This paper deals with the issues associated with the dereverberation of speech or audio signals using deep neural networks (DNNs). They include feature extraction for DNNs from both clean and reverberant signals and DNN construction for generating dereverberant signals. To evaluate the performance of the proposed dereverberation method, artificially processed reverberant speech signals are obtained and a feed-forward DNN is constructed. It is shown that log spectral distortion (LSD) after applying DNN-based dereverberation is reduced by around 1.9 dB, compared with that of reverberant speech signals.
Authors:
Park, Ji Hyun; Jeon, Kwang Myung; Chun, Chanjun; Yoo, Ji Sang; Kim, Hong Kook
Affiliations:
Gwangju Institute of Science and Technology (GIST), Gwangju, Korea; Kwangwoon University, Seoul, Korea(See document for exact affiliation information.)
AES Convention:
141 (September 2016)
eBrief:300
Publication Date:
September 20, 2016
Subject:
Education, Network Audio, & Signal Processing
Click to purchase paper as a non-member or you can login as an AES member to see more options.
No AES members have commented on this paper yet.
To be notified of new comments on this paper you can subscribe to this RSS feed. Forum users should login to see additional options.
If you are not yet an AES member and have something important to say about this paper then we urge you to join the AES today and make your voice heard. You can join online today by clicking here.