Acoustic Analysis of Speech for Gender and Age Classification Using CNN and Machine Learning Techniques

Age and gender classification from voice recordings has significant applications in personalized marketing, human-computer interaction, and speech processing.Recent advancements in this field primarily employ neural networks and traditional machine learning techniques to enhance prediction accuracy.However, several gaps remain, particularly in the robustness and generalizability across diverse languages and datasets, as many existing models rely heavily on manual feature engineering and specific data preprocessing steps.This study proposes a novel convolutional neural network (CNN) architecture that directly learns from raw voice data, bypassing the traditional reliance on manual feature engineering.The main contributions include the introduction of an advanced CNN model designed to enhance feature learning directly from the speech waveform, the implementation of a more generalized system capable of operating effectively across different languages, and the validation of the proposed model against a comprehensive set of metrics.The proposed CNN architecture captures both local and global dependencies in speech data and employs an end-to-end learning approach that directly processes raw voice signals.The proposed CNN model consistently outperformed traditional machine learning approaches, achieving an accuracy of 88.1% for age classification and 96.9% for gender classification.The model's ability to learn discriminative features directly from raw audio data and its generalizability across different languages make it a promising solution for voice-based demographic classification tasks.