A spectral feature process for speech recognition using HMM with MFCC approach

Speech recognition using technology has always been a uphill task. Researchers have been trying since decades to replicate the acoustic system to improve speech recognition by machines. In this paper we tried to improve the feature extraction by modeling it in accordance with the non-linearity of the human audio perception. The human ear is not equally sensitive to all the frequencies of sound within the entire spectrum. From the acoustic information, the speech recognition system tries to extract linguistic information, i.e. to recognize the speaker´s utterance. The speech feature vectors are called observations. The Simulation results show an improvement in the quality metrics of word recognition with respect to computational time, learning accuracy for a speech recognition system.