Reduction of Non-stationary Noise for a Robotic Living Assistant using Sparse Non-negative Matrix Factorization

Abstract

Due to the demographic changes, support by means of assistive systems will become inevitable for home care and in nursing homes. Robot systems are promising solutions but their value has to be acknowledged by the patients and the care personnel. Natural and intuitive human-machine interfaces are an essential feature to achieve acceptance of the users. Therefore, automatic speech recognition (ASR) is a promising modality for such assistive devices. However, noises produced during movement of robots can degrade the ASR performances. This work focuses on noise reduction by a non-negative matrix factorization (NMF) approach to efficiently suppress non stationary noise produced by the sensors of an assisting robot system. 1