Pedestrian gait classification method based on PC-IRNN

Abstract

The invention discloses a pedestrian gait classification method based on PC-IRNN (Personal Computer-International Recurrent Neural Network). According to the method, by constructing the PC-IRNN and performing gradient cutting, the problems that gradient disappearance and gradient explosion phenomena are likely to occur due to long-term dependence of a traditional RNN, input information loss is caused due to the fact that the IRNN utilizes a ReLU activation function, and the calculation cost of an LSTM is too high due to the fact that the network structure is complex are solved. Besides, the method is one of RNNs, and can directly and automatically extract features related to classification from inherent correlation between input sequences of samples, so that the step of extracting the features by predefining convolution kernels of different sizes in a DCNN is omitted, and reduction of network calculation cost and improvement of pedestrian gait classification accuracy are realized.

Description

technical field [0001] The invention belongs to the field of radar target recognition, in particular to a PC-IRNN-based pedestrian gait classification method. Can be used to classify the gait of a pedestrian walking model. Background technique [0002] The movement of pedestrians includes many common forms of micro-motion, which is a combination of various micro-movements and has a variety of postures. Therefore, the classification and recognition of pedestrian gait according to the characteristics of pedestrian micro-motions has always been a research hotspot. Micro-Doppler technology can be based on radar remote observation, not affected by weather, light and other factors, especially in dusty or foggy weather, it can still work well, and can extract highly distinguishable micro-Doppler features, Therefore, radar micro-Doppler features are more and more widely used in pedestrian gait classification and recognition. At present, deep learning is a common method for classif...

AI Technical Summary

Problems solved by technology

[0006] In order to solve the above problems, the purpose of the present invention is to propose a pedestrian gait classification method based on PC-IRNN. The PC-IRNN used in the present invention solves the gradient disappearance and gradient explosion phenomenon that traditional RNN is prone to due to long-term dependence. IRNN The problem of using the ReLU activation function to cause the loss of input information, and the disadvantage of LSTM's high computational cost due to the complex network structure

Images