Response of different window methods in speech recognition by using dynamic programming

In the area of pattern recognition, feature vectors identification is one of the major tasks to make this detection successful. For this reason, feature vectors depict the key distribution of data and by applying the various data analysis techniques, these vectors are obtained. In this paper, for identifying the good feature of speech signal, Mel Frequency Cepstral Coefficients (MFCC) are derived by applying the various window functions. After getting the feature vectors as MFCC, dynamic programming (DP) is used to get the distance for feature matching. In nutshell, to recognize individual speech, this paper presents the performances of various window techniques to make comparison between different voices by using (DP) algorithm. The responses of various types of window techniques including their energy spectrum are analyzed for extracting feature vectors from different speeches. Then this paper reports a DP based speech recognition technique. Result shows that for rectangular window based recognition, the feature extraction is very good and outperforms for word recognition in terms of low DP distance compared to other window methods.