F.J. Fraga (Brazil)
Phoneme Recognition, Speech-to-Text Systems
The implementation of an automatic speech recognition system with unlimited vocabulary, for Portuguese language spoken in Brazil, may be done by two steps: Phoneme recognition and then conversion of the resulting phoneme sequence in a grapheme sequence (letters that form words). This paper presents such a system by briefly describing the phoneme recognizer and then the phonologic–graphemic conversion algorithm in more detail. The algorithm is entirely based on rules that are extracted from Portuguese language structure, allowing the transition from phoneme level to word level without using any kind of lexical entries. In this way the system is able to recognize any word in Portuguese, with no limitation on vocabulary size.
Important Links:
Go Back
