A modular RNN-based method for continuous Mandarin speech recognition.

A new modular recurrent neural network (MRNN)-based method for continuous Mandarin speech recognition (CMSR) is proposed. The MRNN recognizer is composed of four main modules. The first is a sub-MRNN module whose function is to generate discriminant functions for all 412 base-syllables. It accomplis...

Ausführliche Beschreibung

Bibliographische Detailangaben
Veröffentlicht in:	IEEE Transactions on speech and audio processing 9, 3 (2001).
1. Verfasser:	Yuan-Fu Liao
Format:	Artikel
Sprache:	English
Schlagworte:	Chinese. MCE/GPD algorithm. MRNN recognizer. Base-syllable sequence. Base-syllables. Continuous Mandarin speech recognition. Discriminant functions. Dynamic programming-based recognition search module. Intersyllable coarticulation effect. Intersyllable diphone-like units. Modular RNN-based method. Modular recurrent neural network. Multilevel pruning scheme. Sub-MRNN module. Submodules. Syllable boundaries. Three-stage minimum classification error/generalized probabilistic descent algorithm. Time-alignment.