Acoustic sound classification based on artificial neural network

Abstract

The project was motivated by the needs of mobile audio classification system, which can differentiate acoustic events in different genres. The aim of this project is to create an Android mobile application with the capability of detect and classify different types of sounds for surveillance purposes. Research has been done upon various methods and technologies on acoustic classification. It was identified that acoustic event classification did not receive as much attention as human speech recognition. However the basic principles and technologies used to develop the classification systems are similar. The combination of MFCC feature set and Artificial Neural Network classification method was selected as the main methodology of this project. The project was accomplished in the standard Software Development Life Cycle (SDLC), where planning, designing, implementation, testing and optimization were carried out during the project period. The outcome of this project was satisfactory, yet improvements can still be made to enhance the accuracy and performance of the system. This report will cover the research and develop process of the project and will also give suggestions for the future development.