In modern hands-free teleconferencing systems several problems have to be solved in combination, such as corruption of the desired signal by additive noise, acoustic echoes, and room reverberation. In this project the mutual influences of the different subsystems for noise reduction, echo cancellation and listening room compensation (LRC) are studied. Due to the analysis of the subsystems in combination synergies can be exploited.  

Inversion of room impulse responses (RIRs) is still a challenging problem, because RIRs are mixed phase systems, in general, having zeros inside and outside the unit circle in the z-domain. Thus a direct inversion with a stable causal system is not possible. Least squares equalizers approximate the overall concatenated system of LRC-filter and RIR to a given target system and by this lead to optimum equalization in a mean sense given a limited filter order. However, LRC – filters need a reliable estimate of the room impulse response. An acoustic echo canceller identifies the room impulse response and thus can provide an estimate for the equalizer.  This project analyzes the mutual influences of the different subsystems, e.g. the impact of an erroneous RIR estimate on the LRC-filter.

Details