An Analysis of the Performance Evaluation of Syllable Based Tamil Speech Recognition System

Automatic Speech Recognition has been a goal of research for many decades. Many research works have been developed successfully for automatic speech recognition (ASR) of English language. ASR for European languages has not reached their height as ASR in English language. In this work, an implementation of Tamil based automatic speech Recognition System is developed. The ASR has many phases to perform the recognition process. A novel Tamil speech recognition system has been proposed in this work which reduces the complexity and the vocabulary size of the recognition model by applying segmentation at different phases. The temporal features like short term energy, zero crossing rate and the feature vectors based techniques like Mel frequency Cepstral coefficient, linear predictive coding are used for the segmentation. The sound attributes such as Sound Intensity Level, Time Duration and Root Mean Square are used to enhance the effectiveness of the Tamil speech recognition system.