A procedure to generate training sequences for a connected word recognizer using the segmental k-means training algorithm.
Rajendra P. MikkilineniJay G. WilponLawrence R. RabinerPublished in: ICASSP (1988)
Keyphrases
- training algorithm
- hidden markov models
- back propagation
- k means
- neural network
- training process
- neural network training
- word recognition
- conjugate gradient
- support vector machine
- sequence alignment
- speech recognition
- learning rate
- rbf neural network
- cluster analysis
- supervised training
- learning algorithm
- hidden layer
- handwriting recognition
- expectation maximization
- word segmentation
- clustering algorithm
- writer independent
- bp neural network
- radial basis function
- optimization algorithm
- co occurrence
- multi objective
- pairwise
- pattern recognition
- face recognition