Recognition of indoors activity sounds for robot-based home monitoring in assisted living environments

Abstract

In this paper we present a methodology for the recognition of indoors human activities using microphone for robotic applications on the move. In detail, a number of classification algorithms were evaluated in the task of home sound classification using real indoors conditions and different realistic setups for recordings of sounds from different locations - rooms. The evaluation results showed the ability of the methodology to be used for monitoring of home activities in real conditions with the best performing algorithm being the support vector machine classifier with accuracy equal to 94.89%.