Language Agnostic Voice Recognition Model

Abstract

Voice recognition is the ability of a machine to identify a person based on their unique voiceprint. As this task is becoming more important and dominant in everyday people’s lives, this paper is testing different approaches for its implementation. Using a multilanguage database and working with the different frequencies’ characteristics, five machine learning models such as Random Forest, XGBoost, MLP, SVM and Gradient Boosting, along with CNN deep learning model were implemented. The models were trained on three different tasks, gender prediction, age range prediction, and combined gender and age range prediction. These models were evaluated using accuracy, F1-score and MCC score. The results showed that Random Forest outperforms other models by achieving an accuracy of more than 0.9 for all the three classification tasks.