M.S. (Master of Science)
Department of Electrical Engineering
Speech processing systems; Electric filters
The main purpose of this thesis is to compare two different filtering techniques. We have compared the performance of the Wiener filter and the Kalman filter using some constraints. A sample of noisy speech is taken frame by frame and tested with both the filtering techniques. The Linear Prediction Coefficients (LPC) of speech samples of each noisy speech frame are converted into Line Spectral Pairs (LSP ) and then inter frame and intra frame smoothing is performed on them. They are then converted back to the LPC and these LPCs are used in Wiener Filtering and Kalman filtering. The performance of both the algorithms was compared at different signal to noise ratio (SNR). We see that the Kalman filter gives much better results as compared to Wiener filtering at a very low SNR Thus we can say that the Kalman filter has lots of advantages over Wiener filtering. For example, we don’t need to use a Voice Activity Detector in Kalman filtering. We require less number of iterations and also it gives a better performance at low SNR. In the Wiener filtering technique, we need to assume that the speech and noise are wide sense stationary (WSS) but for Kalman filtering we don’t need to worry about the variance of the signal it may be fast changing. Thus we can say that there is a considerable amount of difference between the Wiener filtering and the Kalman filtering techniques.
Shah, Chinmay, "Comparison of speech enhancement techniques" (1997). Graduate Research Theses & Dissertations. 1875.
vi, 67 pages
Northern Illinois University
Rights Statement 2
NIU theses are protected by copyright. They may be viewed from Huskie Commons for any purpose, but reproduction or distribution in any format is prohibited without the written permission of the authors.