An optimal learning method for minimizing spotting errors

A novel design method for word spotting, called MSPE (minimum spotting error), is proposed which guarantees a minimum spotting error situation in a probabilistic sense through MCE/GPD (minimum classification error/generalized probabilistic descent) optimization. MPSE makes it possible to train all trainable parameters consistently; this feature implies an innovative departure from conventional, heuristic approaches to spotter design. Experimental results have demonstrated a very high utilization potential for MSPE.<<ETX>>

[1]  Shigeru Katagiri,et al.  Application of a generalized probabilistic descent method to dynamic time warping-based speech recognition , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[2]  Chin-Hui Lee,et al.  Segmental GPD training of HMM based speech recognizer , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[3]  Biing-Hwang Juang,et al.  Discriminative template training for dynamic programming speech recognition , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[4]  Biing-Hwang Juang,et al.  New discriminative training algorithms based on the generalized probabilistic descent method , 1991, Neural Networks for Signal Processing Proceedings of the 1991 IEEE Workshop.

[5]  L. Rabiner,et al.  A modified K‐means clustering algorithm for use in speaker‐independent isolated word recognition , 1984 .

[6]  Shigeru Katagiri,et al.  Prototype-based discriminative training for various speech units , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[7]  Biing-Hwang Juang,et al.  Discriminative learning for minimum error classification [pattern recognition] , 1992, IEEE Trans. Signal Process..