Robust speech recognition in additive and channel noise environments using GMM and EM algorithm

• Published in: 2004 IEEE International Conference on Acoustics, Speech and Signal Processing Vol. 1; pp. I - 941
• Main Authors: Fujimoto, M., Riki, Y.A.
• Format: Conference Proceeding
• Language: English; Japanese
• Published: Piscataway, N.J IEEE 28.09.2004
• Subjects: Acoustic noise; Additive noise; Applied sciences; Detection, estimation, filtering, equalization, prediction; Exact sciences and technology; Information, signal and communications theory; Microphones; Noise robustness; Signal and communications theory; Signal processing; Signal, noise; Speech analysis; Speech enhancement; Speech processing; Speech recognition; Systems, networks and services of telecommunications; Telecommunications; Telecommunications and information theory; Testing; Transfer functions; Transmission and modulation (techniques and equipments); Working environment noise; Additive noise; Parameter estimation; Microphone; Acoustic signal; Mixture theory; Background noise; Mismatching; Characteristic function; Discrimination; Gaussian process; Vocal signal; Channel estimation; Speech recognition; Noise immunity; Transfer function; EM algorithm; Speech processing
• ISBN: 9780780384842; 0780384849
• ISSN: 1520-6149
• DOI: 10.1109/ICASSP.2004.1326142

In this paper, we evaluated the speech recognition in real driving car environments by using a GMM based speech estimation method and an EM algorithm based channel noise estimation method. The GMM based speech estimation method proposed by Segura et al (2001) was not robust for channel noise such as an acoustic transfer function, a microphone characteristic and so on. To cope with this problem, we propose a channel noise estimation method based on the EM algorithm. Furthermore, we estimate the speech signal more accurately by using a speech GMM and a silence GMM instead of the GMM trained without speech/silence discrimination. Our proposed method has been evaluated on the AURORA3 tasks. In the evaluation results, the proposed method showed the significant improvement in the high-mismatched condition test of AURORA3 tasks.