빔포밍 성능 향상을 위한 여기신호 상호 상관도를 이용한 통계적 목적신호 검출

Abstract

Target signal detection(TSD) is important because it directly affects the performance of generalized sidelobe canceller(GSC) which is widely used for noise reduction. The main problem is that the fluctuation of the non-stationary noise causes it to decrease that the accuracy of the adaptation mode control(AMC) of GSC which decides adaptive filter update. Thus, we propose the speech absence probability (SAP)-based TSD. It is statistical modeling of speech and noise using cross correlation of LP residual of two inputs It is validated that the proposed algorithm achieves the improvement of speech recognition rate.