Title: Investigation of voice disorders and recognising the speech of children with hearing impairment in classical Tamil language

Authors: C. Jeyalakshmi; A. Revathi; V. Krishnamurthi

Addresses: Department of Electronics and Communication Engineering, Trichy Engineering College, Trichy 621132, Tamil Nadu, India ' Department of Electronics and Communication Engineering, Saranathan College of Engineering, Trichy 620012, Tamil Nadu, India ' Department of Electronics and Communication Engineering, College of Engineering, Anna University, Chennai 600005, Tamil Nadu, India

Abstract: Speech from the hearing-impaired persons is highly disordered like dysarthric speech, according to their degree of hearing loss. A continuous speech recognition system for the hearing-impaired in Tamil language has been developed using a short vocabulary and it produces a recognition accuracy of 84.0%. It utilises the alphabet model instead of the phoneme model which improves the recognition rate reasonably. The developed speech recognition system can readily recognise the unintelligible speech and display that into normal text. This can be a very useful assistive device for the hearing-impaired to communicate with their normal peers and to operate all voice-operated devices for command and control applications. Evaluation is also made to compare human and machine recognition accuracy, using ten isolated digits. It shows that, the overall accuracy is 26% by human compared to that of 92.4% by machine, which emphasises the need for the development of our computer-based speech recognition system.

Keywords: deaf children; hearing-impaired children; computer-based speech recognition; MFCC; mel-frequency cepstral coefficients; CMU sphinx toolkit; pronunciation errors; voice disorders; hearing impairment; Tamil language; India.

DOI: 10.1504/IJBET.2015.069402

International Journal of Biomedical Engineering and Technology, 2015 Vol.17 No.4, pp.356 - 370

Received: 16 Jun 2014
Accepted: 06 Nov 2014
Published online: 14 May 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article