1st Joint SIG-IL/Microsoft Workshop on Speech and Language Technologies for Iberian Languages
Porto Salvo, Portugal
In a room where several distant microphones are capturing signals in parallel, the quality of the recorded speech signals strongly depends on the characteristics of the room impulse responses that describe the wave propagation between each source and each microphone. In this paper we present an initial attempt to investigate the possibility of selecting the microphone that offers the best quality of speech. As we want to apply it to an automatic speech recognition system, we aim to select the microphone according to some optimization criterion that has been inferred from the recognition rate in a prior learning process. Several energy-related measures that carry relevant information of the room impulse response are being considered. They should be estimated directly from the speech signal, possibly in real time, but avoiding the need to estimate the whole room impulse response. In this paper, we present the rationale behind the intended investigation, and offer preliminary experiments for a large vocabulary continuous speech recognition task which show how microphone selection using an ideal relative energy measure can largely improve the recognition rate.
Index Terms: microphone selection, reverberation, room impulse response, ASR
Bibliographic reference. Wolf, Martin / Nadeu, Climent (2009): "Towards microphone selection based on room impulse response energy-related measures", In SLTECH-2009, 61-64.