Energy Estimation between Adjacent Formant Frequencies to Identify Speaker´s Gender

This paper describes a method developed for gender identification that combines first three formants, pitch period, and the energy between each adjacent pair of formants. The information provided by the energy, formants, and pitch estimation is combined using a classifier to identify the gender of the speaker. All the parameters are determined for a voiced part of the speech samples. We used five male and five female speech samples /one, two, three, four, five, six, seven, eight, and nine/ for our training set. All speech samples are in clear mode. The sampling frequency is 8 KHz.