Rapid Recognition of Human Behavior Based on Micro-Doppler Feature

In this paper, we use principal component analysis (PCA) and Mel frequency cepstrum coefficient (MFCC) to study the classification of human behavior based on micro-Doppler features. The PCA can reduce linear dimensionality and computational complexity. The MFCC can extract micro-Doppler features of different frequencies. The combination of PCA and MFCC can reduce the time which is used to extract the micro-Doppler features of human behavior and the time which is used to calculate the feature extraction. So, the combination method can realize real-time processing. We collect data from 6 differential behaviors of 12 person by using the frequency modulated continuous wave (FMCW) radar. And then, we classify human activities by using support vector machine which is extracted from PCA and the MFCC feature vectors. Experimental results show that our method has a good recognition rate.