Wavelet-Based Filtration Procedure for Denoising the Predicted CO₂ Waveforms in Smart Home within the Internet of Things

• Main Authors: Jan Vanus, Klara Fiedorova, Jan Kubicek, Ojan Majidzadeh Gorjani, Martin Augustynek
• Format: Article
• Language: English
• Published: MDPI AG 2020-01-01
• Series: Sensors
• Subjects: intelligent buildings; wavelet transformation; prediction; artificial neural network; multilayer perceptron; cloud computing; internet of things; smart home
• Online Access: https://www.mdpi.com/1424-8220/20/3/620
• ISSN: 1424-8220

The operating cost minimization of smart homes can be achieved with the optimization of the management of the building’s technical functions by determination of the current occupancy status of the individual monitored spaces of a smart home. To respect the privacy of the smart home residents, indirect methods (without using cameras and microphones) are possible for occupancy recognition of space in smart homes. This article describes a newly proposed indirect method to increase the accuracy of the occupancy recognition of monitored spaces of smart homes. The proposed procedure uses the prediction of the course of CO₂ concentration from operationally measured quantities (temperature indoor and relative humidity indoor) using artificial neural networks with a multilayer perceptron algorithm. The mathematical wavelet transformation method is used for additive noise canceling from the predicted course of the CO₂ concentration signal with an objective increase accuracy of the prediction. The calculated accuracy of CO₂ concentration waveform prediction in the additive noise-canceling application was higher than 98% in selected experiments.