Multi-mode human body behavior identification method based on sports biomechanics

Abstract

The invention discloses a multi-mode human behavior recognition method based on sports biomechanics. On the basis of the exercise biomechanics, the human body structure is converted into a human bodyskeleton action model with exercise characteristic parameters, action description parameters are determined, and then the biomechanics characteristics are calculated. Describing human body action postures by using the characteristic parameters, and establishing a relation model between the parameters and the action modes; The human body action posture description method based on biomechanics comprises the following steps: analyzing the characteristics of human body action time sequence data, performing boundary detection on continuous actions in multiple modes by using a key frame extraction method based on combination of a curve simplification algorithm and a frame subtraction algorithm, and realizing segmentation and automatic labeling of basic actions and transition actions; And analyzing characteristics of the basic action mode and the transition action mode, and establishing an identification model of the basic action mode and the transition action mode by adopting a multi-model idea to realize identification of multi-mode behaviors.

Description

technical field [0001] The invention belongs to the technical field of human behavior recognition, and in particular relates to a multi-mode human behavior recognition method based on sports biomechanics. Background technique [0002] Sports biomechanics technology has been widely concerned and applied in the fields of ergonomics, sports training, medical diagnosis and rehabilitation training, but it is rarely used in the research of human daily behavior recognition. Existing research mainly conducts action recognition based on discrete features such as time domain and frequency domain of sensor data, and pays less attention to the characteristics of human actions in space, time, and force effects. The expression of actions is not sufficient and the classification is not detailed enough. Therefore, for sensor-based continuous multimodal behavior recognition, there are still many problems that need to be refined and studied in depth. Contents of the invention [0003] The ...

AI Technical Summary

Problems solved by technology

[0002] Sports biomechanics technology has been widely concerned and applied in the fields of ergonomics, sports training, medical diagnosis and rehabilitation training, but it is rarely used in the research of human daily behavior recognition.

Existing research mainly conducts action recognition based on discrete features such as time domain and frequency domain of sensor data, and pays less attention to the characteristics of human actions in space, time, and force effects. The expression of actions is not sufficient and the classification is not detailed enough.

Images