DocumentCode :
699957
Title :
Robust audio speaker segmentation using one class SVMS
Author :
Kadri, Hachem ; Davy, Manuel ; Rabaoui, Asma ; Lachiri, Zied ; Ellouze, Noureddine
Author_Institution :
Unite de Rech. Signal, Image et Reconnaissance des Formes, ENIT, Tunis, Tunisia
fYear :
2008
fDate :
25-29 Aug. 2008
Firstpage :
1
Lastpage :
5
Abstract :
This paper presents a new technique for segmenting an audio stream into pieces, each one contains speeches of only one speaker. Speaker segmentation has been used extensively in various tasks such as automatic transcription of radio broadcast news and audio indexing. The segmentation method used in this paper is based on a discriminative distance measure between two adjacent sliding windows operating on preprocessed speech. The proposed unsupervised detection method which does not require any pre-trained models is based on the use of the exponential family model and 1-SVMs to approximate the generalized likelihood ratio. Our 1-SVM-based segmentation algorithm provides improvements over baseline approaches which use the Bayesian Information Criterion (BIC). The segmentation results achieved in our experiments illustrate the potential of this method in detecting speaker changes in audio streams containing over-lapped and short speeches.
Keywords :
Bayes methods; audio streaming; maximum likelihood estimation; speaker recognition; speech processing; 1-class SVM; BIC; Bayesian information criterion; adjacent sliding windows; audio stream segmentation; exponential family model; generalized likelihood ratio approximation; one class SVM segmentation algorithm; overlapped speech; pretrained model; robust audio speaker segmentation; short speech; speech preprocessing; unsupervised detection method; Discrete wavelet transforms; Kernel; Mel frequency cepstral coefficient; Signal processing; Speech; Support vector machines;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Signal Processing Conference, 2008 16th European
Conference_Location :
Lausanne
ISSN :
2219-5491
Type :
conf
Filename :
7080489
Link To Document :
بازگشت