Improving Accuracy in Human Voice Recognition for Gender Classification Using Machine Learning Algorithms

This study focused on improving the accuracy of human voice recognition, specifically for distinguishing between male and female voices, by comparing two algorithms. The Random Forest algorithm, known for its robust performance and recursive feature elimination capabilities, was used to enhance feature selection for voice classification. Our experiments involved a dataset of 112 voice samples, with the goal of classifying each sample as either male or female. The Random Forest algorithm achieved an impressive accuracy of 97.79 % in predicting gender based on voice. In comparison, the Novel Decision Tree algorithm, which was also optimized using the Random Forest method, yielded an accuracy of 96.74%. The study employed a G-power value of 80 % with 10 samples for analysis. Statistical tests revealed a significant difference between the two algorithms, with a p-value of 0.090, indicating that the Random Forest method is statistically superior in this context ($p<0.05$) with a 95% confidence interval. Hence further the Random Forest algorithm demonstrated significantly better performance than the Novel Decision Tree algorithm for gender voice recognition, making it a more effective choice for this application.