Title :
A Novel Speaker Change Detection Algorithm
Author :
Yu, Xiaoqing ; Tan, Haiying ; Wan, Wanggen
Author_Institution :
Shanghai Univ., Shanghai
Abstract :
To cope with the problem of speaker voice change detection, a novel speaker change detection algorithm is proposed in this paper. The new algorithm is developed based on traditional BIC algorithm in order to choose the voice segmentation point firstly, and then analyze the BIC change trend to decide which point is the real speaker voice change point. To compare with the normal BIC detection algorithm, the precision and recall rate of this new one is improved a lot. And also the new one has lower computational complexity compared with the traditional one.
Keywords :
speaker recognition; speech processing; BIC detection algorithm; computational complexity; speaker change detection algorithm; speaker voice change detection; voice segmentation; Algorithm design and analysis; Change detection algorithms; Computational complexity; Detection algorithms; Gaussian processes; Hidden Markov models; Human voice; Indexing; Speech analysis; Speech recognition;
Conference_Titel :
Communications, Circuits and Systems, 2007. ICCCAS 2007. International Conference on
Conference_Location :
Kokura
Print_ISBN :
978-1-4244-1473-4
DOI :
10.1109/ICCCAS.2007.4348127