I've always had an interest in how the various electronics that I use everyday work. As my college career progressed, I've developed an interest in how electronics and technologies are applied for use in Space. I have a passion for space exploration and hope to one-day join the European Space Agency with the skills I've obtained from IADT.
This project is designed to allow a user to input an audio clip where noise and distortion is present but unwanted. Using techniques that apply to the field of Machine Learning (principally Linear Regression and Gradient Descent) the system attempts to reduce the noise that is present in the audio clip. This project was implemented using MATLAB. The technique of Gradient Descent was used to intelligently tune a digital audio filter so that noise was removed. Using Machine Learning, the system does not need to know anything about the audio it is about to process and it creates a unique system, with no human intervention, that removes the noise present.
The main objective of this project was to create a Machine Learning system with the ability to create a digital filter that could remove noise present in degraded audio clips. Applications include audio restoration, noise control in industrial environments and music technology.
The project was successful in reducing distortion in several different types of audio clips. Very large filter systems were synthesised and noise levels were reduced significantly, without any human intervention. The principle of using Machine Learning for audio restoration was successfully demonstrated.
My Thesis is an extensive report on the research that I undertook for the project, the design and development of the project and the testing and implementation of the project. The aim was to develop a series of software modules using Machine Learning (ML) techniques such as Linear and Logistic Regression and Gradient Descent. The application of this project was in the field of Audio Processing, whereby a user can restore or enhance (depending on the function that the user desires) audio signals to improve the quality of the audio or restore an audio signal that has been degraded.
Machine Learning is “a type of artificial intelligence in which computers use huge amounts of
data to learn how to do tasks rather than being programmed to do them”. (Oxford Dictionary)
I developed an ML system that can restore, enhance and process audio to a point where the user only has to input the audio file through the ML interface and choose what they want to do with the file (restore, enhance) and let the ML do the processing without with no further user interaction required.