SpokenWord identification for Malayalam using Artificial Neural Network

This paper focuses on developing a syllable based speaker independent speech recognition for Malayalam language. An ANN model is proposed for an automatic syllabification to understand the isolated word utterances as syllables. The learning was performed with isolated word utterances of multiple speakers after pre-processing. Pre-processing involves noise removal, framing, segmentation, filtering and feature extraction. It is found that ANN shows satisfactory result for neutral and gender independent speech identification with an accuracy of 75% in an average for 4 different experiments. The work has also been extended by proposing a deep learning architecture for Automatic Speech Recognition(ASR) for better accuracy.