What is that sound? An automated sound classification system using machine learning

• Main Author: Cher, Jensen
• Other Authors: Andy Khong W H
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Electrical and electronic engineering::Computer hardware, software and systems; Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/150286
• Institution: Nanyang Technological University

In modern times nowadays, the need for automation is becoming more prevalent as companies in the Information technology sector seek to produce automated solutions to improve their customers’ performance and productivity. One of the demands of today’s is the automation of audio classification, which can differentiate different genres of sounds. The project aims to create and investigate machine learning models and algorithms with the capability to detect and classify different types of sound based on their audio features. In this report, machine learning models such as Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and a combination of both Convolutional Recurrent Neural Network (CRNN) were applied as researches has been done upon CNN and RNN individually and combination of both, the researchers found the combination of both produces better results. A combination of the MFCC feature set and the machine learning models mentioned was chosen to be the main methodology idea for this project. The overall outcome of this project was satisfactory with an accuracy achieving 72% classification results on the CRNN model. Improvements to the dataset and the models can still be made to enhance and upgrade the models’ accuracy, efficiency, and performance.