The classification and recognition of human activities plays an important role in the fields of security, crime prevention, medical surveillance, etc. The swings of the limbs during the movement of the human body introduce rich micro-Doppler information. Different human activities correspond to different micro-Doppler distributions. Therefore, the discrepancies in micro-Doppler distributions can be utilized to classify different human activities. At the same time, the deep learning method based on image domain has been widely utilized in the classification and recognition of radar targets. To improve human activities classification performance, this paper proposes a human activities classification method based on complex-value convolutional neural network (CV-CNN). The motion capture data are utilized to obtain the motion trajectories of various parts of the human, and then the radarechoes of human can be got. After that, the time-frequency (TF) analysis is utilized to obtain the TF representations, and the activities sample database. A CV-CNN structure is built to train micro-Doppler characteristics are analyzed to establish a human the complex-value TF data of different human activities. The experiment results which the classification accuracy reaches 99.81% verify the effectiveness of the proposed method. The results under different training sample proportions and SNRs demonstrate the robustness of the proposed method, which is 14.22% higher than that of CNN when the ratio of training set is 10%, and 11.34% higher than that of CNN when the SNR is -5dB.

Radar echoes simulation has played a significant role in human detection and classification in the scenarios, e.g., antiterrorism, rescue after a disaster and medical, where the real-measured data are generally unavailable and limited. Therefore, a novel radar echoes simulation method of human movements is proposed based on motion capture (MOCAP) and electromagnetic (EM) calculation. First, we generate the trajectories of body segments from the true shape and MOCAP data of a human body. On the basis of that, the radar echoes are simulated by calculating the EM scattering characteristics, i.e., radar cross sections (RCSs) of all the gestures of each body segment's trajectory. Meanwhile, the micro-Doppler characteristics induced by the micromotion of human body segments are modulated in simulated radar echoes. Finally, comparisons between the simulated radar echoes and measured ones prove the validity of the proposed method. Some refinement for RCS calculation of the human body will be investigated in our future work.

In the field of target classification, detecting a ground moving target that is easily covered in clutter has been a challenge. In addition, traditional feature extraction techniques and classification methods usually rely on strong subjective factors and prior knowledge, which affect their generalization capacity. Most existing deep-learning-based methods suffer from insufficient feature learning due to the lack of data samples, which makes it difficult for the training process to converge to a steady-state. To overcome these limitations, this paper proposes a Wasserstein generative adversarial network (WGAN) sample enhancement method for ground moving target classification (GMT-WGAN). First, the micro-Doppler characteristics of ground moving targets are analyzed. Next, a WGAN is constructed to generate effective time–frequency images of ground moving targets and thereby enrich the sample database used to train the classification network. Then, image quality evaluation indexes are introduced to evaluate the generated spectrogram samples, with an aim to verify the distribution similarity of generated and real samples. Afterward, by feeding augmented samples to the deep convolutional neural networks with good generalization capacity, the classification performance of the GMT-WGAN is improved. Finally, experiments conducted on different datasets validate the effectiveness and robustness of the proposed method.

In order to address problem about the channel phase error, a channel phase calibration method based on deep learning is proposed. Using data mining to replace the traditional method can not only improve the flexibility and stability of the method, but also achieve better results. Firstly, we use the frequency response function to model the channel characteristics, and the channel mismatch model is established to simulate the errors of the channel. Secondly, the error generated by the channel is introduced into the signal to generate the analog data set. Through the training and fitting, we achieved the all-phase calibration. At the same time, a variety of different channel parameters are simulated, and the generalization ability of different channel parameters get verified. Finally, the model network is evaluated in the form of test standard deviation. According to the results, the standard deviation can be controlled within 3°, which proves the effectiveness of this method. In this paper, Octave was used to generate the simulated data set for preprocessing, PyCharm platform was used to build the neural network, and the model was trained based on TensorFlow.