A perceptually-weighted deep neural network for monaural speech enhancement in various background noise conditions

Liu, Qingju; Wang, Wenwu; Jackson, Philip JB; Tang, Y

Abstract

Deep neural networks (DNN) have recently been
shown to give state-of-the-art performance in monaural speech
enhancement. However in the DNN training process, the perceptual
difference between different components of the DNN
output is not fully exploited, where equal importance is often
assumed. To address this limitation, we have proposed a new
perceptually-weighted objective function within a feedforward
DNN framework, aiming to minimize the perceptual difference
between the enhanced speech and the target speech. A perceptual
weight is integrated into the proposed objective function, and
has been tested on two types of output features: spectra and
ideal ratio masks. Objective evaluations for both speech quality
and speech intelligibility have been performed. Integration of our
perceptual weight shows consistent improvement on several noise
levels and a variety of different noise types.

Citation

Liu, Q., Wang, W., Jackson, P. J., & Tang, Y. (2017, August). A perceptually-weighted deep neural network for monaural speech enhancement in various background noise conditions. Presented at EUSIPCO 2017, the 25th European Signal Processing Conference, Kos Island, Greece

• Presentation Conference Type: Other
• Conference Name: EUSIPCO 2017, the 25th European Signal Processing Conference
• Conference Location: Kos Island, Greece
• Start Date: Aug 28, 2017
• End Date: Sep 2, 2017
• Acceptance Date: May 25, 2017
• Deposit Date: Jun 1, 2017
• Publisher URL: https://www.eusipco2017.org/
• Additional Information: Event Type : Conference