A Holistic Cyber-Physical System to Detect Anomalies in Smart Home Appliances
No Thumbnail Available
Links to Files
Permanent Link
Collections
Author/Creator
Author/Creator ORCID
Date
2017-01-01
Department
Information Systems
Program
Information Systems
Citation of Original Publication
Rights
This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu
Distribution Rights granted to UMBC by the author.
Distribution Rights granted to UMBC by the author.
Abstract
This theses focuses on detecting the anomalies of appliances using their acoustic signature and power consumption data. In this work, we advocate a holistic cyber-physical system (CPS) which helps recognize the sound and energy patterns generated by a variety of everyday appliances and provide better monitoring of them. We focus on the appliance health estimation by combining its acoustic signal and power consumption data, provide just-in-time feedback to the customer, and encourage them to take appropriate precaution to avoid any increase in electricity bill or any physical damage or loss. As every appliance creates a unique sound which is not audible to human ears, we captured those different frequencies sound which provides a lot of knowledge about the appliance'sworking condition. We also investigated the abnormal power consumption of an HVAC system so that we can detect malfunctions and potential safety concerns. This system eliminates the potential risk of various types of sensor failure which are used in a smart home. Finally, we correlated and established the relationship between appliance'spower consumption and acoustic signature. We extracted the features like fast Zero Crossing Rate (ZCR), Audio Energy, Energy Entropy, Spectral Centroid, Spectral Spread, Fourier transform (FFT), Small Time Fourier Transform (STFT), Mel-frequency Cepstral Coefficient (MFCC) from the acoustic signal. We trained and tested the model with different classifiers such as support vector machine, ensemble, J48 decision tree, OneR and showed up to 90% accuracy.