Emotion Speech Recognition for Safety Driving

Liew, Soon Seng (2024) Emotion Speech Recognition for Safety Driving. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Full Text-Liew Soon Seng.pdf Restricted to Registered users only Download (5MB)

Abstract

The goal of this project is to develop a speech emotion recognition system for improving safety while driving. The system will analyse the driver's speech patterns and recognize emotions such as anger, fearful, and sad, which are common causes of accidents. The system will use machine learning algorithms to accurately identify emotions in real-time and provide appropriate feedback to the driver, such as alerts drivers to take a break. The system will be developed using a large dataset of recorded speech samples and will be trained and tested using state-of-the-art techniques. The process of the system involves database building including IV sound recording with native speakers, using the extraction feature MFCC, Mel and Chroma to frame the input data of recording, training the classifiers and testing, troubleshooting and final result evaluation. The result of this system is expected to reach an accuracy between 80-90%. However, there are many uncertainties such as noise, human issues and limited resources will affect the final result which produce final accuracy of 78.74%. The successful implementation of this system will lead to increased safety on the roads and a reduction in accidents caused by emotional driving.