CONF Motlicek_ICASSP2013-2_2013/IDIAP ACCENT ADAPTATION USING SUBSPACE GAUSSIAN MIXTURE MODELS Motlicek, Petr Garner, Philip N. Kim, Namhoon Cho, Jeongmi EXTERNAL https://publications.idiap.ch/attachments/papers/2013/Motlicek_ICASSP2013-2_2013.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/Motlicek_Idiap-RR-38-2013 Related documents IEEE - The 38th International Conference on Acoustics, Speech, and Signal Processing (ICASSP) Vancouver, BC, Canada 2013 7170-7174 1520-6149 10.1109/ICASSP.2013.6639054 doi This paper investigates employment of Subspace Gaussian Mixture Models (SGMMs) for acoustic model adaptation towards different accents for English speech recognition. The SGMMs comprise globally-shared and state-specific parameters which can efficiently be employed for various kinds of acoustic parameter tying. Research results indicate that well-defined sharing of acoustic model parameters in SGMMs can significantly outperform adapted systems based on conventional HMM/GMMs. Furthermore, SGMMs rapidly achieve target acoustic models with small amounts of data. Experiments performed with US and UK English versions of the Wall Street Journal (WSJ) corpora indicate that SGMMs lead to approximately 20% and 8% relative improvements with respect to speaker-independent and speaker-adapted acoustic models respectively over conventional HMM/GMMs. Finally, we demonstrate that SGMMs adapted only with 1.5 hours can reach performance of HMM/GMMs trained with 18 hours. REPORT Motlicek_Idiap-RR-38-2013/IDIAP ACCENT ADAPTATION USING SUBSPACE GAUSSIAN MIXTURE MODELS Motlicek, Petr Garner, Philip N. Kim, Namhoon Cho, Jeongmi Accented speech Acoustic model adaptation Automatic Speech Recognition Under-resourced data EXTERNAL https://publications.idiap.ch/attachments/reports/2013/Motlicek_Idiap-RR-38-2013.pdf PUBLIC Idiap-RR-38-2013 2013 Idiap Rue Marconi 19, Martigny, Switzerland November 2013 This paper investigates employment of Subspace Gaussian Mixture Models (SGMMs) for acoustic model adaptation towards different accents for English speech recognition. The SGMMs comprise globally-shared and state-specific parameters which can efficiently be employed for various kinds of acoustic parameter tying. Research results indicate that well-defined sharing of acoustic model parameters in SGMMs can significantly outperform adapted systems based on conventional HMM/GMMs. Furthermore, SGMMs rapidly achieve target acoustic models with small amounts of data. Experiments performed with US and UK English versions of the Wall Street Journal (WSJ) corpora indicate that SGMMs lead to approximately 20% and 8% relative improvements with respect to speaker-independent and speaker-adapted acoustic models respectively over conventional HMM/GMMs. Finally, we demonstrate that SGMMs adapted only with 1.5 hours can reach performance of HMM/GMMs trained with 18 hours.