An SVM Based Speaker Independent Isolated Malayalam Word Recognition System
P Thushara, C Gopakumar · 2014
As technology advances man-machine interaction is becoming an unavoidable activity. So an effective method of communication with machines enhances the quality of life. If it is able to operate a system by simply commanding, then it will be a great blessing to the users. Speech is the most effective mode of communication used by humans. So by introducing voice user interfaces the interaction with the machines can be made more user friendly. This paper implements a speaker independent speech recognition system for limited vocabulary Malayalam Words in Raspberry Pi. Mel Frequency Cepstral Coefficients (MFCC) are the features for classification and this paper proposes Radial Basis Function (RBF) kernel in Support Vector Machine (SVM) classifier gives better accuracy in speech recognition than linear kernel. An overall accuracy of 91.8% is obtained with this work.