Arm muscular effort estimation from images using Computer Vision and Machine Learning

TítuloArm muscular effort estimation from images using Computer Vision and Machine Learning
Publication TypeConference Paper
Year of Publication2015
AuthorsAbraham L, Bromberg F, Forradellas R
Conference Name1st International Conference on Ambient Intelligence for Health
Date PublishedDec
Conference LocationPuerto Varas, Chile
Palabras clavecolor histograms, LBP, muscle arm effort, random forests, SVM
Abstract

A problem of great interest in disciplines like occupational medicine, ergonomics, and sports, is the measurement of biomechanical variables involved in human movement and balance such as internal muscle forces and joint torques. This problem  is solved by a two-step process: data capturing using impractical, intrusive and expensive devices that is then used as input in  complex models for obtaining the biomechanical variables of interest. In this work we present a first step towards capturing input  data through a more automated, non-intrusive and economic process, specifically weight held by an arm subject to isometric contraction as a measure of muscular effort. We do so, by processing RGB images of the arm with computer vision (Local Binary Patterns and Color Histograms) and estimating the effort with machine learning algorithms (SVM and Random  Forests). In the best case we obtained an FMeasure= 70.68%, an Accuracy= 71.66% and a mean absolute error in the predicted weights of 554.16 grs (over 3 possible levels of effort). Considering the difficulty of the problem , it is enlightening to achieve over random results indicating that, despite the simplicity of the approach, it is possible to extract meaningful information for the  predictive task. Moreover, the simplicity of the approach suggests many lines of further improvements: on the image capturing  side with other kind of images; on the feature extraction side with more sophisticated algorithms and features; and on the knowledge extraction side with more sophisticated learning algorithms. 

Miembros del DHARMa que son autores:: 
Peer reviewed?: 
1
Internacional?: 
1