Web Based Network Speech Recognition

Abstract

In modern technological world there is a continuous flow of improvements, as it has been with computers, tablets and smart phones. There is a need to explore alternatives for effectively using those new devices and systems. Effective usage of those devices may include: virtual keyboards rather than physical ones, the use of a touch screen in-stead of a mouse, using location based services to find events that are happening around us instead of accessing a Website in order to get the information, talking to a smart phone or tablet in order to call, text or do tasks, thus replacing the physical interaction via a touchscreen or keyboard. The main contribution of this thesis is to design and im-plement a Web-based Network Speech Recognition system using Open Source compo-nents and new emerging technologies. This system can take audio queries from a Web browser, feeds them into the Speech engine and returns the result back to the Web browser client. Web-based Network Speech Recognition systems already have been built by Google, Nuance and many other companiess. Implementation however differs in various ways, such as the use of WebSockets in real time or the use of HTTP Request / Response method. The system developed in the theis is entirely composed of open source elements: the speech recognition engine that serves the speech recognition re-quests, and a Web Server to receive the audio stream from the Web browser clients. The designed system efficiency is high and it can serve multiple clients and it provides good processing power making it able to manage heavy load operations with reasonable ef-fort.