Second European Conference on Speech Communication and Technology

Genova, Italy
September 24-26, 1991


Integrated Phoneme-Function Word Architecture of Hidden Control Neural Networks for Continuous Speech Recognition

Bojan Petek, Alex H. Waibel, Joseph M. Tebelskis

School of Computer Science, Carnegie Mellon University, Pittsburgh, Pennsylvania, USA

We present a context-dependent, phoneme and function word based, Hidden Control Neural Network (HCNN) architecture for continuous speech recognition. The system can be seen as a large vocabulary extension of the word-based HCNN system proposed by Levin [Levin90]. Initially, we analysed the context-independent HCNN modeling principle in the framework of the Linked Predictive Neural Network speech recognition system [Tebelskis91] and found that it results in a 6% increase of the word recognition accuracy at perplexity 402. Significant savings in the resource requirements and computational load for the HCNN implementation can be achieved. In speaker-dependent recognition experiments with perplexity 111, the current versions of the LPNN and HCNN systems achieve 60% and 75% word recognition accuracy, respectively. Keywords: Automatic Speech Recognition, Hid- den Control Neural Network, Large vocabulary recognition, Context-dependent modeling, Function-word modeling.

Full Paper

Bibliographic reference.  Petek, Bojan / Waibel, Alex H. / Tebelskis, Joseph M. (1991): "Integrated phoneme-function word architecture of hidden control neural networks for continuous speech recognition", In EUROSPEECH-1991, 1407-1410.