A challenging task in recognizing the speech of the Hearingimpaired using normal hearing models in classical Tamil language

Journal of Engineering Research

jeya lakshmi¹  Revathi dhanabal² 

[1] K.RAMAKRISHNAN COLLEGE OF ENGINEERING;Sastra Univerisity

关键词: Mel frequency Perceptual linear predictive coefficients (MFPLP);    Vector quantization;    Clustering;    Centroids;    Recursive least square (RLS) filtering;    Speech recognition;    Hearing impaired (HI);    Normal hearing (NH).;   

DOI  :  

学科分类：社会科学、人文和艺术（综合）

来源: Kuwait University * Academic Publication Council

【 摘 要 】

Objective:We develop a system to recognize the speeches of Hearing impaired children using normal hearing models. Background:Hearing impaired speakers normally use sign language to communicate with others though they have perfect vocal structure. Their speeches are much distorted and may not be understood even by the teachers and parents. So, it is necessary to develop the system for recognizing their speeches especially in the native language and we have considered Tamil speaking hearing impaired persons in our work. Method:Performance of the system is analysed by applying the hearing impaired speeches directly on the models developed using the speeches of the normal speakers. This work mainly highlights the use of speeches of hearing impaired speakers for testing only and it is not necessary to create the database for hearing impaired which is difficult to create. Results: Using RLS filtering selected normal speech is subsequently applied to the models and performance is evaluated. Recognition accuracy is 100% for clustering model with cluster size 256. Conclusion:Recognized speeches will be heard and interpreted clearly by the normal speakers and the system can be used to help the hearing impaired at large and it helps to improve their status in society.

【 授权许可】

Unknown