A human-centered Web-based tool for the effective real-time motion data collection and annotation from BLE IoT devices



Bardoutsos, Andreas, Markantonatos, Dimitris, Nikoletseas, Sotiris, Spirakis, Paul G ORCID: 0000-0001-5396-3749 and Tzamalis, Pantelis
(2021) A human-centered Web-based tool for the effective real-time motion data collection and annotation from BLE IoT devices. In: 2021 17th International Conference on Distributed Computing in Sensor Systems (DCOSS), 2021-7-14 - 2021-7-16.

[img] PDF
ISIoT_2021_paper_5.pdf - Author Accepted Manuscript

Download (682kB) | Preview

Abstract

The effective utilization of real-world data is an integral part of any IoT monitoring or AI-assisted system. Thus, data collection and annotation is an important step towards the successful development and realization of such systems. Nevertheless, in order to create reliable datasets, current data collection and annotation methodologies often require a controlled environment while also the presence of the volunteer contributing to the process, or any subject for that matter, and an expert, monitoring the procedure, is mandatory. These processes are heavily restrained by the recent COVID-19 pandemic outbreak.To address such issues, in this paper we propose a human-centered Web-based dataset creation and annotation tool that utilizes the Web Bluetooth API. The user can effectively collect gestures from a nearby device that supports the BLE protocol, assign tags to the collected data, and store them remotely, in real-time. The data storage, as well as its annotation, can also be performed remotely by an expert stakeholder. An off-the-shelf wearable sensorial device has been used indicatively for our tool demonstration purposes. To the best of our knowledge, this is the first attempt that exploits the Web Bluetooth API capabilities for the development of a Browser-based real-time data collection, storage, and annotation tool. Our tool can be also expanded to other applications that use the sensing device with only minor configuration changes and is also operable through any smart-device that supports a Web-Browser. Furthermore, our tool's performance matches that of native applications'. Finally, the tool is successfully deployed and validated by integrating it into our ongoing ML platform that is related to allergic rhinitis gesture recognition.

Item Type: Conference or Workshop Item (Unspecified)
Uncontrolled Keywords: the Internet of Things (IoT), Web Bluetooth API, dataset creation, machine learning, mHealth, eHealth, data annotation
Depositing User: Symplectic Admin
Date Deposited: 21 Nov 2022 14:52
Last Modified: 17 Mar 2024 13:27
DOI: 10.1109/DCOSS52077.2021.00067
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3166304