7th International Conference on Spoken Language Processing

September 16-20, 2002
Denver, Colorado, USA

Recognition of Continuous Speech Segments of Monophone Units Using Support Vector Machines

Weifeng Lee, C. Chandra Sekhar, Kazuya Takeda, Fumitada Itakura

Nagoya University, Japan

In this paper, we study support vector machine based approaches for acoustic modeling of subword units in continuous speech. Classification of subword unit segments is considered as a multi-class pattern recognition problem. In conventional approaches for multi-class pattern recognition using support vector machines, learning involves discrimination of each class against all the other classes. We propose a close-class-set discrimination method suitable for large-classset pattern recognition problems. In the proposed method, learning involves discrimination of each class against a subset of classes confusable with it and included in its close-class-set. We consider different criteria for identification of close-class-sets. We study the effectiveness of the proposed method in reducing the complexity of multi-class pattern recognition systems. We present our studies on recognition of continuous speech segments of 41 mono-phone units in a large corpus of Japanese speech.

Full Paper

Bibliographic reference.  Lee, Weifeng / Sekhar, C. Chandra / Takeda, Kazuya / Itakura, Fumitada (2002): "Recognition of continuous speech segments of monophone units using support vector machines", In ICSLP-2002, 2653-2656.