Isolated Word Recognition using Polynomial Classifier

This paper presents an isolated word recognition using polynomial classifier. Along with the high accuracy, speech recognition applications also required the low complexity and less storage space, which is achieved using the polynomial classifier. Speech features used are the well-known Mel-Frequency Cepstral Coefficient (MFCC). The performance of the said classifier is tested for MFCC of size 12 to 22 and the best one is selected for the further analysis. The effect of % overlap between the two frames is also evaluated. We also provide the performance comparison of polynomial classifier with the other classifiers like Vector Quantizer (VQ) and Dynamic Time Warping (DTW). The recognition using polynomial classifier is found faster than the VQ and DTW and also requires less storage space, however it is found that the recognition rate using polynomial classifier is slightly less than the two.