Human body motion state classification method based on improved convolutional neural network

Abstract

The invention firstly simplifies a human body target motion into a compound motion of a typical motion form, and puts forward a human body walking and running empirical model according to measured biomechanical experiment data. On the basis of an XYZ convention, an Euler rotation matrix is used for establishing and perfecting human body motion animation models under a state of marching on the spotand a walking and running state; then, on the basis of a continuous wave signal, an integral radar echo model of a human body walking and running target and a radar echo model of each position of thetarget are constructed; and a Matlab simulation tool is used for analyzing radar echo characteristics under different human body motion states and generating a corresponding time domain graph throughthe short-time Fourier transform so as to lay a theoretical foundation for the moving human body target characteristic extraction and the walking and running state of the human body. The time domaingraph of the human body motion is divided into a training set, a verification set and a testing set, the improved convolutional neural network is input for training, the network parameter is regulatedto realize model convergence, and the network can correctly classify the human body motion state.

Description

technical field [0001] The invention belongs to the technical field of human body motion state classification based on radar detection, and in particular relates to a human body motion state classification method based on an improved convolutional neural network. Background technique [0002] Human target detection has important applications in anti-terrorism, post-disaster search and rescue and other occasions. Radar-based human motion state classification technology has become a hot spot in recent years. For the human body in motion, the life detection radar uses the micro-Doppler modulation of the radar transmission signal through the swing of the human trunk and limbs, extracts the micro-Doppler characteristics of the echo signal, and realizes the detection of the human body and the identification of the movement state. The modulation process of motion on the echo signal is a non-stationary process, and the limb swing amplitude is much larger than the vibration amplitude...

AI Technical Summary

Problems solved by technology

However, the conventional convolutional neural network does not optimize the network design for the characteristics of human motion state, so it requires more data set training and more network layers, which easily leads to over-fitting of the network.

Images