Browsing by Author "Hsieh, SF"

Now showing 1 - 5 of 5

A Double-talk resistant echo cancellation based on iterative maximal-length correlation
(2000)
Jenq, JC
;
Hsieh, SF
;
電信工程研究所
;
Institute of Communications Engineering
A novel acoustic echo canceller for speakerphone system based upon Iterative Maximal Length Correlation (IMLC) algorithm is proposed. This algorithm is robust to double talk because of the nature of the ML sequence; it also iteratively reduces the far-end speech's effects. This simple structure was found to perform very well. We also perform a rigorous convergence analysis and derive the lower bound for the coefficients error. Computer simulations also confirm our theoretical results.
Acoustic echo cancellation using iterative-maximal-length correlation and double-talk detection
(2001-11-01)
Jenq, JC
;
Hsieh, SF
;
電信工程研究所
;
Institute of Communications Engineering
The conventional maximal-length-correlation (MLC) algorithm to estimate room impulse response for adaptive echo cancellation (AEC) is disturbed by both far-end and near-end speeches. In this paper, a new iterative-maximal-length-correlation (IMLC) algorithm is proposed to reduce the far-end speech interference. To avoid the near-end interference, a new double-talk detection (DTD) method is proposed by tracking the squared coefficients errors of the AEC filter. This DTD method has well-separated detection margins among single-talk (ST), double-talk (DT), and echo path changes. Statistical analysis and computer simulations confirm that our proposed IMLC-DTD algorithm outperforms conventional methods.
Common acoustical-poles/zeros modeling for 3D sound processing
(2000)
Chen, MC
;
Hsieh, SF
;
電信工程研究所
;
Institute of Communications Engineering
HRTFs(Head Related Transfer Functions) are a set of transfer functions from different spatial locations to both ears. They constitute an important database in 3-D sound signal processing. However, such an enormous data set makes real-time implementations difficult. Inspired by the resonant characteristic of pinnae, we seek a common-pole/zero model to reduce this vast data set. A Shanks common-poles model is proposed to improve previous Prony work. Based on a linearized least-squares criterion, a novel approach using iterative prefiltering and recursive updating the common poles is proposed and outperforms previous works.
Common-acoustic-poles/zeros approximation of head-related transfer functions
(2001)
Liu, CJ
;
Hsieh, SF
;
電信工程研究所
;
Institute of Communications Engineering
Common-acoustic-poles/zeros (CAPZ) approximation is a more efficient way to model head-related transfer functions (HRTF's). In CAPZ approximation, a group of HRTF's can share a set of poles but use their own zeros. Previous CAPZ works, such as the Prony, Shanks and iterative prefiltering methods, were all based on the linearized least-square criterion. A new state-space approach, jointly balanced model truncation, is proposed by using singular value decomposition of a joint Hankel matrix. The proposed approach can choose the suitable order of IIR filters for HRTF's approximation according to the distribution of singular values but previous works can't. The proposed method is also modified to permit different orders for pole and zero. Computer simulations of these approaches are included for comparison.
Decision of double-talk and time-variant echo path for acoustic echo cancelation
(2003-11-01)
Jenq, JC
;
Hsieh, SF
;
電信工程研究所
;
Institute of Communications Engineering
A new double-talk detection (DTD) method is proposed for acoustic echo cancelation (AEC) using the iterative maximal-length-correlation (IMLC) algorithm. Based on the assumption of uniformly distributed time-variant echo-path-change, we develop a simplified likelihood ratio test and plot the detection performance by a receiver operating characteristic. Computer simulation shows the proposed DTD method is effective in discriminating double-talk from time-variant echo-path-change.