SPEAKER INDEPENDENT PHONEME RECOGNITION USING NEURAL NETWORKS

Phoneme recognition is important for successful development of speech recognizers in most real world applications. While speaker dependent phoneme recognizers have achieved close to 100 % accuracy, the speaker independent phoneme recognition systems have poor accuracy not exceeding 75%.In this paper we describe a two-module speaker independent phoneme recognition system for all-Indian English speech. The first module performs classification of phonemes recognition using Probabilistic neural networks. The second module executes the recognized phonemes from the classified phonemes employing Recurrent Neural Networks. The system was trained by Indian English speech consisting of 1000 words uttered by 50 speakers. The test samples comprised 500 words spoken by a different set of 30 speakers. The recognition accuracy is found to be 98 % which is well above the previous results