Punjabi Automatic Speech Recognition Using HTK
Mohit Dua, Virender Kadyan, Shelza Dua · 2012
This paper aims to discuss the implementation of an isolated word Automatic Speech Recognition system (ASR) for an Indian regional language Punjabi. The HTK toolkit based on Hidden Markov Model (HMM), a statistical approach, is used to develop the system. Initially the system is trained for 115 distinct Punjabi words by collecting data from eight speakers and then is tested by using samples from six speakers in real time environments. To make the system more interactive and fast a GUI has been developed using JAVA platform for implementing the testing module. The paper also describes the role of each HTK tool, used in various phases of system development, by presenting a detailed architecture of an ASR system developed using HTK library modules and tools. The experimental results show that the overall system performance is 95.63 % and 94.08%.