Efficient Gender Classifier for Arabic Speech Using CNN With Dimensional Reshaping
Abdulhamed Mohammed Jasim, Sohaib R. Awad, Fahad Layth Malallah, Jassim M. Abdul-Jabbar · 2021 7th International Conference on Electrical, Electronics and Information Engineering (ICEEIE) · 2021
Nowadays, gender detection is an open research in the field of soft biometrics. Human gender detection can be implemented depending on a voice or a speech. In this paper, an improved recognition rate of an individual gender detection is approached. The methodology starts with a proposed pre-processing technique, which is converting 1-dimensional (1D) into 3-dimensional (3D) data voice. Then the whole 3D voice samples are passed to the convolutional neural network for both training and testing (predicting) the class whether it is a male or a female. The number of the convolutional neural network (CNN) layers is 19 distributed into 4 rounds, in which each round contains sub-layers. The performance is extracted by using a database for Arabic speech that contains 6644 voice samples. The average result of the recognition rate is achieved up to 98.91%, which outperforms the state of the art using the same database.