Research On The Theory And Method Of Human Motion Recognition For The Carrier-Free UWB Radar

As the size of the radar hardware platform becomes smaller and smaller,the cost becomes lower and lower.The application of indoor radar-based human motion recognition has become a reality,which can be realized in a low-cost device with simple architecture.The main advantage of the carrier-free UWB radar human motion recognition system is that the carrier-free UWB radar has extremely high resolution,can capture the finer movement changes of the human motion,and has strong anti-multipath fading and anti-interference against the indoor complex environment.In this paper,different types of human motion echo signals are captured based on the carrier-free UWB radar.The human motion echo signals are analyzed by the traditional statistical methods and transform domain methods,the main features of the echo signals are extracted and classified.The specific research work is as follows:1.Firstly,the human motion echo signals of the carrier-free UWB radar is analyzed,and the human motion of the carrier-free UWB radar is constructed.A novel human motion feature extraction method based on the principal component analysis(PCA)and discrete cosine transform(DCT)is proposed.The classification and recognition of the ten different types of human motion were carried out.The experimental results show that the recognition rate of the data set divided into training sample sets and test sample according to 3:1 can reach 100%,and the recognition rate of the data set divided into training sample and test sample set according to 1:1 can reach 99%.2.The human motion echo feature extraction method combining PCA and DCT is used for the few-shot of human motion recognition research.Randomly select a part of the sample data as the training set and the rest of the sample data as the test set.Through the multiple experiments to average strategy as the final human motion type recognition rate.The experimental results show that the human motion feature extraction method based on the PCA and DCT also has a significant effect on the recognition of the few-shot of human motion.3.The carrier-free UWB radar human motion echo signal is a nonlinear non-stationary signal.The 2D-VMD feature extraction algorithm is used to extract the features of different types of radar-based human motion signal.The carrier-free UWB radar human motion echo signal is decomposed by the 2D-VMD algorithm,and several BIMFs are obtained,which represent the central frequency component of different parts of human motion.The experimental results show that several BIM's decomposed by the 2D-VMD algorithm can well represent the human motion echo signal characteristics of this type of human motion,and it is an effective tool for feature extraction of radar-based human motion.4.Feature engineering plays an important role in human motion recognition.A good feature extraction method can bring unexpected effects to human motion recognition.However,most research is focused on micro-Doppler research.When the micro-Doppler spectrograms of two people are similar or the movement is irregular,the final recognition accuracy will drop sharply.The feature engineering of human motion recognition of the carrier-free UWB radar based on the two-dimensional empirical mode decomposition(BEMD)is researched.Based on the measured ten different types of the human motion experiments,the experimental results show that the feature extraction by the BEMD algorithm is very effective in the field of human motion recognition of the carrier-free UWB radar.Then,the feature extraction based on the BEMD algorithm model framework and the feature extraction of human motion micro-Doppler spectrogram using the convolutional neural network is compared,we found that the BEMD algorithm based feature extraction process and the convolutional neural network under deep learning is very similar,and can further explore the mechanism of extracting features inherent in convolutional neural network according to the process of the BEMD algorithm.