Variable step-size adaptive algorithms for acoustic echo cancellation in hands-free portable devices

It appears your Web browser is not configured to display PDF files. Download adobe Acrobat or click here to download the PDF file.

Click here to download the PDF file.


Eldeeb, Rania




Hands-free communication is becoming more ubiquitous in many of today’s practical devices such as mobile phones, tablets, teleconferencing systems, Voice over Internet Protocol (VoIP) conferencing systems, hearing aids, Bluetooth® accessories, and hands-free car kits. Therefore there is an increased demand for more robust acoustic echo cancellation and noise cancellation to provide higher quality conversations. Acoustic echo cancellers use adaptive filters to identify the acoustic echo path impulse response. Non-stationarity of the acoustic environment and long lengths of the echo paths impose challenges on the performance of adaptive algorithms.

This thesis proposes new techniques to improve the performance of adaptive algorithms in the context of acoustic echo cancellation by varying the step-size parameter in different ways. Two techniques are proposed and incorporated into existing variable step-size Normalized LeastMean-Square (NLMS) based algorithms: the gradient-induced technique and the thresholding technique. In the gradient-induced technique, an estimate of the gradient of the coefficients’ adaptation is incorporated into the filter coefficients’ update rule. It is shown that the proposed technique improves the overall performance of selected Proportionate NLMS (PNLMS) based algorithms with a modest increase in complexity. It also allows algorithms to achieve comparable performance with less overall complexity when compared to more complex algorithms. The thresholding technique is incorporated into the NLMS algorithm and the NonParametric Variable Step-Size NLMS (NPVSS-NLMS) algorithm forming new classes of thresholded NLMS-based algorithms. Simulations show that the proposed thresholded algorithms outperform their original counterparts in terms of convergence speed and tracking capability with a modest increase in computations.


Electrical engineering.
Artificial intelligence.
System theory.




Carleton University

Thesis Degree Name: 

Master of Applied Science: 

Thesis Degree Level: 


Thesis Degree Discipline: 

Engineering, Electrical and Computer

Parent Collection: 

Theses and Dissertations

Items in CURVE are protected by copyright, with all rights reserved, unless otherwise indicated. They are made available with permission from the author(s).