Human motion state discrimination method based on densely connected convolutional neural network

Abstract

The invention provides a gait data acquisition mode based on multi-dimensional information fusion and a motion state discrimination method based on a densely connected convolutional neural network. Inthe gait data acquisition process of multi-dimensional information fusion, acceleration and angular velocity information and plantar pressure information of a left shank (LS), a right shank (RS) anda waist (L5) of an experimental subject during straight going, left turning, right turning and stair climbing are measured at the same time, the operability is high, and the complexity is low. According to the densely connected convolutional neural network for the original gait sequence, feature extraction is not needed any more, priori knowledge is not needed, meanwhile, feature selection is matched, the number of needed matched measurement devices and the number of needed features are reduced, and the complexity of a network model is reduced. And the input of the next layer is formed by connecting the input feature map and the output feature map of the convolution layer of the middle part, so that the accuracy of the network is obviously improved. The accuracy of motion state judgment (straight going, left turning and right turning) of single-person multi-tasks (straight going, left turning and right turning) on the self-collection data set can reach 99.1%, and compared with 91.79% of an SVM, the accuracy is improved to some extent.

Description

technical field [0001] The invention provides a human body motion state discrimination method based on a densely connected convolutional neural network, which provides a multi-dimensional information fusion gait information collection method, and provides a new analysis method for motion state discrimination oriented to gait information, belonging to Human gait recognition and pattern recognition field. Background technique [0002] The human gait recognition technology aims to analyze the gait acceleration data of the moving object and realize the qualitative judgment of the moving object's gait. Previous Gait Analysis Based on Video or Image Sequence (Ben Yuye, Xu Sen, Wang Kejun. A Review of Pedestrian Gait Feature Expression and Recognition. Pattern Recognition and Artificial Intelligence, 2012,25(1):71-81.(BEN X Y ,XU S,WANG K J,et al.Review on Pedestrian Gait Feature Expression and Recognition.PatternRecognition and Artificial Intelligence,2012,25(1):71-81.)), vulnera...

AI Technical Summary

Problems solved by technology

Previous Gait Analysis Based on Video or Image Sequence (Ben Yuye, Xu Sen, Wang Kejun. A Review of Pedestrian Gait Feature Expression and Recognition. Pattern Recognition and Artificial Intelligence, 2012,25(1):71-81.(BEN X Y ,XU S,WANG K J,et al.Review on Pedestrian Gait Feature Expression and Recognition.PatternRecognition and Artificial Intelligence,2012,25(1):71-81.)), vulnerable to scene lighting changes, moving target occlusion, etc. Influenced by factors, it is not conducive to subsequent machine learning

Images