Sound source recognition for human robot interaction

Abstract

A very important aspect in developing human-robot interaction (HRI) is the ability to recognize people by sound source recognition. In this paper, we introduce an intelligent audio human detection system that is able to recognize user’s voice, and identified it from background sound. The sound sources recognition for human robot interaction is investigated using an unsupervised learning algorithm, neighborhood linear embedding (NLE), which is able to extract the intrinsic features such as neighborhood relationships, global distributions and clustering property of a given data set. Furthermore, motivated by the scale adaptivity of human’s perception, several scale invariant metrics are designed to enhance the intrinsic feature extraction performance of NLE. Simulations on different sound sources recognition are studied to demonstrate effective applications of the scale invariant NLE algorithm for robust sound recognition and identification to improve auditory system of robot for human robot interaction.