Badenhorst, JDavel, MHBarnard, E2012-02-242012-02-242011-11Badenhorst, J, Davel, MH and Barnard, E. Trajectory behaviour at different phonemic context sizes. The 22 International Symposium of the Pattern Recognition Association of South Africa, Prasa 2011, Vanderbijlpark, South Africa, 22-25 November 2011http://hdl.handle.net/10204/5600The 22 International Symposium of the Pattern Recognition Association of South Africa, Prasa 2011, Vanderbijlpark, South Africa, 22-25 November 2011The authors propose a piecewise-linear model for the temporal trajectories of Mel Frequency Cepstral Coefficients during phone transitions. As with conventional Hidden Markov Models, the parameters of the model can be estimated for different phonemic context sizes, but their model allows for an intuitive understanding of the impact of context size. They find that the most detailed models, predictably, match the coefficient tracks best - but when data scarcity forces them to use less detailed models, different types of context modelling (clustered triphones versus biphones) have complimentary behaviours. The authors discuss how this complimentarity may be useful for data-efficient ASR.enPhonemic context sizesTemporal trajectoriesMel Frequency Cepstral CoefficientsPhone transitionsConference PresentationBadenhorst, J., Davel, M., & Barnard, E. (2011). Trajectory behaviour at different phonemic context sizes. PRASA. http://hdl.handle.net/10204/5600Badenhorst, J, MH Davel, and E Barnard. "Trajectory behaviour at different phonemic context sizes." (2011): http://hdl.handle.net/10204/5600Badenhorst J, Davel M, Barnard E, Trajectory behaviour at different phonemic context sizes; PRASA; 2011. http://hdl.handle.net/10204/5600 .TY - Conference Presentation AU - Badenhorst, J AU - Davel, MH AU - Barnard, E AB - The authors propose a piecewise-linear model for the temporal trajectories of Mel Frequency Cepstral Coefficients during phone transitions. As with conventional Hidden Markov Models, the parameters of the model can be estimated for different phonemic context sizes, but their model allows for an intuitive understanding of the impact of context size. They find that the most detailed models, predictably, match the coefficient tracks best - but when data scarcity forces them to use less detailed models, different types of context modelling (clustered triphones versus biphones) have complimentary behaviours. The authors discuss how this complimentarity may be useful for data-efficient ASR. DA - 2011-11 DB - ResearchSpace DP - CSIR KW - Phonemic context sizes KW - Temporal trajectories KW - Mel Frequency Cepstral Coefficients KW - Phone transitions LK - https://researchspace.csir.co.za PY - 2011 T1 - Trajectory behaviour at different phonemic context sizes TI - Trajectory behaviour at different phonemic context sizes UR - http://hdl.handle.net/10204/5600 ER -