Speech Enhanced at IWAENC 2008

BY MICHAEL L. SELTZER

From September 14-17, 2008, the 11th International Workshop on Acoustic Echo and Noise Control (IWAENC) was held at the University of Washington in Seattle, Washington, USA.  IWAENC is a workshop held every other year that focuses on speech signal processing, with a particular emphasis on issues related to hands-free speech communication. While this field (and this workshop) originated when researchers were trying to solve the issues of speakerphones connected to landline telephones, it has received a recent resurgence of activity as a result of the widespread use of Voice over IP (VoIP) technologies and multi-party communication.

The workshop featured two keynote speeches. The first was Professor Peter Vary of RWTH Aachen University was entitled "Speech enhancement by conditional estimation: noise reduction, error concealment, and bandwidth extension, what makes the difference?". In this talk, Prof. Vary discussed the sources of degradation of the speech signal in wireless communication systems and showed how solutions to recovering from these separate sources of degradation are all applications of conditional estimation, if the problem is cast in a Bayesian framework. This very interesting unifying view was supported by several examples of algorithms and experimental results.

The second keynote address was given by Professor Shihab Shamma of the University of Maryland. His talk, entitled "Cortical spectrotemporal modulations in audio and speech processing", took a physiological view of the processing of the auditory cortex. He discussed how modulation in time and frequency is used to encode discriminative information about speech and music and how these modulations are represented and extracted by the brain. Performing lab experiments on live ferrets, he was able to measure detectors for various spectrotemporal modulations in the ferret's brain. He then proposed ways in which these measurements can be used to create algorithms for the enhancement of noisy speech and the assessment of speech quality.

The workshop lasted three days and featured oral sessions on acoustic echo cancellation, microphone array processing, and single channel speech enhancement. There were multiple poster sessions, but the content of the poster sessions was deliberately mixed so a variety of topics was present at all sessions. The first day of the workshop featured a demo session as well, where live demonstrations included blind source separation and ultrawideband (44.1 kHz) echo cancellation.

The IWAENC Technical Committee also presented the Eberhard Hansler Best Student Paper Award in honor of the workshop founder, Prof. Hansler of the University of Darmstadt. The award went to Tiago Falk for his paper "A non-intrusive quality measure of dereverberated speech". Congratulations to Tiago!

The full workshop proceedings including all papers and the presentations of the keynote speakers are available online. 

IWAENC 2008 Workshop Proceedings