Voice recognition and marking using mel-frequency cepstral coefficients

Abstract

A real-time voice recognition and marking system was developed in this study to automatically identify different voices of speakers. A microphone array was installed for audio reception. Pre-emphasis, framing and Hamming window, fast Fourier transform, mel-frequency, and mel-frequency cepstral coefficients with processing times of 0.001, 0.305, 0.205, 0.049, and 0.546 s, respectively, were used in the system. The total processing time was less than 1.5 s. Unique eigenvalues were obtained for each sound. The results indicated that the proposed system, which is an example of intelligent recording, can be used to automatically record speech in meetings or during classes.