An acoustic-phonetic feature-based system for automatic phoneme recognition in continuous speech

An acoustic-phonetic feature- and knowledge-based system for the automatic segmentation, broad categorization and fine phoneme recognition of continuous speech is described. The system uses an auditory-based front-end processing and incorporates new knowledge-based algorithms to automatically segment the speech into phoneme-like segments that are further categorized into 4 main categories: sonorants, stops, fricatives and silences. The final outputs from the system are 19 class phonemes which contain 7 stops, 6 fricatives, nasals and semivowels, 4 vowel classes and silences. The system was tested on continuous speech from 30 speakers having 7 different dialects from the TIMIT database which were not used in the design process. The results are 92% accuracy for the segmentation and categorization, 86% for the stop classification, 90% for the fricative classification, 75% for the nasal and semivowel extraction and 82% for the vowel recognition. These results compare favorably with previous phoneme classification results.