자동차 주행 환경에서의 음성인식 시스템 개발

This paper presents a speech recognition system which works successfully in the traffic environments. To guarantee high degree of success in the various noisy situations, our approach deals with noise itself from three different standpoints : 1) An implementation of automatic capture of voice-input and its exact segmentation by removal of noise, 2) a construction of reference patterns with training data including background noise, 3) a selection of feature vector which is robust in noise. The performance of our system is verified by on-line and off-line test with isolated 33 words including 10 digits. Finally, we give a brief description of our speech recognition system capable of controlling several electric components equipped in passenger car.