This document proposes a system for classifying gender and emotions from human voice using machine learning algorithms. It involves preprocessing voice data, removing noise using hidden Markov models, extracting features using discrete wavelet transforms, and classifying gender and emotion using k-nearest neighbors. The system combines gender and emotion classification into a single system, achieving higher accuracy than systems that only classify one or the other. Evaluation shows the proposed system achieves 97% accuracy, outperforming existing systems with accuracies of 75-86%.