Marcos A. Cantu1, H. Steven Colburn2, Volker Hohmann1
1University of Oldenburg, Department of Medical Physics and Acoustics and the Cluster of Excellence Hearing4All, Oldenburg, Germany
2Boston University, Department of Biomedical Engineering, Boston, MA, USA
Interfering speech has rapid spectrotemporal fluctuations that established noise reduction algorithms have difficulty suppressing without a concomitant loss, or distortion, of binaural cues for spatial hearing. An ongoing project at the University of Oldenburg (UOL) involves development and evaluation of prototype Short-Time Target Cancellation (STTC) assistive listening devices that can enhance speech intelligibility of a target talker, and attenuate interfering talkers, while still preserving binaural cues for spatial hearing. The STTC processing computes a ratio mask (i.e., a time-varying spectral gain) that can be applied to the binaural signals at the Left and Right ears, thereby attenuating the interfering talkers. The STTC processing is causal and memoryless, with low requirements in terms of memory size and computational power, and is designed to run online in real-time without training or any a priori knowledge about the number or locations of interfering sound sources; only an assumed “look” direction is needed. The STTC processing can be used either to filter the binaural signals at the Left and Right ears or as a postfilter for adaptive beamforming. Where adaptive beamforming processing computes a complex-valued filter-vector, the STTC processing computes a real-valued time-varying spectral gain; the two approaches are compatible and our evaluation results indicate that their combination has an additive effect. Although the STTC processing, and adaptive beamforming, can be implemented with standard in-ear or behind-the-ear (BTE) hearing aid earpieces, better performance can be achieved via a small microphone array integrated into the frame of a pair of eyeglasses. Evaluation results, using simulations in virtual acoustic environments, indicate that these prototype STTC assistive listening devices can effect enhancement of a target talker, and attenuation of interfering talkers, in both anechoic space and reverberation.
Acknowledgements: This work was supported by the Cluster of Excellence EXC 2177 Hearing4All, funded by the German Research Foundation (DFG), and by NIH/NIDCD grants R01DC000100 and R01DC015429. The content herein is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.