Speech Emotion Recognition

Abstract

This Project Looks At How To Create An Engaging Single-player Game Stressing Smooth Controls And Real-time Interaction Developed In Unity Game Engine. A Responsive Gaming Experience Is Possible For Players Who Can Move Smoothly Between Various States Including Walking, Running, Crouching, And Standing Still. The Game Features A Countdown Timer Indicating The Conclusion Of A Match And A Restart Choice, As Well As Systems For Managing Enemy Spawning And Scoring. Weapons Make Fight To Seem Dynamic And Varied By Allowing Both Semi-automatic And Full-automatic Shooting Modes. This Research Proposes A Real-time Speech Emotion Recognition (SER) System That Classifies Human Emotions From Audio Input Using A Machine Learning Pipeline. The System Utilizes The RAVDESS Dataset For Training And Extracts Acoustic Features Such As Mel Frequency Cepstral Coefficients (MFCC), Chroma And Spectral Contrast. A Multilayer Perceptron (MLP) Classifier Is Used For Emotion Prediction, Recognizing Eight Distinct Emotional States. Real-time Audio Can Be Recorded And Analyzed By The Model, Which Adaptively Improves Itself Using High Confidence Predictions. The Proposed System Is Capable Of Dynamic Learning, Thus Continuously Enhancing Performance Over Time. This Approach Facilitates The Integration Of Emotional Intelligence In Applications Such As Virtual Assistants ,mental Health Monitoring, And Interactive Voice-based Systems.

Keywords