Human identification using micro-motion and LPCP-Cov neural networks

A method of human identification using micro-doppler data based on a lightweight neural network is proposed. It can identity various gaits and different person effectively, and the amount of computation and the number of parameters of the model can be reduced also. The principle like that used in linear phase finite impulse response filter in digital signal processing is utilized here. The proposed LPCP-Conv kernel can lower the storage cost and the computational complexity of the model. The results of experiments show that the proposed algorithm can decrease the quality of computation and the number of parameters effectively, as well as improving the accuracy. The LPCP-Conv kernel is friendly to hardware accelerator with low power design for reducing the cost of memory access.