Sparse DNN-based speaker segmentation using side information

Sparse deep neural networks (SDNNs) for speaker segmentation are proposed. First, the SDNNs are trained using the side information that is the class label of the input. Then, speaker-specific features are extracted from the super-vector feature of the speech signal by the SDNNs. Lastly, the label of...

Full description

Saved in:
Bibliographic Details
Published in:Electronics letters 2015-04, Vol.51 (8), p.651-653
Main Authors: Ma, Yong, Bao, Chang-Chun
Format: Article
Language:English
Subjects:
audio databases
Bayes methods
Bayesian information criterion method
BIC method
continuous speech stream
deep auto‐encoder networks method
feature extraction
input class label
k‐means clustering
Labels
multispeaker speech stream corpus
neural nets
Neural networks
pattern clustering
SDNN
Segmentation
Segments
side information
sparse deep neural networks
sparse DNN‐based speaker segmentation
speaker recognition
speaker‐specific feature extraction
Speech
Speech and audio processing and translation
speech frame
speech signal
Streams
supervector feature
TIMIT database
Vector quantization
Citations: Items that this one cites
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Sparse deep neural networks (SDNNs) for speaker segmentation are proposed. First, the SDNNs are trained using the side information that is the class label of the input. Then, speaker-specific features are extracted from the super-vector feature of the speech signal by the SDNNs. Lastly, the label of each speech frame is obtained by K-means clustering, which is used to segment different speakers of a continuous speech stream. The performance evaluation using the multi-speaker speech stream corpus generated from the TIMIT database shows that the proposed speaker segmentation algorithm outperforms the Bayesian information criterion method and the deep auto-encoder networks method.
ISSN:0013-5194
1350-911X
1350-911X
DOI:10.1049/el.2015.0298