Authors:
Chinazunwa Uwaoma
and
Gunjan Mansingh
Affiliation:
The University of the West Indies, Jamaica
Keyword(s):
Smartphone, Machine Learning, Algorithms, Respiratory, Sound Analysis, Classification, Symptoms.
Related
Ontology
Subjects/Areas/Topics:
Biological Inspired Sensors
;
Computer Vision, Visualization and Computer Graphics
;
Human-Machine Interfaces
;
Image and Video Analysis
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Machine Learning in Control Applications
;
Robotics and Automation
;
Signal Processing, Sensors, Systems Modeling and Control
;
Time-Frequency Analysis
Abstract:
This paper explores the capabilities of mobile phones to distinguish sound-related symptoms of respiratory conditions using machine learning algorithms. The classification tool is modeled after some standard set of temporal and spectral features used in vocal and lung sound analysis. These features are extracted from recorded sounds and then fed into machine learning algorithms to train the mobile system. Random Forest, Support Vector Machine (SVM), and k-Nearest Neighbour (kNN) classifiers were evaluated with an overall accuracy of 86.7%, 75.8%, and 88.9% respectively. The appreciable performance of these classifiers on a mobile phone shows smartphone as an alternate tool for recognition and discrimination of respiratory symptoms in real-time scenarios.