Findings about selecting body parts to analyze human activities through skeletal tracking joint oriented devices

Abstract

Analyzing activities (either static postures or movements) made by a user is a complex process that can be done through a wide range of approaches. One part of these existing approaches support doing the recognition focusing their analysis on specific body parts. In fact, in previous publications a method was introduced for activity recognition (Body-Angles Algorithm) capable of analysing only using a single sample of those activitites and allowing the selection for each activity which are the relevant joints. But being able to analyse the body of the user selecting only a subset of the same, has both advantages and disadvantages. Therefore throughout this article we will expose those disadvantages, the applied solution to mitigate them and the results of an evaluation destined to clear which body parts make it easier to obtain high accuracy rates in recognition. Through this work we aim to give the scientific community lessons learned about the usage of different body areas in the analysis of activity recognition.