Voice Frequency-Based Gender Classification Using Convolutional Neural Network

Abstract

Voice-based gender classification is a fundamental task in speech processing with applications in biometric systems, human–computer interaction, and demographic analytics. This study proposes an efficient deep learning approach for gender recognition using frequency-domain acoustic features. A structured dataset comprising 21 statistical voice features is employed to train a one-dimensional Convolutional Neural Network (1D CNN). Unlike conventional approaches that rely on raw audio or spectrogram representations, the proposed method operates directly on pre-extracted frequency features, thereby reducing computational complexity while preserving discriminative information. The dataset is preprocessed through label encoding and feature standardisation before being fed into the model. The trained network achieves a test accuracy of 97.00% and an AUC score of 1.00, demonstrating excellent classification capability. The results confirm that structured acoustic features, when combined with deep learning, provide a robust and scalable solution for voice-based gender classification.