Adaptive Noise Cancelation in Speech Signals

This thesis deals with adaptive algorithms and their application in echo cancelation. It tries to tackle the ever-lasting problem of computational complexity vs. rate of convergence. It describes some well-known adaptive methods as well as a novel concept which has been entitled Optimal Step-Size (OSS). In the first part of the thesis, two common methods, the Normalized Least Mean Squares (NLMS) and the Recursive Least Mean Squares (RLS), are described and their properties analyzed. It is shown what deficiencies they have when applied for echo cancelation. In the second part of the thesis it is shown that, by modifying the NLMS and the RLS, better results can be achieved. In the last part of the thesis a novel approach is described, called OSS. The proposed method possesses certain features that make it robust in non-stationary Environment. Experimental evaluations show that its performance is comparable with the conventional methods and in certain cases it outperforms them.