A New Feature for Speech Segments Extraction with Hidden Markov Models

숨은마코프모형을 이용하는 음성구간 추출을 위한 특징벡터

  • Published : 2008.03.30


In this paper we propose a new feature, average power, for speech segments extraction with hidden Markov models, which is based on mel frequencies of speech signals. The average power is compared with the mel frequency cepstral coefficients, MFCC, and the power coefficient. To compare performances of three types of features, speech data are collected for words with explosives which are generally known hard to be detected. Experiments show that the average power is more accurate and efficient than MFCC and the power coefficient for speech segments extraction in environments with various levels of noise.


  1. Abdulla, W. H. (2002). HMM-based techiques for speech segments extraction. Scientific Programming, 10, 221-239
  2. Abdulla, W. H. and Kasabov, N. K. (1999). Two pass hidden Markov model for speech recognition systems. In Proceeding of the ICICS'99
  3. Acero, A., Crespo, C., Torre, C. de la and Torrecilla, J. C. (1993). Robust HMM-based endpoint detector, In Proceeding of the EuroSpeech, 3, 1551-1554
  4. Ganchev T., Fakotakis N. and Kokkinakis G. (2005). Comparative evaluation of various MFCC implementations on the speaker verification task. In Proceeding of the 10th International Conference on Speech and Computer, SPECOM 2005, 1, 191-194
  5. Haeb-Umbach, R. (1999). Investigations on inter-speaker variability in the feature space. In Proceeding of the IEEE ICASSP'99, 1, 397-400
  6. Rabiner, L. R. and Juang, B. H. (1993). Fundamentals of Speech Recognition. Prentice Hall PTR, New Jersey
  7. Seok, J. W. and Bae, K. S. (1999). Endpoint detection of speech signal using wavelet transform. The Journal of the Acoustical Society of Korea, 18, 57-63