Loading...

State-Space Model for Speech Enhancement in VoIP

Rahimi, Alaa | 2012

503 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 43614 (52)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Ghorshi, Mohammad Ali
  7. Abstract:
  8. Speech enhancement in noisy environments improves the quality and intelligibility of speech and reduces communication fatigue. High performance speech enhancement models are based on Bayesian estimation models, requiring estimations of the parameters of the functions that describe the likelihood and the prior distributions of the signal and noise processes. Two Bayesian speech enhancement models which are used in this thesis are Bayesian-Kalman filter and Bayesian MAP estimation.In real time applications including VoIP, in addition to additive noise, packet loss or packet delays might also occur. In real time communications the receiver terminal replaces silence for the duration of lost speech segments. On the other hand, in high quality communication systems in order to avoid quality reduction, a suitable model is required to replace the missing segments of speech. Hence, in this thesis, two models for packet loss replacement are also proposed; the first model is based on the linear prediction low order AR model and is combined with Kalman filter noise reduction model. The second proposed model is based on MAP coefficient estimation and is added to the MAP estimation noise reduction model.To improve the performance and quality of the enhancement, a final model, which includes a combination of Kalman filter for noise reduction, MAP estimation for parameter estimation of the lost segments and a low order AR model for packet loss replacement, is proposed. Performance evaluation and result comparison of the proposed models are also included.
  9. Keywords:
  10. Kalman Filters ; State Space ; Voice Over Internet Protocol (VOIP) ; Speech Enhancement

 Digital Object List

  • محتواي پايان نامه
  •   view

 Bookmark

No TOC