Environment Classification via Blind Roomprints Estimation

TL;DR

This paper introduces a new environment classification method for speech recordings that uses multi-band RT60 analysis of blind channel estimates, achieving high accuracy without relying on reverberation tail selection.

Contribution

The novel approach eliminates the need for reverberation tail selection by leveraging multi-band RT60 analysis of blind channel estimates for environment classification.

Findings

Achieves up to 93.6% accuracy on ACE corpus recordings.

Does not require reverberation tail selection.

Effective in classifying environments from speech recordings.

Abstract

In this paper we present a novel approach for environment classification for speech recordings, which does not require the selection of decaying reverberation tails. It is based on a multi-band RT60 analysis of blind channel estimates and achieves an accuracy of up to 93.6% on test recordings derived from the ACE corpus.