Bangla Home Assistant Using CMU Sphinx Offline Speech Recognition

Abstract

We live in an era of science and technology. Our everyday life is influenced in various ways by technologies. Anywhere we look, we see the use of technology. With the increase of modern technologies, we are going to a point where almost every work is done by our tech devices. And the way of interactions has also been changed. Now various types of interactions are being developed which makes using our devices easier than ever. Such a type of interaction is speech recognitions. It is now very popular in the current era because all the tech giants are focusing on speech recognition with their “Personal assistants” like Google Assistant, Amazon Alexa, and Microsoft Cortana. In a speech recognition system, the user uses his/her voice to give command to a device. But sadly, all of these works are being done in English or other popular languages. Our mother tongue, Bengali was left out. So, our goal was to make a voice interactable device that can be used offline for executing tasks. And all of these would be in Bengali. So, we did some research on speech recognitions and made a recognizer that understands commands given in Bengali. And with that, we made a home assistant device so that we can bring out the maximum usability of the recognizer. Because when you are at home, doing some other work, the voice recognizer would be very much helpful. In the future, incorporation of AI with machine learning will extend the idea further.