CONF Imseng_ICASSP_2010/IDIAP An Adaptive Initialization Method for Speaker Diarization based on Prosodic Features Imseng, David Friedland, Gerald Gaussian Mixture Models Prosodic features Speaker Diarization EXTERNAL https://publications.idiap.ch/attachments/papers/2010/Imseng_ICASSP_2010.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/Imseng_Idiap-RR-02-2010 Related documents Proceedings IEEE International Conference on Acoustics, Speech and Signal Processing Dallas, USA 2010 March 2010 4946-4949 The following article presents a novel, adaptive initialization scheme that can be applied to most state-ofthe-art Speaker Diarization algorithms, i.e. algorithms that use agglomerative hierarchical clustering with Bayesian Information Criterion (BIC) and Gaussian Mixture Models (GMMs) of frame-based cepstral features (MFCCs). The initialization method is a combination of the recently proposed “adaptive seconds per Gaussian” (ASPG) method and a new pre-clustering and number of initial clusters estimation method based on prosodic features. The presented initialization method has two important advantages. First, the method requires no manual tuning and is robust against file length and speaker count variations. Second, the method outperforms our previously used initialization methods on all benchmark files that were presented in the 2006, 2007, and 2009 NIST Rich Transcription (RT) evaluations and results in a Diarization Error Rate (DER) improvement of up to 67% (relative). REPORT Imseng_Idiap-RR-02-2010/IDIAP An Adaptive Initialization Method for Speaker Diarization based on Prosodic Features Imseng, David Friedland, Gerald EXTERNAL https://publications.idiap.ch/attachments/reports/2009/Imseng_Idiap-RR-02-2010.pdf PUBLIC Idiap-RR-02-2010 2010 Idiap January 2010 The following article presents a novel, adaptive initialization scheme that can be applied to most state-ofthe-art Speaker Diarization algorithms, i.e. algorithms that use agglomerative hierarchical clustering with Bayesian Information Criterion (BIC) and Gaussian Mixture Models (GMMs) of frame-based cepstral features (MFCCs). The initialization method is a combination of the recently proposed “adaptive seconds per Gaussian” (ASPG) method and a new pre-clustering and number of initial clusters estimation method based on prosodic features. The presented initialization method has two important advantages. First, the method requires no manual tuning and is robust against file length and speaker count variations. Second, the method outperforms our previously used initialization methods on all benchmark files that were presented in the 2006, 2007, and 2009 NIST Rich Transcription (RT) evaluations and results in a Diarization Error Rate (DER) improvement of up to 67% (relative).