Waist vertical spine muscle activity identification and prediction method based on unidirectional video signals

Abstract

The invention relates to the technical field of ergonomics, in particular to a waist vertical spine muscle activity recognition and prediction method based on unidirectional video signals. The methodat least comprises the following steps: a measurement step firstly includes a signal acquisition sub-step, and when the signal acquisition sub-step is carried out, the electromyographic signals of a right trapezius muscle and longissimus lumbar muscles on two sides and motion information of key points of main parts of the whole body are synchronously acquired; a two-step clustering identificationstep comprises a pre-clustering sub-step and a quasi-clustering sub-step, and also comprises a step of constructing a clustering feature tree so as to divide the clustering feature tree into a plurality of sub-clusters; and a prediction step is based on the artificial neural network, multi-person comparison is carried out on signals for a long time. A part of target actions are recognized in the complex and long-time labor process, and on the basis of action recognition, muscle power generation percentage prediction can be conducted on tasks trained in a laboratory.

Description

technical field [0001] The invention relates to the technical field of ergonomics, in particular to a system for recognizing and predicting the activity of the erector spinae of the waist. Background technique [0002] There are following defects in the prior art. [0003] Because the conditions of electromyography testing require the contact interface to be stable, factors such as sweating and skin stretching will cause drastic changes in the results, and under working conditions, it is difficult to carry out large-scale electromyography (emg) acquisition for in-depth research on the causes of injuries and diseases Still restricted. The main reason is that the monitoring of human actions in the working environment faces a variety of possibilities, and the result of direct measurement takes too long, and many actions need to be eliminated based on classification and then identified. [0004] Also because it is difficult to simultaneously track EMG data and 3D motion data. ...

AI Technical Summary

Problems solved by technology

Modeling approaches are too complex to account for the influence of interacting factors such as subjective habits or subtle pain sensations (Butler et al., 2010)

Images