A Spatial Spectral Mask for Dual-Microphone Systems

With significantly increasing human/machine interaction, the acoustic equipment or any terminal of a multimedia service system is challenged to ensure the speech intelligibility, speech quality or comfortable perception for the listener. Therefore, speech enhancement is very necessary procedure for front-end processing in most speech applications, such as: high quality audio recording, dialog systems, dictation systems, audio-video conference, computer games. Recently, scholars and researchers have paid much attention to studying microphone array (MA) technology for both extracting desired target speech components at a certain direction and suppressing background from other directions. Generalized Sidelobe Canceller (GSC) beamformer is suitable installed for MA-based acoustic devices, which includes hearing aids, mobile phones, teleconferencing system, audio-video controlled device. Because of many reasons, such as the different microphone sensitivities or the error of direction of arrival of interest signal, the GSC beamformer’s output often corrupted and speech quality degraded. In this article, the authors suggested a spatial spectral mask to enhance GSC beamformer’s evaluation. The illustrated simulation shows the improvement of speech quality in the terms of signal-to-noise ratio (SNR) from 2.5 to 6.8 dB and speech distortion reduced to 11.2 dB. The advantage of the author’s method is the capability of incorporating into multi-channel systems.