A Comparative Study of Recognition Technique Used for Development of Automatic Stuttered Speech Dysfluency Recognition System

Objectives: This paper is an attempt to compare the work done around the world for development of stuttered speech database and approaches for analysis of stuttered speech and recognition system. Methods/Statistical Analysis: In particular we have compared the different methods adopted by the researchers around the world for development of speech database and the techniques implemented on these developed databases. We have compared the databases on the basis of utterances, gender, age group, speech dysfluencies and type of samples. The recognition systems are compared on the basis of feature used, classification techniques and the accuracy. Findings: Speech recognition based application is getting more popularized and now being implemented at various places. However, the developed speech recognition systems cannot handle the speech dysfluencies. Very less work had been carried out till date for stuttered speech recognition system. The work for Indian languages is very negligible. The only work carried out is for Kannada. There is no major contribution for other Indian Languages. This paper shows the current status and the notable work carried in other languages. Application/Improvements: There is a need to develop more such systems for other Indian languages which will be very helpful for multilingual society like India.