English Language Speech Recognition using MFCC and HMM

The Recognition of speech is a process, which can be outlined as understanding of human speech, processing it into a machine-readable format and utilizing its realization for real time applications. English, the international language, holds the largest vocabulary among the languages, and it is often used for giving commands and for speech recognition in many areas. This paper implements and exploits Mel Frequency Cepstral Coefficient and Hidden Markov Model Techniques to evaluate the competence of speech recognition. This paper displays the Periodogram and Spectrogram of different English commands and commonly used English phrases which are observed after the evaluation of varying pitches for unique English phrases used as commands inserted using the graphical user interface. The result thus obtained displays differences of multiple users speaking commands and commonly used English phrases in multiple accents under unsupervised learning and partially known Hidden Markov Model.