Towards a Toolkit for Free-Living Wearable Development

Towards a Toolkit for Free-Living Wearable Development
UbiComp/ISWC 2022

Blaine Rothrock¹
Alexander Curtiss¹
Juyang Bai¹
Josiah Hester²

Introduction

We are part of Normalizing Timing of Rhythms Across Internal Networks of Circadian Clocks (NTRAIN) DARPA funded project team, responsible for an external wearable hub (extHub) development that enable an implant to release pharmacy to person in need.
The extHub is a battery powered wearable device that provides power, communication, and user interface to monitor and control the implant. Meanwhile, the extHub serves as the gateway for all data and control, and comprises a multi modal sensor suite to understand human behavior, such as heart rate and body temperature. The extHub connects to an phone application that provide an interface for user to input and control. Figrue 1 is the architecture of extHub development.

Hardware & Firmware

Hardware

The extHub is centered around a number of hardware devices (Figure 2): printed circuit boards with necessary sensors, energy management, communication encapsulated in a comfortable sleeve, which securely communicates with a phone application. extHub_hardware The conductor board (Figure 3) contains the processor, buttons, status LEDs, RTC, battery, haptic motor contacts, and sleeve connector on the top side. The bottom side contains multi-wavelength PPG sensor used to calculate heart rate, accelerometer used to measure people movement, and FRAM. extHub_conductor The sleeve interface board (SIB) contains breakout connectors for external sensors, such as skin temperature sensor and ECG electrodes. The SIB also connects to the main battery to boost it up for the rest of system. extHub_sib The conductor board case enclose the conductor board, haptic sensor and battery.

Firmware

We use Nordic nRF5340 system on chip (SoC) running Zephyr real-time operating system (RTOS) to develop extHub. We develop drivers for hardware componets and sensors on board or external for data collection and interact with user. We also develop BLE driver with extensive embedded C/C++ API provided by Zephyr for the communication between extHub and phone application.

Demo

Data Science

The extHub is responsible for running sensing algorithm for motion/activity information, heartrate, skin temperature and body temperature. PTT: We collect PPG and ECG data. L2norm: We collect accelerometry data from an onboard ISM330 inertial measurement unit (IMU) to infer human activity information using L2norm.

Toolkit for Free Living Wearable Development

We propose a toolkit for free-living wearable development, which includes a wearable device, a mobile app, and a cloud platform.
(Please check our paper "Towards a Toolkit for Free Living Wearable Development" for more information.)

Citation

@inproceedings{10.1145/3544793.3560399,
author = {Rothrock, Blaine and Curtiss, Alexander and Bai, Juyang and Hester, Josiah},
title = {Towards a Toolkit for Free Living Wearable Development},
year = {2023},
isbn = {9781450394239},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3544793.3560399},
doi = {10.1145/3544793.3560399},
abstract = {Real-world Data collection and analysis is a significant pain point in wearable device development, requiring multidisciplinary skills in: embedded systems, application development, data science, and domain expert knowledge. In this work, we first build motivation based on previous experiences in wearable development, then introduce a toolkit for data collection and iterative development to reduce engineering efforts for free living experimentation of wearable devices. This toolkit utilizes Bluetooth Low Energy and an adaptive mobile application to help researchers quickly test new hardware, collect meaningful data, and assist in developing embedded algorithms with minimal intermediary code changes. We demonstrate the utility of our toolkit by collecting data in-the-wild from multiple sensors using a prototype wearable and a ground truth heart rate sensor. In addition, we demonstrate the toolkit’s capabilities with a baseline throughput test. Finally, we show how this tool has helped in early development of a new custom device. Our work is released as open source and welcomes contributions in an effort to broaden the tookit’s utility for the wearable research community.},
booktitle = {Adjunct Proceedings of the 2022 ACM International Joint Conference on Pervasive and Ubiquitous Computing and the 2022 ACM International Symposium on Wearable Computers},
pages = {379–385},
numpages = {7},
keywords = {Wearables, Machine Learning, Health},
location = {Cambridge, United Kingdom},
series = {UbiComp/ISWC '22 Adjunct}
}