M.S. (Master of Science)
Department of Electrical Engineering
Speech processing systems; Demodulation (Electronics); Fuzzy algorithms; Fuzzy systems
Speech can be modeled as short bursts of vocal energy separated by silence gaps. During a typical conversation, talk-spurts comprise only 35% of each party's speech while the remaining 65% are silence. Wireless communication systems can achieve significant gains in spectral and energy efficiency by disconnecting the user from the spectral resource during the silence periods. Therefore, a simple and robust voice activity detector is critical for a wireless system's efficiency and speech quality. This thesis presents a new voice activity detection algorithm based on fuzzy logic. The fuzzy soft decisions are made based on four parameter values: energy level, spectral distribution, periodicity, and stationarity. The input signal to the proposed fuzzy voice activity detection algorithm is subdivided into 15-msec frames. Each frame is processed and measured for relative energy content, percentage of signal energy below 1000 Hz, and the periodicity. Stationarity between adjacent frames is also measured as a continuous variable using the spectral covariance method. All four parameter values are extracted from every speech frame based on conventional speech processing techniques, and then act as fuzzy input variables. The fuzzy rule base, on which the fuzzy inference phase is based, consists of 16 empirically designed fuzzy rules. The thesis concludes with both objective and subjective test results illustrating the performance of the proposed fuzzy voice activity detection algorithm in office, portable and mobile environments. Both objective and subjective tests show that the proposed voiced activity detection algorithm is practical for mobile and portable wireless systems. The fuzzy algorithm's simplicity also augurs its attractiveness for real-time implementation using a portion of a single DSP chipset.
Zhou, Tao, "Fuzzy voice activity detection based on speech feature extraction techniques" (2000). Graduate Research Theses & Dissertations. 3232.
x, 99 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.
Includes bibliographical references (pages -99)