The student (Sara Qusay Salah Eddin) obtained her master's degree from the Department of Computer Sciences / University of Technology for her thesis (Audio Classification based on Polynomial Interpolation) In this thesis, two main algorithms were proposed for gender classification based on sound. A new algorithm was used to classify between Gender based on self-association. In the beginning, the signal is reworked to fill the blanks between the signals, then a convoluted signal is made to a cross-correlated signal. Finally, gender is classified by cross-correlation. The sound is defined as a male voice if the band is at the volume value between 85 Hz and 155 Hz, while the female voice is selected if the band is in the volume value between 165 Hz and 255 Hz. The 2270 sample audio data (including 1132 female and 1138 male) is used to evaluate the proposed algorithms. The result of the algorithm is obtained 99% accuracy rate and 33% call rate when using the standard sound. The result is when using real audio 83% accuracy rate, the call rate is 1.0%.
It was on Monday 30/4/2018 at the discussion hall the department. The discussion committee was composed of (Dr. Matheel Emad Alddin Abdel Moneim and Dr. Sakina Hassan Hashem) from the University of Technology / Computer Sciences Department and (Dr. Raghad Abdel Aziz) from the University of Baghdad / Faculty of Education Ibn Rushd and the presence of the supervisor of the student (Dr. Nidaa' Fleih Hassan).
|
Source: Department's Media |
