Streaming wearable streaming and management platform (SWSP)

Project description

Prolonged or intense stress can lead to mayor physical and psychological health problems. Good stress management contributes to reducing these complaints.

The newest wearable sensors make it possible to measure biometrics that reflect stress levels reliably, but information is very scattered. The goal of this project is to create an online platform which can be used to process streaming data from different wearables and interpret the information to estimate stress levels in the wearer.

In addition to the creation of this SWSP platform research is done to encompass the ability to connect a big pool of existing wearables such as the Moodmetric ring or Movesense to the existing scalable system without the need for updates to the platform’s infrastructure and code.

The project also created an inclusive environment that encouraged the exploration of a new sensing system that can measure stress values.

Context

The project focuses on the domain of (digital or e-) health and well-being. Specifically focusing on stress management for people suffering of PPS or dementia. The key to effective stress measurement is preventative rather than reactive solutions to issues caused by prolonged stress. Active measurement of biomarkers that provide insights into (experiences emotional) stress can be used to proactively identify and address stress.

In recent years, e-health has been explored through wearables, but the selection and gathering of information from these devices lacks centralization and data-processing.

By combining information gathered from a selection of wearables a data pool can be created that offers stress estimation using an AI model. This creates a unified platform that can offer insights into the complexity of stress. Offering a more unified management opportunity for individuals, doctors, healthcare workers and people in care.

Results

Software back-end
This project runs on a microservices software architecture powered by the programming language developed by Google; Go. This architecture was specifically designed to be efficient at processing enormous amounts of data. In the entire processing pipeline, there are three services each serving their own critical role. The first service receives all the raw data from the wearables and stores the data. This service is the entry point and will see the most traffic. That is why it has no other responsibilities than storing data. The second service houses data processing algorithms. This is the only service executing actual logic. The third and final service is responsible for communicating the processed data to front-end applications.

Embedded communication
Communication is an important part of a wearable Because there are a lot of communication options, and the project contains challenging communication requirements it’s important to choose the communication protocol which is best suited for the different use cases. An overview of IoT related protocols was created in which 16 properties of 15 different communication protocols were research so the protocols can be compared to each other.

The overview is used to give advice for the three different use cases (Dementia Intramural, Dementia Extramural and PPS) which communication protocol is best suited. For the intramural use case BLE mesh is advised, for the extramural use case LTE-M with a side note to keep monitoring the progress of LoRa is advised and for the PPS use case BLE point-to-point is advised.

For the PPS use case a quick proof of concept was created. The proof of uses the Stress Patch from Kinetic Analysis as wearable, communicates via BLE and sends it data to the software back-end.

Embedded sensing
The Stress Patch proposed by Kinetic Analysis has seen significant improvements and enhancements. Sensing management, correction and validation have been researched and implemented ensuring higher reliability and accuracy of the system. Additional safeguards have been introduced using a watchdog alongside measurement frequency control and the continual calibration of skin heat flux measurements. This made the sensor more robust and dependable.

In addition to this a sensor that integrates the dielectric properties of particles in a medium to measure stress dependent particles in the blood using indirect dielectrophoresis (iDEP) has been modelled. This sensing method is proposed to offer better insights into the rise of stress-related hormones rather than measuring the results of prolonged hormone changes such as heartrate variability.

About the project group

The project group is part of the first iteration of the Master of Applied IT at Fontys. All three students have received their Bachelor’s in IT at Fontys: Matthijs and Femke in Hardware Interfacing, and Pieter Edoardo in Software Engineering.

This project was the focus of the first semester and thus we spend about 4 days a week working on the project. In this master, research takes the central stage, which meant a thorough analysis of the design challenge and possible solutions was required. This enabled the group to share fresh insights instead of just using existing ideas to provide a system.