An Enhanced Speech Recognition Algorithm Using Levinson-Durbin, DTW and Maximum Likelihood Classification

In this paper, we applied techniques such as Levinson-Durbin, DTW and maximum likelihood classification to achieve an enhanced speech recognition algorithm. Speech recognition has been adversely affected by noise and some other impairments factors making speech difficult to be recognized. Speech is distorted by a background noise and echoes, electrical characteristics. We used the combinatorial approach of Levinson-Durbin, DTW and maximum likelihood classification to develop a system for speech recognition. The system compares the speech with phonetics lattice and database of enrolled speeches from different speakers and output the enrolled speech with the recognition Id and name of the identified speech if it found a match. Otherwise error of unknown is returned to show speech mismatch. The system is able to distinguish two different speakers at any point in time using speech identity and phonetic pattern analysis