14thAnnual Conference of the International Speech Communication Association

Lyon, France
August 25-29, 2013

Target-to-Non-Target Directional Ratio Estimation Based on Dual-Microphone Phase Differences for Target-Directional Speech Enhancement

Seon Man Kim, Hong Kook Kim

GIST, Korea

In this paper, we propose a dual-microphone target-directional speech enhancement system utilizing target-to-non-target directional signal ratio (TNR) based on dual-microphone phase differences in adverse noise environments. One of the important issues associated with multi-microphone speech enhancement methods is the effective utilization of spatial cues such as phase differences for target-speech estimation within noisy speech. To this end, a TNR estimation method is presented based on phase differences between dual-microphone signals. Then, the estimated TNR is incorporated into a Wiener filter to obtain a masking filter for speech enhancement. Consequently, it is shown from a perceptual evaluation of speech quality that the performance of the proposed speech enhancement system outperforms those of conventional single- or dual-microphone speech enhancement systems based on a Wiener filter, beamformer, and phase-error-based filter under noise conditions with a signal-to-noise ratio ranging from 0 to 20 dB.

Full Paper

Bibliographic reference.  Kim, Seon Man / Kim, Hong Kook (2013): "Target-to-non-target directional ratio estimation based on dual-microphone phase differences for target-directional speech enhancement", In INTERSPEECH-2013, 3254-3258.