Hierarchical phoneme classifier for Hindi speech
Abhinav Singhvi, Prashant Gupta, Sudip Sanyal · 2008
In this work we develop a phoneme segmentation and phoneme recognition for Hindi speech recognition. Novel features like the number of dominating frequencies, the primary and secondary dominating frequencies are used in addition to standard features like zero crossing, length of the phoneme and energy. The results obtained with standard data sets show that the system has an accuracy of more than 85%.