Machine learning algorithm to learn and predict aircraft/ engine type using engine noise data

This report serves as illustrative purpose of learning experiences and project executions which author had undertaken during AY2019/20 for Final Year Project (FYP). It summarized the invaluable knowledge and important technique acquired through accomplishing various task associated with FYP. Remote...

Full description

Saved in:
Bibliographic Details
Main Author: Jimmy
Other Authors: Sameer Alam
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2020
Subjects:
Online Access:https://hdl.handle.net/10356/140654
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:This report serves as illustrative purpose of learning experiences and project executions which author had undertaken during AY2019/20 for Final Year Project (FYP). It summarized the invaluable knowledge and important technique acquired through accomplishing various task associated with FYP. Remote Control Tower (RTC) serves as a cost saving alternative for small regional airports. Aircraft Identification System within RTC is expected to minimize the workload of air traffic officers and increase their spatial awareness. Current aircraft identification research focuses heavily on visual based aircraft classification through computer vision. Author then proposed an auditory aircraft identification system through machine learning model. Author conducted both onsite and online aircraft sound data collection. The audio data undergone various audio signal process such as filtering and audio event detection to remove noise within the audio files. Several audio spectral features such as Mels Coefficient, Spectral Contrast were extracted from the processed audio files. Author proceed to construct several sound classification deep learning models. These models were trained using spectral features and were able to make prediction of aircraft type based on the aircraft sound. Time delay Neural Network (TDNN) model was the best performing model with F1-score of 0.67 on 3 different aircraft class namely A320, B738 and B77W.