SERVICE ROBOT SPEECH ENHANCEMENT METHOD USING ACOUSTIC MICRO-SENSOR ARRAY

摘要

Good interactive voice response and perception function is the important technology of intelligent service robot. Service robot works under noisy and multiple-path reflection environment generally. Speech enhancement technology is applied to increase the robustness and intelligence of input signal. Conventional single channel speech enhancement algorithms always improve the Signal-Noise-Ratio of target signal at the price of inevitable speech distortion. In this paper, we introduced an acoustic micro-sensor (microphone) array system as a frontend device to enhance desired signal polluted by diffuse noise. A Modified General Sidelobe Canceller (GSC) algorithm is proposed to achieve full-band stable gain and directional beam response from given angle range. An adaptive post-filter approach will further improve the performance of whole system. Experiments based on real data show that this method is efficient to improve the robustness and intelligence of service robot's interactive voice function.