Article
Human Activity Classification Using Multilayer Perceptron
Ojan Majidzadeh Gorjani *
,†
, Radek Byrtus , Jakub Dohnal , Petr Bilik , Jiri Koziorek
and Radek Martinek
Citation: Majidzadeh Gorjani, O.;
Byrtus, R.; Dohnal, J.; Bilik, P.;
Koziorek, J.; Martinek, R. Human
Activity Classification Using
Multilayer Perceptron. Sensors 2021,
21, 6207. https://doi.org/
10.3390/s21186207
Academic Editor: Ki H. Chon
Received: 18 August 2021
Accepted: 8 September 2021
Published: 16 September 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Department of Cybernetics and Biomedical Engineering, Faculty of Electrical Engineering and Computer Science,
VSB—Technical University of Ostrava, 700 30 Ostrava, Czech Republic; radek.byrtus@vsb.cz (R.B.);
jakub.dohnal@vsb.cz (J.D.); petr.bilik@vsb.cz (P.B.); jiri.koziorek@vsb.cz (J.K.); radek.martinek@vsb.cz (R.M.)
* Correspondence: ojan.majidzadeh.gorjani@vsb.cz; Tel.: +420-597-325-856
† Current address: Faculty of Electrical Engineering and Computer Science, VSB—Technical University of
Ostrava, 708 00 Ostrava, Czech Republic.
Abstract:
The number of smart homes is rapidly increasing. Smart homes typically feature functions
such as voice-activated functions, automation, monitoring, and tracking events. Besides comfort
and convenience, the integration of smart home functionality with data processing methods can
provide valuable information about the well-being of the smart home residence. This study is aimed
at taking the data analysis within smart homes beyond occupancy monitoring and fall detection.
This work uses a multilayer perceptron neural network to recognize multiple human activities from
wrist- and ankle-worn devices. The developed models show very high recognition accuracy across
all activity classes. The cross-validation results indicate accuracy levels above 98% across all models,
and scoring evaluation methods only resulted in an average accuracy reduction of 10%.
Keywords:
human activity recognition; artificial neural network (ANN); intelligent buildings (IB);
smart home (SH)
1. Introduction
The availability and affordability of smart home technology have driven the rapid
increase in the number of smart homes. Typically, smart home technologies enable voice-
activated functions, automation, monitoring, and tracking events such as the status of
windows and doors, entry, and presence detection. Besides comfort and convenience, the
integration of smart home functionality with the Internet of Things (IoT) and other com-
munications systems creates new possibilities for assisting and monitoring the well-being
of aged or disabled people [
1
]. In particular, activity recognition within smart homes can
provide valuable information about the well-being of the smart home residence. Such
information can be utilized to automatically adjust the ambient conditions of the rooms
with the use of heating, ventilation, and air conditioning (HVAC). Another use of this infor-
mation could be the detection of irregularities within the residence’s activities that indicate
that assistance is required or a medical emergency. In general, human activity recognition
systems can be applied to many fields, such as assisted living, injury detection, personal
healthcare, elderly care, fall detection, rehabilitation, entertainment, and surveillance in
smart home environments [2].
In general, human activity recognition is formulated as a classification problem. It is an
important research topic in pattern recognition and pervasive computing [
3
]. A significant
amount of literature concerning machine learning techniques has focused on the automatic
recognition of activities performed by people and the diversity of approaches and meth-
ods to address this issue [
4
,
5
]. Minarno et al. [
6
] compared the performance of logistic
regression and support vector machine to recognize activities such as lying down, standing,
sitting, walking, and walking upstairs or downstairs. Guan et al. [
7
] tackled this issue
using wearable deep LSTM learners for activity recognition. Ramamurthy et al. [
8
] noted
that deep learning methods applied to human activity recognition commonly represent the
Sensors 2021, 21, 6207. https://doi.org/10.3390/s21186207 https://www.mdpi.com/journal/sensors
