Mohammad Reza Hasanabadi
Mohammad Asgari

This article uses Gaussian Mixture Speaker Models known as GMM to detect the signal on its spectral shape. In Fact analyzing the signal information in some cases depends on its spectral shape, therefore using models such as GMM helps recover the information. This model has attained more than 90% identification accuracy using 30 seconds in a clean atmosphere and more that 80% accuracy in less conditions. The simulation is based on MATLAB GUI. The algorithm of this work is according to hidden markoy models (HMM) which provides a highly reliable approcach foe Speech recognition. This paper focuses on Farsi digits in isolated words form covering zero to ten. This system is able to recognize speech waveform by translating the speech waveforms into a set of vector features using Mel Frequency Cepstral Coefficients (MFCC) techniques.Keywordsextraction;modelsrecognizers were introduced. This field of studspeed to develop unhastened this development. Speech recognitiontranslation of spoken words into its equivalent text. It is alsocalledor simplyHealthcare, Military, Aerospace andTheDTW and neural networks.Generally Speech recognition is divided into speakerindependent and speakerused in this paper is based on speaker as reference pattern. Asillustspeech and viaalgorithmand consequently detects input digits.sections, the detailsintroduced.T

Speech recognition; Farsi digits; Signal feature extraction; Mel Frequency Cepstral Coefficients; Hidden markov models; Matlab GUI