EUROSPEECH 2003 - INTERSPEECH 2003
8th European Conference on Speech Communication and Technology

Geneva, Switzerland
September 1-4, 2003

        

Automatic Generation of Context-Independent Variable Parameter Models Using Successive State and Mixture Splitting

Soo-Young Suk, Ho-Youl Jung, Hyun-Yeol Chung

Yeungnam University, Korea

A Speech and Character Combined Recognition System (SCCRS) is developed for working on PDA (Personal Digital Assistants) or on mobile devices. In SCCRS, feature extraction for speech and for character is carried out separately, but recognition is performed in an engine. The recognition engine employs essentially CHMM (Continuous Hidden Markov Model) structure and this CHMM consists of variable parameter topology in order to minimize the number of model parameters and to reduce recognition time. This model also adopts our proposed SSMS (Successive State and Mixture Splitting) for generating context independent model. SSMS optimizes the number of mixtures through splitting in mixture domain and the number of states through splitting in time domain. The recognition results show that the proposed SSMS method can reduce the total number of Gaussian up to 40.0% compared with the fixed parameter models at the same recognition performance in speech recognition system.

Full Paper

Bibliographic reference.  Suk, Soo-Young / Jung, Ho-Youl / Chung, Hyun-Yeol (2003): "Automatic generation of context-independent variable parameter models using successive state and mixture splitting", In EUROSPEECH-2003, 2501-2504.