Trung Nguyen Thanh Undergraduate Dissertation 2016/17

Webportal for the collection of disordered speech from people with severe physical impairments

Supervised by H.Christensen

Abstract

Nowadays, speech has become one of the most popular methods to control different electronic devices such as TV, mobile phone, home automation system, etc. However, most mainstream speech recognition systems do not work effectively on people with disordered speech due to the special characteristics of these unique voices. Therefore, many systems (e.g. homeService, VIVOCA) were specifically designed to work with speech disorders and in order to make those as powerful as the standard systems, huge amounts of speech data from people with dysarthria are required. However, due to many constraints in the current recording process, it is difficult for researchers to collect that much data in a short time period.

By taking advantage of modern web-based technologies, the aim of this project is to develop a webportal that allows people suffering severe physical impairments with associated speech disorders to easily record their speech online and helps in building up a large collection of disordered speech in a shorter amount of time compared to the existing methods for research studies and training the statistical models underpinning the automatic speech recognition technology.