M.S. (Master of Science)
Department of Electrical Engineering
Teleconferencing--Equipment and supplies; Noise control--Equipment and supplies; Echo; Adaptive filters
The objective of this thesis is to develop new algorithms for line echo cancellation, acoustic echo cancellation, howling suppression and speech detection. The gradient lattice-ladder algorithm is shown to have superior convergence feature to the least mean square (LMS)-type algorithms when applied as line echo cancellers. In order to model the long-impulse-response acoustic echo and to avoid a high order requirement of a finite impulse response (FIR) filter, an infinite impulse response (IIR) lattice algorithm based on a pole-zero structure is presented to cancel the acoustic echo with fast convergence, reduced computation as well as guaranteed stability. The new algorithm provides a one-dimensional lattice structure of the autoregressive moving average (ARMA) model and thus obviates the matrix manipulations of conventional ARMA structures. A new howling controller based on a constrained AR model is developed that has fast convergence rate, time and order recursive property. The model is formulated by constructing a sequence of orthogonal polynomials. Based on this method, a class of HR notch filters with constrained poles and zeros are developed to efficiently remove the howling components from the speech signal. An orthogonality-based speech detector is presented to prevent the interference of double-talk condition and track the variation of echo paths. The detector examines the convergence of the echo cancellers instead of the explicit communication modes. Extensive computer simulations as well as real-time experimentations on the ADSP 21020 are carried out to examine the presented algorithms.
Liu, Lei, "A novel adaptive notch filter for howling control and an IIR-lattice algorithm for acoustic echo cancellation in teleconferencing applications" (1994). Graduate Research Theses & Dissertations. 444.
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.