Acoustic Information for Multisensory Occupancy Detection using Machine Learning
Loading...
URL
Journal Title
Journal ISSN
Volume Title
Sähkötekniikan korkeakoulu |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2021-12-13
Department
Major/Subject
Autonomous Systems
Mcode
ELEC3055
Degree programme
Master's Programme in ICT Innovation
Language
en
Pages
89
Series
Abstract
In this thesis, the combination of passive infrared (PIR) and audio sensors for indoor room occupancy detection was demonstrated, with a special focus on audio signal processing and feature extraction using Machine Learning techniques. The lighting automation industry uses extensively PIR or motion sensors for presence detection due to their simplicity and low cost, but their small sensitivity for subtle movements makes them inefficient in certain situations. The occasional undesired behavior is often compensated with long timeout values but this leads to excessive energy consumption. We are presenting a computationally efficient audio feature extraction method and using exhaustive grid search for Machine Learning model architecture optimization. The best-performing models provided the basis for the embedded implementation of a real-time occupancy device, which combines audio and PIR information for improved presence detection. The prototype performance was measured against the traditional PIR based solutions demonstrating lower false-negative predictions with shorter timeout values.Description
Supervisor
Bäckström, TomThesis advisor
Kallas, JuhaKeywords
deep learning, digital signal processing, embedded AI, lighting control, presence detection