Research of a non-specific person noise-robust speech recognition system
WiCOM'09 Proceedings of the 5th International Conference on Wireless communications, networking and mobile computing
Comparison studies of LS_SVM and SVM on modeling for fermentation processes
ICNC'09 Proceedings of the 5th international conference on Natural computation
GA-SVM based framework for time series forecasting
ICNC'09 Proceedings of the 5th international conference on Natural computation
Optimization of SVM parameters based on PSO algorithm
ICNC'09 Proceedings of the 5th international conference on Natural computation
ICNC'09 Proceedings of the 5th international conference on Natural computation
ICNC'09 Proceedings of the 5th international conference on Natural computation
A novel knowledge discovery model for fishery forecasting
FSKD'09 Proceedings of the 6th international conference on Fuzzy systems and knowledge discovery - Volume 3
Multi-class classification for Wuhan area's TM image based on support vector machine
FSKD'09 Proceedings of the 6th international conference on Fuzzy systems and knowledge discovery - Volume 5
Climate prediction by SVM based on initial conditions
FSKD'09 Proceedings of the 6th international conference on Fuzzy systems and knowledge discovery - Volume 5
Fuzzy SVM controller for robotic manipulator based on GA and LS algorithm
FSKD'09 Proceedings of the 6th international conference on Fuzzy systems and knowledge discovery - Volume 6
Location-Aware Caching for Semantic-Based Image Queries in Mobile AD HOC Networks
International Journal of Multimedia Data Engineering & Management
An efficient multiple-kernel learning for pattern classification
Expert Systems with Applications: An International Journal
A vector-valued support vector machine model for multiclass problem
Information Sciences: an International Journal
A MapReduce-based distributed SVM ensemble for scalable image classification and annotation
Computers & Mathematics with Applications
| Hi-index | 35.69 |
We present a discriminative training algorithm, that uses support vector machines (SVMs), to improve the classification of discrete and continuous output probability hidden Markov models (HMMs). The algorithm uses a set of maximum-likelihood (ML) trained HMM models as a baseline system, and an SVM training scheme to rescore the results of the baseline HMMs. It turns out that the rescoring model can be represented as an unnormalized HMM. We describe two algorithms for training the unnormalized HMM models for both the discrete and continuous cases. One of the algorithms results in a single set of unnormalized HMMs that can be used in the standard recognition procedure (the Viterbi recognizer), as if they were plain HMMs. We use a toy problem and an isolated noisy digit recognition task to compare our new method to standard ML training. Our experiments show that SVM rescoring of hidden Markov models typically reduces the error rate significantly compared to standard ML training.