DISTANT TALKER IDENTIFICATION BY NONLINEAR PROGRAMMING AND BEAMFORMING IN SERVICE ROBOTICS

Abstract

In service robotics speech interaction is assuming increasing interest. In this contest, in order to perform properly their tasks, a robot needs to identify the persons who is talking to. In this paper an algorithm for distant talker identification is presented: it is based on beamforming in the frequency domain to reduce directional noise and nonlinear programming for detecting the identity of the speaker. The results obtained show good identification performances in real in-door environment.