Search

Aiming at the high computational cost of traditional compressed sensing algorithms, this paper proposed a sparse imaging algorithm based on depth expansion model from the perspective of sparse signal recovery. Firstly, it constructed a complex sparse reconstruction network VAMP-Net. In VAMP-Net, it divided complex regressive echo signal into real part and imaginary part as input. Secondly, it substituted the input into the iterative block based on VAMP algorithm. Finally, it carried out the optimal nonlinear sparse transformation by convolutional neural module to obtain the recovered real part and imaginary part signals, and then merged them to obtain the restored target image. As for the proposed algorithm, this paper used artificial data sets to conduct simulation experiments under different target density, iteration times and noise environment, and compared with the traditional iterative shrinkage threshold algorithm and deep learning reconstruction algorithm. Then it used the measured data with different sparsity for field measurement verification. Experimental results show that the image reconstructed by this algorithm has better performance in NMSE, TBR, reconstruction speed and memory usage. [ABSTRACT FROM AUTHOR]

In order to solve the problems of poor anti-interference and invasion of user privacy in traditional methods of using cameras for human activity recognition, this paper proposed a human activity recognition network based on spatiotemporal attention of millimeter wave radar 3D point cloud data to achieve accurate perception of intelligent application context. The network firstly used a secondary sliding time window to accumulate and separate the point cloud data generated by human activities as the input of the classifier, then used the PointLSTM unit to aggregate point features and states according to the point cloud coordinate relationship to extract the time sequence features of human activities, and then spliced temporal-spatial features, reduced the overall network computation and enhanced the network’s aggregation ability for local featured through sampling grouping modules, and finally used a stacked attention module to deeply fuse global and local features in temporal-spatial point cloud data to complete the accurate classification of human activities. This paper used millimeter wave radar to collect point cloud datasets of various human activities, the experimental results show that the average accuracy of the proposed spatiotemporal attention network can reach 98.64%,which can effectively identify complex and small-difference human activities, and meet the requirements of the human activity recognition system. [ABSTRACT FROM AUTHOR]

Radar echo extrapolation has been widely used in short-term heavy precipitation nowcasting. Aiming at the problem that traditional radar echo extrapolation methods do not make full use of massive historical meteorological data, which leads to low forecast accuracy, this paper proposed a new radar echo extrapolation method based on DPCNN. With the addition of dynamic probability calculation layer, DPCNN was able to calculate the corresponding probability convolution kernel for different radar echo input sequences and used it in the subsequent extrapolation calculations. Dynamic probability calculation layer enabled the network to make corresponding probability adjustments according to different input sequences in the prediction stage and enhance the connection between extrapolation and input sequences. The extrapolation experiment of short-term heavy precipitation in a certain area verifies the validity of the model from four aspects: extrapolated image, CSI, FAR, and POD. [ABSTRACT FROM AUTHOR]

In order to solve the problem of high time complexity in the signal recognition process by using convolution neural network, this paper proposed a new algorithm based on denoising auto-encoder and convolution neural network. Firstly, it performed the STFT on the radar emitter signal to obtain the time frequency image. And it used gray and threshold. It put processed image vectorization into the denoising auto-encoder. Then the denoising auto-encoder extracted the feature data. It put the reconstituted image matrix into the convolution neural network and used softmax classifier as classification and recognition. The simulation shows that the model using DAE is better than the original model and the time complexity decreases greatly. The recognition rate can reach 80% under SNR = - 6 dB. Besides, compared with the performance of traditional way, the recognition effect is obviously improved. [ABSTRACT FROM AUTHOR]

The advantage of deep neural network to automatically learn the deep characteristics of data was studied. This paper proposed a signal sorting method based on multilayer deep belief networks, in order to solve the problems of time consuming in traditional radar signal selection, feature redundancy and so on. Based on the improved algorithm of depth of stacked multilayer model, it overcame the problem of insufficient to the single model of learning ability, and deeply studied the essential features of the different signal, and fused the posterior probability of the model to make a classification decision, so as to further improve the signal recognition rate. It used this method to sort 7 different types of radar emitter signal sorting. Compared with other performance signal sorting method, the experimental results show that this method obtains better classification results, and exhibits strong learning ability to nature features, thus it verifies the effectiveness and superiority of this algorithm. [ABSTRACT FROM AUTHOR]

When the interference enters from the direction of the main-lobe of the radar, adopting the adaptive beamforming algorithm will lead to problems such as beam distortion and migration of the main-lobe. Aiming at the problem of interference entering from the radar main-lobe, this paper proposed an anti-jamming method based on the design of the transmitted waveform. When the signal encoded by Frank is transmitted, it constructed the matched filter and the blocking filter by using the orthogonality of the Frank code, and passed the received data through the two filters to obtain two signals. Then it suppressed the interference by the adaptive cancellation method, thus suppressed the main-lobe interference. Finally, simulation results show that the method of waveform design can effectively suppress interference from main-lobe entry. [ABSTRACT FROM AUTHOR]

To deal with the parameter estimation problem in multi-function radars( MFR) based on stochastic context-free grammar( SCFG) model,this paper developed a new method of MFR parameter estimation problem based on Earley parsing called EIO algorithm on the basis of traditional inside-outside( IO) algorithm and viterbi-score( VS) algorithm. This method pre-processed the intercepted radar data by combining IO algorithm with Earley parsing to handle arbitrary grammar production rules. It utilized the EIO algorithm to realize the learning of grammatical production rules probabilities and MFR parameter estimation. Theoretical analysis and simulation results show that the modified algorithm can reduce the computation complexity,memory complexity and running time while keeping the same level of estimation accuracy,validates the correctness and effectiveness of this method. [ABSTRACT FROM AUTHOR]

In order to solve the problem of complex parameter configuration in radar emitter signal recognition, this paper found a tree-based machine learning pipeline optimization method, based on the study of machine learning parameter optimization. This method used GP to generate the tree-based machine learning pipeline, and got the best machine learning pipeline with the best parameters through the evolution of the pipeline and parameters. The pipeline could include any combination of feature processing and modeling to achieve the learning and recognition of the original data set. This paper compared recognition effect of this method and the SVM method in the radar emitter signal feature sets of two different dimensions. In contrast, this method does not require cumbersome parameter configuration, and the highest accuracy is improved by more than 6%. It is proved that the tree-based machine learning pipeline obtained by this method has obvious advantages. [ABSTRACT FROM AUTHOR]

In order to observe and analyze the space of mutual electromagnetic interference from different aspects and perspectives, this paper took advantage of the characteristics of the boundaries of the isosurface to divide the radars mutual electromagnetic interference space. It measured the degree of the radars mutual electromagnetic interference by energy flux density and interference margin in the warship formation. And it used a progressive grid sampling extremum octree to reconstruct the isosurface quickly. It employed the differential display mode of the isosurface to create a real-time interactive 3D chart with a top view, a front view and a side view, while displaying the simulation results of the radars mutual interference in the warship formation dynamically. This paper simulated 6 ships with 16 single-pulse radars to interfere each other. The result shows this method is feasible and effective. It can be applied to the real-time 3D visual simulation of the electromagnetic environment of the electromagnetic interference in the large scene. It can also be used in the 3D visual simulation of the UAV radar electromagnetic interference research and prevention. [ABSTRACT FROM AUTHOR]