Enhancing Phoneme Recognizer Performance with a Simple Rule-based Language Model

The phoneme classification inaccuracy at the acoustic phonetic level is a major weakness in most speech recognition systems. However, the inaccuracy will violate phonotactic constraints at the acoustic phonetic level. A better performance is expected if a language model is adopted in a recognition system for post-processing phoneme estimates and making corrections with a set of explicit rules of the language used. We developed a simple language model with a set of rules to correct phoneme classification errors made by a Hidden Markov Model based phoneme recognizer. The experimental results indicate that about 20% of recognition errors can be corrected.