Hoi-Rin Kirn
HMnet Evaluation for Phonetic
Environment Variation of Training
Data
Abstract:Evaluation of the SSS algorithm has usually been performed on the aspects of recognition
performance and number of the free parameters to be estimated, which is related to robustness
of the model. In this work, I have investigated how the variation of phonetic environments of
training data affects on the HMnet generated by SSS algorithm. By analyzing the results, I think
we could understand more better the SSS algorithm's behavior and limitation.