectional scanning angle, the ResNet-ECA signal recognition network can determine any number of targets, rather than a specific number of targets calibrated according to the training set labels.

[0027] Beneficial effects:

[0028] (1) Traditional methods rely on strong assumptions about the physical signal model. The beamforming-based ResNet multi-target DOA estimation algorithm of the present invention can effectively and accurately distinguish the source angle of each target under multi-target interference.

[0029] (2) The present invention beneficially combines the traditional method with the deep learning algorithm. The traditional LCMV beamforming algorithm can enhance the signal in each scanning direction by selecting a suitable weighting vector and suppress interference in other directions, thereby greatly reducing interference between targets and environmental noise interference.

[0030] (3) By combining the one-dimensional signal obtained by detector sampling with the signal weighted by the LCMV beamforming algorithm and using the ResNet-ECA signal recognition neural network, it is possible to effectively identify whether the phase of each detector sampling signal is aligned with the phase of the synthetic signal. The more aligned the phase is, the greater the possibility of the existence of a target in that direction.

[0031] (4) The weighted vector for phase adjustment can be adjusted appropriately according to the accuracy requirements of the scanning angle or the specific environment, without the need to retrain the RecNet-ECA network. At the same time, because the weighted vector is a digital signal calculated during the system design process, omnidirectional scanning can be completed in parallel without having to be substituted into the network one by one, saving calculation and judgment time.

[0032] (5) It is relatively easy to use a neural network to discriminate the angles of multiple targets in the present invention. It only needs to determine whether there is a target in a certain direction. Compared with other neural networks that are not combined with beamforming and need to determine the location of each target based on the classification probability of different combinations of targets in various directions, the computational complexity of the present invention is lower than that of other existing target DOA estimation methods.

[0033] (6) The present invention can be widely used in wireless communications, target detection and tracking, speech processing, self-driving cars, environmental monitoring and other fields, and also provides new ideas and methods for monitoring underwater targets in the ocean. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1This is a flow chart of the estimation algorithm of the present invention.

[0035] Figure 2 This is one of the visualizations of the dataset and the beamforming effect.

[0036] Figure 3 This is the second visualization of the data set and beamforming effect.

[0037] Figure 4 This is the third visualization of the data set and beamforming effect.

[0038] Figure 5 This is the fourth visualization of the data set and beamforming effect. DETAILED DESCRIPTION

[0039] The present invention will be further described below in conjunction with the accompanying drawings, but the present invention is not limited to the following embodiments.

[0040] like Figure 1-5 The figure shows a specific embodiment of a ResNet multi-target signal DOA estimation algorithm based on beamforming. First, the sound source signals emitted by multiple drones are received through a circular array, and the LCMV classic beamforming algorithm is used to perform beam scanning in 360 degrees in all directions. The signal phase is adjusted for each direction to enhance the signal and reduce noise and the mutual influence between multi-target signals. Subsequently, the output data of each microphone after phase adjustment and the beamforming output data are merged, and then the ResNet-ECA target signal DOA estimation algorithm is used to train the data to obtain a DOA estimation model. The model is used to classify and identify whether there is a target in the scanning direction, thereby determining the final source direction of the target.

[0041] This embodiment provides a ResNet multi-target signal DOA estimation algorithm based on beamforming, and the method specifically includes the following steps:

[0042] Step S1: Use an array composed of omnidirectional sensors to receive surrounding sound signals. All detectors in the array sample the signals and record the signals received by the circular array at time t.

[0043] In this embodiment, the microphone array is assumed to be an 8-element circular array with a radius of 0.15m. The coordinate system is established with the center of the circular array as the reference point. The array scans clockwise with the positive direction of the coordinate axis Y axis as 0°, and the range is Φ∈[0,2π). The number of sampling snapshots is 400, the sound speed is 340m / s, and the sampling frequency is f sThe frequency is 4000Hz, and the detector can sample the target signal 400 times each time. Therefore, the sound clip signal is intercepted every 600 time points as a detection, which can ensure that all detectors can fully sample the signal. Each detector samples 400 time points, and 8 detectors detect at the same time to get an 8*400 matrix data.

[0044] Step S2: The signal received by each detector in the array at time t is scanned in all directions using the LCMV beamforming algorithm. For each azimuth scan, the signal received by each detector is first multiplied by a suitable weighting vector (suitable means: data selected according to actual needs), phase adjusted, and then added. Each signal after phase adjustment and the added signal form a set of data for determining the existence of the signal in the scanning direction.

[0045] In this embodiment, a suitable weighting vector is selected to perform signal enhancement processing on the signal received at the scanning angle, and suppress signals in other directions. If during the scanning process, the scanning direction is the same as the direction of the sound signal source of the drone, then the signal in this direction is the strongest after enhancement, that is, the propagation delay of each array element is basically made up, and the array element output signals are basically in phase. The output signals of each array element after phase adjustment are weighted and summed to obtain the enhanced signal, and the enhanced 1*400 signal data is spliced ​​with the 8 array element output data after phase adjustment (8*400 matrix data) to obtain a 9*400 matrix data.

[0046] Step S3: During the ResNet network training process, the signal data set after phase adjustment in each direction is made into a complete ResNet network training set, and three labels are added when naming, namely, array scanning angle, target incidence angle and prior judgment result (whether there is a target). The above training set is used as the input of the ResNet-ECA signal recognition network and input into the ResNet network. After the training is completed, a multi-target DOA estimation model is obtained;

[0047] Figure 2-5 The detector output data and its synthetic signal of this embodiment are used to produce a visualization of the training set and beamforming effect. Figure 3 This is an example of the input network training set. The first 8 signals show the signals of the 8 array elements after phase adjustment, and the 9th signal is the final weighted synthetic output data of the 8 microphone signals. It can be seen that at some moments marked by the boxes, the phases of the 8 microphone data are basically aligned and added to form a more obvious peak in the 9th signal. Figure 3-5 The signal processing process under the signal-to-noise ratio of 10dB, 0dB, and -5dB is shown respectively. Figure 3-5 The child Figure 1 ,2 The three figures represent the signal of the first row of the data set without noise, the signal of the first row of the data set with noise, and the signal of the ninth row of the data set. Figure 3-5 The child Figure 2 This shows that as the signal-to-noise ratio continues to decrease, the signal gradually becomes distorted. When the signal-to-noise ratio is -5dB, the signal is almost completely covered by noise. Figure 3-5 The child Figure 3 It can be clearly seen that after the phase adjustment of the 8-segment signal, the noise and interference signals can be effectively removed.

[0048] In this embodiment, when constructing a training set, due to the symmetry of the circular array, the array scanning angle and the incident angle of the two UAV signal sources are set within the angle range of 0-90°. The three angles are freely combined every 10 degrees, with a total of 1000 angle combinations. When the scanning angle coincides with the incident angle of any of the two signal sources, it is labeled "with target", and in other cases it is labeled "with target". Data sampling is performed 30 times for each combination, and a total of 30,000 data are established as a data set.

[0049] The sorted data set is input into the pre-built and initialized ResNet-ECA neural network for training. By iteratively training a large amount of data and gradually updating the parameters, a multi-target DOA estimation model is finally obtained.

[0050] Step S4: During the network inference process, the detector array is placed in the required environment, and steps S1 and S2 in the embodiment are repeatedly executed. The signal received by the array and processed by the LCMV beamforming algorithm is used as input data and input into the multi-target DOA estimation model described in step S3. The model output determines whether there is a target at each scanning angle.

[0051] In this embodiment, the accuracy and root mean square error are used to evaluate the method of the present invention. When the signal-to-noise ratio is 15dB, the beamforming-based ResNet multi-target signal DOA estimation algorithm of the present invention has an accuracy of 96.73% for DOA estimation of two targets in the air, and a root mean square error of 2.53°. By comparing the DOA estimation results of ResNet, GCC-PHAT+ResNet and beamforming+MobileNet for multiple targets, it can be obtained that the DOA estimation accuracy of ResNet for multiple targets is 31.51%, and the root mean square error is 9.63°, the DOA estimation accuracy of GCC-PHAT+ResNet for multiple targets is 85.12%, and the root mean square error is 5.81°, and the DOA estimation accuracy of beamforming+MobileNet for multiple targets is 82.48%, and the root mean square error is 7.39°. It can be obtained that the method of the present invention effectively improves the accuracy of DOA estimation compared with the existing methods.

[0052] In addition, when there are three or more targets in the air, the method of the present invention can still maintain a DOA estimation accuracy of more than 96% and a root mean square error of less than 5°, which further demonstrates that the present invention exhibits strong robustness and generalization ability in multi-target DOA estimation, while reducing the complexity of signal processing in multi-target estimation.

[0053] Finally, it should be noted that the present invention is not limited to the above embodiments, and there are many variations. All variations that can be directly derived or associated with the content disclosed by ordinary technicians in this field should be considered as the protection scope of the present invention.

Claims

1. A ResNet multi-target signal DOA estimation algorithm based on beamforming, characterized in that: The method includes: using a detector array, applying the LCMV beamforming algorithm, scanning the space, performing phase adjustment during the scanning process, achieving signal enhancement for signals from the scanning direction, and suppressing interference signals from other directions, and then using the residual network ResNet to transform the multi-target DOA estimation problem into a binary classification problem, and adding an efficient channel attention mechanism to the residual network ResNet to enhance the model's extraction of signal features, thereby improving the accuracy and robustness of target DOA estimation.

2. According to claim 1, a beamforming-based ResNet multi-target signal DOA estimation algorithm is characterized in that The method comprises the following steps: Step S1: The circular array receives the drone signal, calculates the distance from the sound source to each array and samples accordingly; Step S2: using the LCLV beamforming algorithm to adjust the phase of the sampled signal; Step S3: splicing the phase-adjusted array output with its weighted sum signal; Step S4: Use the ResNet-ECA network to extract features from the data and perform training to obtain a multi-target DOA estimation model; Step S5: Use the trained model to infer the newly adjusted and spliced ​​data; Step S6: The network determines whether there is a target at each scanning angle according to the multi-target DOA estimation model.

3. According to claim 2, a ResNet multi-target signal DOA estimation algorithm based on beamforming is characterized in that: The step S1 specifically includes: using an array composed of omnidirectional sensors to receive surrounding sound signals, all detectors in the array sample the signals, and recording the signals sampled and received by the circular array at time t.

4. According to claim 2, a ResNet multi-target signal DOA estimation algorithm based on beamforming is characterized in that: The steps S2 to S3 are specifically as follows: for the signal received by each detector in the array at time t, the LCMV beamforming algorithm is used to scan in all directions. For the scan in each azimuth, the signal received by each detector is first multiplied by a weighted vector, phase adjusted, and then added. Each signal after phase adjustment and the added signal are combined into a set of data for determining the existence of the signal in the scanning direction.

5. The ResNet multi-target signal DOA estimation algorithm based on beamforming according to claim 2, characterized in that: The step S4 is specifically as follows: during the network training process, the phase-adjusted signal data set in each direction is made into a complete network training set, and three labels are added when naming, namely, the array scanning angle, the target incident angle and the prior judgment result. The above training set is used as the input of the ResNet-ECA signal recognition network and input into the ResNet network. After the training is completed, a multi-target DOA estimation model is obtained.

6. The ResNet multi-target signal DOA estimation algorithm based on beamforming according to claim 2, characterized in that: In step S5, during the network inference process, the detector array is placed in the required environment, and steps S1 and S2 are repeated. The signal received by the array and processed by the LCMV beamforming algorithm is used as input data and input into the multi-target DOA estimation model in step S4. The model output is then used to determine whether there is a target at each scanning angle.

7. The ResNet multi-target signal DOA estimation algorithm based on beamforming according to claim 2, characterized in that: The input signal of the ResNet network includes the phase-adjusted signals of each detector and the sum of all adjusted signals; phase adjustment means that the signal received by the array is multiplied by a weighted vector according to the array scanning angle through the LCMV beamforming technology.

8. The ResNet multi-target signal DOA estimation algorithm based on beamforming according to claim 2, characterized in that: The output of the ResNet-ECA signal recognition network is to judge whether there is a target in each scanning direction, rather than the arrangement of the scanning angle and the target incident angle. The ResNet network needs to record the number of targets detected in all scanning directions.

9. The ResNet multi-target signal DOA estimation algorithm based on beamforming according to claim 2, characterized in that: The input dataset of the ResNet-ECA signal recognition network is a two-dimensional matrix, whose composition is generated by the following steps: The signal output by each detector is a one-dimensional signal, which is a sampling of the signal over a period of time. In a single scanning angle calculation, it is still a one-dimensional signal after phase adjustment by the beamforming algorithm. The sum of the adjusted signals of each detector is also a one-dimensional signal. These signals together constitute a two-dimensional data matrix.

10. The beamforming-based ResNet multi-target signal DOA estimation algorithm according to claim 5, characterized in that: The input 2D data matrix of the ResNet-ECA signal recognition network can be a 2D data matrix obtained from a single scanning angle and / or a giant 2D data matrix consisting of all 2D data matrices corresponding to all scanning angles; When the input of the ResNet network is only a two-dimensional matrix obtained from a scanning angle, the output of the ResNet network is the judgment whether there is a target signal at this scanning angle; when the input of the ResNet network is a two-dimensional data matrix obtained from omnidirectional scanning angles, the output of the ResNet network is the number of targets judged by the ResNet network in all directions and the location of each target.

Images