University of Southampton
Adaptive IIR filtering for acoustic echo cancellation on a mobile handset
Abstract
dc:description.abstract<p>This thesis is concerned with the problem of acoustic echo cancellation in mobile handsets. The acoustic echo path in a mobile handset is due to the acoustic coupling between the loudspeaker and microphone, which changes depending on the particular handset design and on the handset orientation. To cancel this echo effectively, the nature of this acoustic echo path in normal use must be fully understood. This thesis identifies the possible echo sources on a mobile handset and reports on echo path measurements taken from a typical handset design in various handset orientations. It has been found that resonant acoustic echo path responses are obtained in normal handset use.</p> <p>The resonant nature of the echo path response motivates an investigation of IIR filter models, as well as more traditional FIR models of the echo path. From the reported results of offline modelling experiments it is clear that, not only do IIR filter models give benefits in terms of complexity and performance over FIR models, but that the HR filter model also needs to be adaptive.</p> <p>The modelling performances of adaptive FIR and both Equation Error and Output Error adaptive IIR algorithms have been investigated in this thesis. The steady state performance of certain Output Error adaptive IIR algorithms has been shown to be superior to equivalent adaptive FIR algorithms, both in the presence and the absence of microphone disturbance noise. The tracking performance of these Output Error adaptive IIR algorithms for different time variations in the echo path response, also show that these algorithms can also have better tracking performance than equivalent adaptive FIR algorithms.</p> <p>In the handset acoustic echo cancellation application robust operation for input speech signals is necessary at low echo to microphone noise levels. A modified form of the NLMS Newton Simplified Gradient Output Error adaptive IIR algorithm is developed in this thesis to satisfy these requirements.</p>
Degree
thesis:*- Name dc:type.qualificationname
- Ph.D.
- Level dc:type.qualificationlevel
- doctoral
- Grantor dc:publisher.institution
- University of Southampton
- Year dc:date.issued
- 2004
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Craney, Edward Peter
Rights
- Language dc:language
- English, eng