Recognition of Marathi Isolated Spoken Words Using Interpolation and DTW Techniques

Abstract

This paper contains a Marathi speech database and isolated Marathi spoken words recognition system based on Mel-frequency cepstral coefficient (MFCC), optimal alignment using interpolation and dynamic time warping. Initially, Marathi speech database was designed and developed though Computerized Speech Laboratory. The database contained Marathi isolated words spoken by the 50 speakers including males and females. Mel-frequency Cepstral Coefficients were extracted and used for the recognition purpose. The 100% recognition rate for the isolated words have been achieved for both interpolation and dynamic time warping techniques.