Elevato-CDR: Speech Enhancement in Large Delay and Reverberant Assisted Living Scenarios

Speech enhancement algorithms are needed for Assisted Living (AL) environments, which are characterized by large inter-microphone spacings, significant phase-wraparound, and high reverberation times (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T_{60}$</tex>). We contribute Elevato-CDR, which addresses the task of enhancing speech given the large relative delays and high levels of reverberation, which are characteristic of AL scenarios. The purpose of this paper is to evaluate Elevato-CDR's performance and compare it with benchmark reverberation schemes, using speech mixtures which experience reverberation and background noise. Our findings indicate that Elevato-CDR successfully outperforms classical dereverberation algorithms in terms of subjective (MARS, PESQ) and objective (SDR, SRR) performance measures in environments where the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T_{60}=4.1\mathrm{s}$</tex> and the inter-microphone spacing is up to 3.7 m. These findings demonstrate that combining a robust, big relative delay estimation technique, called the Elevatogram with a coherence-based reverberation attenuation method extends the useful range of operation of speech enhancement to AL environments.