Evaluation of Algorithms for Blind Reverberation Time Estimation

Proposal for a Research Internship


Evaluation of Algorithms for Blind Reverberation Time Estimation


The reverberation time, T60, is one of the most important quantities to describe
the acoustical properties of a room. It is defined as the time interval in which the
sound energy decays by 60 dB after switching off the exciting sound source and
describes the amount of room reverberation. Knowledge about the reverberation
time is exploited by various algorithms for speech and audio signal processing, e.g.,
in the context of automatic speech recognition or speech dereverberation. As a
consequence, numerous algorithms for the blind estimation of the reverberation
time from a noisy and reverberant speech signal have been published in recent
years, which are mostly considering single-channel recordings, e.g., [1].
The variety of proposals motivates a thorough performance evaluation of different
algorithms for T60 estimation. The aim of this work is to implement state-of-the-art
algorithms for blind T60 estimation and to evaluate their estimation accuracy for
different scenarios (noise level, room reverberation etc.) by means of the evaluation
dataset and measures used for the ACE Challenge [2]. This investigation should
reveal the pros and cons of different approaches and to provide hints for possible
This research work requires basic knowledge about statistical signal processing and
MATLAB programming skills.



Prof. Dr.-Ing. Walter Kellermann


Dr.-Ing. Heinrich L¨ollmann and M.Sc. Andreas Brendel, Room 05.024/05.018,