적응형 문턱값을 가지는 2차 조건 사후 최대 확률을 이용한 통계적 모델 기반의 음성 검출기 = Statistical Model-Based Voice Activity Detection Using the Second-Order Conditional Maximum a Posteriori Criterion with Adapted Threshold

국문 초록 (Abstract)

본 논문에서는 음성의 통계적 모델에 기반한 음성 검출기 (voice activity detection, VAD)의 성능 향상을 위해 2차 조건 사후 최대 확률 (second-order conditional maximum a posteriori, second-order CMAP)기법을 적용한 우도비 테스트 (likelihood ratio test, LRT)를 제안한다. 제안된 알고리즘은, 기존의 통계적 모델에 기반한 음성 검출기와 CMAP 기반의 음성 검출기를 분석한 다음, 직전 2 프레임에서 음성의 존재와 부재에 대한 조건부 확률에 따라 실시간으로 적응형 문턱값을 구하여 기하 평균한 우도비와 비교하는 음성검출 결정법 (decision rule)을 제시한다. 제안된 알고리즘을 비정상 (non-stationary) 잡음환경에서 기존의 통계적 모델에 기반한 음성 검출기, CMAP 기반의 음성 검출기와 비교하였으며, 향상된 성능을 보였다.

다국어 초록 (Multilingual Abstract)

In this paper, we propose a novel approach to improve the performance of a statistical model-based voice activity detection (VAD) which is based on the second-order conditional maximum a posteriori (CMAP). In our approach, the VAD decision rule is expressed as the geometric mean of likelihood ratios (LRs) based on adapted threshold according to the speech presence probability conditioned on both the current observation and the speech activity decisions in the pervious two frames. Experimental results show that the proposed approach yields better results compared to the statistical model-based and the CMAP-based VAD using the LR test.

참고문헌 (Reference)

1 J.-H. Chang, "Voice activity detector employing generalised gaussian distribution" 40 (40): 1561-1563, 2004

2 K. Srinivasant, "Voice activity detection for cellular networks" 85-86, 1993

3 J.-H. Chang, "Voice activity detection based on multiple statistical models" 54 (54): 1965-1976, 2006

4 J. W. Shin, "Voice activity detection based on conditional MAP criterion" 15 : 257-260, 2008

5 J. Ramirez, "Statistical voice activity detection using a multiple observation likelihood ratio test" 12 (12): 689-692, 2005

6 Y. C. Lee, "Statistical model-based VAD algorithm with Wavelet Transform" E89-A (E89-A): 1594-1600, 2006

7 J. Ramirez, "Speech/non-speech discrimination based on contextual information integrated bispectrum LRT" 13 (13): 497-500, 2006

8 Y. Ephraim, "Speech enhancement using a minimum mean-square error short-time spectral amplitude estimator" ASSP-32 (ASSP-32): 1190-1121, 1984

9 N. S. Kim, "Spectral enhancement based on global soft decision" 7 (7): 108-110, 2000

10 L. R. Rabiner, "Fundermentals of Speech Recognition" Prentice-Hall 1993

11 Y. D. Cho, "Analysis and improvement of a statistical model-based voice activity detector" 8 (8): 276-278, 2001

12 J. Sohn, "A voice activity detector employing soft decision based noise spectrum adaptation" 365-368, 1998

13 J. Sohn, "A statistical model- based voice activity detection" 6 (6): 1-3, 1999