RESEARCH ARTICLE


Modelling Human Activity using Smartphone Data



Suriya Badrinath1, Raja Muthalagu1, *
1 Department of Computer Science, Birla Institute of Technology & Science Pilani – Dubai Campus, UAE


Article Metrics

CrossRef Citations:
0
Total Statistics:

Full-Text HTML Views: 468
Abstract HTML Views: 245
PDF Downloads: 220
Total Views/Downloads: 933
Unique Statistics:

Full-Text HTML Views: 280
Abstract HTML Views: 139
PDF Downloads: 135
Total Views/Downloads: 554



Creative Commons License
© 2021 Badrinath and Muthalagu

open-access license: This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International Public License (CC-BY 4.0), a copy of which is available at: https://creativecommons.org/licenses/by/4.0/legalcode. This license permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

* Address correspondence to this author at Department of Computer Science, Birla Institute of Technology – Science Pilani – Dubai Campus, UAE; E-mail: raja.m@dubai.bits-pilani.ac.in


Abstract

Background:

Over time, multichannel time series data were utilized for the purpose of modeling human activity. Instruments such as an accelerometer and gyroscope which had sensors embedded in them, recorded sensor data which were then utilized to record 6-axes, single dimensional convolution for the purpose of formulating a deep CNN. The resultant network achieved 94.79% activity recognition accuracy on raw sensor data, and 95.57% accuracy when Fast Fourier Transform (FFT) knowledge was added to the sensor data.

Objective:

This study helps to achieve an orderly report of daily Human activities for the overall balanced lifestyle of a healthy human being.

Methods:

Interfacing is done using Arduino Uno, Raspberry-Pi 3, heart rate sensor and accelerometer ADXL345 to generate real time values of day-to-day human activities such as walking, sleeping, climbing upstairs/downstairs and so on. Initially, the heart pulse of our four tested individuals is recorded and tabulated to depict and draw conclusions all the way from “Low BP” to “Heavy Exercise”. The convolution neural network is initially trained with an online human activity dataset and tested using our real time generated values which are sent to the MAC OS using a Bluetooth interface.

Results:

We obtain graphical representations of the amount of each activity performed by the test set of individuals, and in turn conclusions which suggest increase or decrease in the consistency of certain activities to the users, depicted through our developed iOS application, “Fitnesse”.

Conclusion:

The result of this works is used to improve the daily health routines and the overall lifestyle of distressed patients.

Keywords: Convolution neural network, Arduino, Raspberry Pi, IOS application, Human activity recognition, Smart phone.