Gender classification through acoustic analysis
Sonia Maria D’Souza, Ramesh Prasad Ramanujam, M Reddy Ugesh, Mohammed Faisal, Mitrani Paul, Ganga D Benal · 2025
The paper aims to use Acoustic Analysis for Gender Classification and to use Machine Learning Algorithm for Gender classification then compare it to produce Accuracy. This dataset contains audio recordings while Classifier is built and tested with extracted features from the audio recordings. The algorithms used are Decision Tree Classifier, K-Neighbors Classifier, Logistic Regression, Support Vector Machine (SVM), SVM with RBF Kernel, and Gradient Boosting Classifier. The Decision Tree Classifier returns a tree-based model of the classification of gender based on the extracted acoustic features. K — Neighbors Classifier is used to predict gender based on the closeness of data points in the gender data. Logistic Regression provides a probabilistic model that maps the acoustic feature vector to a male/female classification. SVM (With and Without RBF Kernel) : It uses a hyperplane to separate the the gender classes in the feature space. The Gradient Boosting Classifier takes a number of weak learners and couples them together to make more accurate predictions. Performance of these algorithms in terms of accuracy, precision, recall and F1 score is then evaluated in the study. In addition, Receiver Operating Characteristic (ROC) curve represents a further determination of the discrimination of the classifiers. In this paper, we show that through the acoustic analysis provide results which data are separate with gender group and its classified accurately proposed appropriate method. To summarize, acoustic gender classification is successfully accomplished Rounding off the ensemble, the Gradient Boosting Classifier, which, besides producing better predictions as compared to a weak single learner, merely adds subtle complexity by marrying its weak single learners to together broadcast itself better, accentuating the power of ensemble methods acting harmoniously together in the pursuit of gender classification using acoustic properties. A thorough analysis of the performance of these strong algorithms is done through a set of metrics such as accuracy, precision, recall, and F1 score. In addition, the ROC curve is used as a standard method of assessing the discriminative power of the classifiers we analysed.