On System Identification and Acoustic Echo Cancellation

Sammanfattning: The topic of acoustic echo cancellation has received a lot of interest over the years. Even though the topic is relatively old, the increasing demand for hands-free telephony has made it more important than ever before. Furthermore, in recent years new applications such as Internet Protocol (IP) telephony and stereophonic hands-free telephony have emerged that also require acoustic echo cancellation in order to function properly.This thesis is partly about acoustic echo cancellation which has the purpose to remove the acoustic echoes from the loudspeaker sound signals picked up by the microphones in hands-free telephony systems. If the attenuation of the echoes is small, as it is in a hands-free telephony setup, good acoustic echo cancellation is required for the setup to work well. Several methods are presented in this thesis, addressing various acoustic echo cancellation areas such as doubletalk detection, stereophonic acoustic echo cancellation and ordinary acoustic echo cancellation. While some of the results are basically extensions of existing acoustic echo cancellation algorithms, others are more innovative in the sense that they offer solutions to previously unsolved problems or outperform by far existing algorithms.The second part of this thesis is about system identification which is also a relatively old, but still very active topic. System identification deals with the problem of building mathematical models of dynamic systems and its applications are manifold. One particular area where system identification is needed is the aforementioned acoustic echo cancellation application, where models of the acoustic paths between the loudspeakers and the microphones need to be determined. This thesis presents some explanations and additions to two well-known system identification algorithms. Furthermore, it provides new solutions to two previously unsolved system identification problems.