Machine Learning
A Tutorial on Support Vector Machines for Pattern Recognition
Data Mining and Knowledge Discovery
Speech recognition in mobile environments
Speech recognition in mobile environments
Support vector machines for speech recognition
Support vector machines for speech recognition
Integrating time alignment and neural networks for high performance continuous speech recognition
ICASSP '91 Proceedings of the Acoustics, Speech, and Signal Processing, 1991. ICASSP-91., 1991 International Conference
A comparison of methods for multiclass support vector machines
IEEE Transactions on Neural Networks
Hi-index | 0.00 |
Extensive research on continuous speech recognition (CSR) marks the current trend in Automatic Speech Recognition (ASR) course. Topics concerning CSR include acoustic modeling, segmentation, language modeling and word decoding. This paper describes an efficient method for recognizing words derived from continuous speech signal. Speech boundaries are very obscure due to overlapping context of the acoustic units, designated as the coarticulation effects. Considering larger segments such as words, helps tapering off much of this issue. Nevertheless, the problem still exists between interconnected words thus addressing the aim of our study. Like any other static classifier, Support Vector Machines (SVM) inherently unable to receive variable input length. Our solution is to extend the size of acoustic segment incrementally rather than shifting a fixed block over the utterance. Recognition is determined via voting the highest posterior probability score for a word segment by means of confidence measure. We proposed a two-fold embedded grammar strategy that eliminates both the decoding process and out of vocabulary errors. We argue that these notions have some advantages over previous works, hence could yield better result.