Gender Determination in Human Voice Signals using Synaptic Efficacy Function-based Leaky Integrate and Fire Neuron Model

Abstract

Today's technological advances have made it possible for people to determine the gender of the speaker from an audio signal. Numerical data such as frequency types, spectral and entropy constitute acoustic information of audio signals. Recently, artificial intelligence-based learning models with high success rates have started to attract attention in various fields. There are many studies on deep learning models on audio signals. In this study, spiked neural networks with a different architectural structure, inspired by deep learning models, were used. The dataset used in the study consists of parameters based on acoustic information including human speech and voices. By using the determined data set, the spiked neural network model was trained and gender determination was achieved. As a result, 98.84% overall accuracy success was achieved in the classification process in this proposed study. With the experimental analyzes carried out in this study, it was observed that the spiked neural network model was successfully run and high performances were obtained.