Enhanced Gender Recognition in Telugu Speech Using Augmented Transgender Audio

View at publisher Summarise with VIRAT
Authors: G. S. Pavan, Kalal Pavan Teja, Karella Udayram, Paluru Pavan Kumar, Madasamy Anand Kumar
Cited by: 0
References: 3

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This study presents a comprehensive approach to gender classification in Telugu audio using Mel-Frequency Cepstral Coefficients (MFCC) and pitch features. This work fills a critical gap in inclusivity within speech-based systems by expanding the problem to include transgender voices, building on earlier research that mainly concentrated on binary (male/female) classification. The proposed system has two primary objectives: robust preprocessing and feature extraction, and multi-model classification for gender prediction. Each audio sample was processed to extract MFCCs and pitch from a carefully selected and balanced dataset of male, female, and transgender voices that was standardized to 16 kHz. Five popular machine learning models—Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Multi-Layer Perceptron (MLP), and XGBoost—were trained and evaluated. Data augmentation techniques, including minor pitch shifting and noise addition, were applied to improve model generalization, only for limited transgender samples to reduce data imbalance. The results show that the integration of the MFCC and pitch features with augmentation significantly improved performance, achieving a total accuracy of over 97% in both both SVM and MLP models. This study underscores the effectiveness of feature fusion and data balancing in multi-class gender classification and promotes inclusivity in modern voice recognition systems.

Fields of Study & Subjects

Speech Recognition and Synthesis Music and Audio Processing Speech and Audio Processing

Publication details

Journal / source
International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
Pages
1–6
Publication date
1 February 2026
Language
EN