Improved MFCC and LPC algorithm for bundelkhandi isolated digit speech recognition
Abhishek Dixit, Abhinav Vidwans, Pankaj Sharma · 2016
Automatic speech recognition in different languages, spoken in different areas of any country, is one of the major research area in the field of signal processing. This paper presents an improved MFCC algorithm for Bundelkhandi digit speech recognition. Here speech digit features are extracted by using modified Mel Frequency Cepstral Coefficient algorithm (MFCC). In this modified MFCC algorithm, one additional filter, named as 1-D median filter, is used. Now this modified MFCC algorithm is applied on Bundelkhandi digits dataset to gather the speech features. Similarly 1-D Median filter is used with LPC algorithm to capture the Bundelkhandi digits dataset. Finally speech features, which are collected from modified MFCC and modified LPC algorithm, are combined and used as an input to back propagation neural network for recognizing particular digits. Experiments are performed on Bundelkhandi digit dataset (0 to 9), made with the help of sound recorder. These results show that proposed algorithm performance is better than simple MFCC and LPC algorithm.