Vision-aided speech recognition system for a small four-legged robot

Speech Recognition is one of important means for service robots to communicate with human naturally, but the noises from various kinds of sources in real environment prevent to achieve better results. For our small four-legged robot aimed to support people´s daily life in domestic surroundings, not only the general environmental noises but also the noise from robot itself, led to very poor recognition rate even using a microphone array. In this paper, a vision aided auditory subsystem is proposed to improve the speech recognition results by suppressing the noise reception, which control a directional microphone mounted on the head towards the mouth of communicating people determined by the camera on the same head. The experiment and demonstration test indicated the efficacy of the new design.