Human body skeleton motion sequence behavior identification method

Abstract

The invention discloses a human body skeleton motion sequence behavior identification method. The method includes acquiring human body skeleton node coordinates; serially connecting the nodes corresponding to the limbs of the human body skeleton to form the motion characteristic expressions of the limbs and the trunk; serially connecting the motion characteristic expressions of the limbs and the trunk to form the vector expression of the human body skeleton; arranging the vector expressions corresponding to the frames in the human body skeleton motion sequence according to time sequence to obtain a three dimensional matrix; performing the normalization of the values in the three-dimensional matrix and the normalization of the dimensions to obtain the image expression corresponding to the human body skeleton motion sequence; adaptively extracting the texture characteristic expression in the image expression by means of a convolutional neural network; and performing behavior class determination based on the texture characteristic expression and determining the behavior class of the human body skeleton motion sequence by voting. The method can accurately identify people's behaviors based on the human body skeleton motion sequence without the need for complex data preprocessing.

Description

technical field [0001] Embodiments of the present invention relate to the technical fields of computer vision, pattern recognition and deep learning, and in particular to a method for recognizing human skeleton movement sequence behaviors. Background technique [0002] The revival of neural network theory has promoted the rapid development of artificial intelligence technology. In today's society, intelligent robots and driverless cars are about to enter people's lives. Intelligent transportation, intelligent video surveillance and smart cities all require computers to automatically analyze human behavior. At present, depth camera technology combined with high-precision human skeleton estimation algorithm can directly provide the skeleton sequence corresponding to the human body movement process, and based on the skeleton sequence, human behavior can be accurately recognized. [0003] The traditional behavior recognition algorithm based on the human skeleton sequence is ma...

AI Technical Summary

Problems solved by technology

[0003] The traditional behavior recognition algorithm based on the human skeleton sequence is mainly to design a classifier after coding on the basis of manual feature extraction to realize behavior classification. The manual feature extraction process is relatively cumbersome, and it is usually separated from the subsequent feature coding and classification process. Although it can be cascaded to form a system, it is not conducive to practical application due to low efficiency

In addition, the training and testing of traditional methods are usually carried out on small data sets. When the amount of data increases, the computational complexity of the model is unbearable for general hardware conditions, and it is difficult to play a role in practical applications.

Images