Uso de la transformación de Fourier de orden fraccional al determinar los coeficientes cepstral en las frecuencias mel para la verificación de locutores
Edgar F. Maldonado Orduz, David Daniel Bertel Mendoza, Yezid Torres Moreno · Entre ciencia e ingeniería · 2013
The voice is a natural biometric attribute for the verification and recognition of speakers. Using the Fourier transform of fractional order, it is possible to obtain characteristics of the speech signal in the time-frequency space with an added degree of freedom. The speech is represented in a different way compared to the traditionally Fourier transformation. This paper compares the performance of using the fractional Fourier transformation in the speakers’ verification in text-dependent systems, instead of the common representation using the standard Fourier transformation, of the Mel Frequencies Cepstral Coefficients. The results show that, for an appropriate choice of the order of the fractional Fourier transformation, an improvement has been obtained in the verification of the speaker.