Modified Amplitude Spectral Estimator for Single-Channel Speech Enhancement

DOI

10.2991/ameii-16.2016.213How to use a DOI?

Keywords

Speech Enhancement, Amplitude Spectral Estimation, Decision-Directed, Soft-Decision

Abstract

The widely used amplitude spectral subtraction for speech enhancement suffers from the unreasonable assumption that the phase spectra of clean speech and noise signal must being uniform. To avoid this problem, Zhu has presented an extensive analysis of amplitude spectral subtraction, and proposed a quantitative analysis based speech spectral recovery (QASSR) approach. However, the residual noise of QASSR approach is still highly perceptible due to the small suppression ratio in speech absent periods. In this paper, a modified algorithm for speech amplitude spectral estimation is proposed by incorporating the decision-directed (DD) method and soft decision technique. Firstly, the a priori SNR in each frame and frequency bin is estimated using the DD method, and then an over attenuation factor is derived from the soft decision technique to further suppress the residual noise in noise-only frames. The performance of our proposed algorithm is evaluated and compared with that of QASSR approach, and the simulation results showed the improved performance of the proposed algorithm in various noise conditions.

Copyright

© 2016, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Zhenhui Zhai
AU  - Shifeng Ou
AU  - Ying Gao
PY  - 2016/04
DA  - 2016/04
TI  - Modified Amplitude Spectral Estimator for Single-Channel Speech Enhancement
BT  - Proceedings of the 2nd International Conference on Advances in Mechanical Engineering and Industrial Informatics (AMEII 2016)
PB  - Atlantis Press
SP  - 1115
EP  - 1120
SN  - 2352-5401
UR  - https://doi.org/10.2991/ameii-16.2016.213
DO  - 10.2991/ameii-16.2016.213
ID  - Zhai2016/04
ER  -