BROWSE
Detail
Conference Paper
Non-Speech Section Detection on Media Contents
Cited - time in 
Share 
Share
- Authors
- Inseon Jang, ChungHyun Ahn, Jeongil Seo, Younseon Jang
- Issue Date
- 2017-01
- Citation
- International Workshop on Advanced Image Technology (IWAIT) 2017, pp.1-2
- Language
- English
- Type
- Conference Paper
- Abstract
- This paper addresses a problem of non-speech section detection for the DVS (Descriptive Video Service) authoring, whose goal is to discriminate the non-speech section where an audio description can be inserted in the media contents which involve the presence of various sounds. The proposed method is based on the Deep Neural Network (DNN) trained with the audio features extracted from the center channel signal of a full-mix stereo audio. Jointly exploiting the inter-channels structure of the broadcast audio and speech signal characteristics, it provides superior performance on the error rate and the convergence speed compared with the conventional method. Experiments on real broadcast audio confirm the high performance of the proposed method.
- KSP Keywords
- Audio Features, Conventional methods, Deep neural network(DNN), High performance, Non-speech, Signal characteristics, Speech Signals, Stereo audio, Video service, audio description, convergence speed