UWE Bristol Engineering showcase 2015

Philip Jacobs
BEng (Hons) Robotics
Project Supervisor
Dr Praminda Caleb-Solly
Developing Internet of Things Enabled Smart Products to Support
Ambient Assisted Living
Sensor Mount
When designing the mount in CAD safety, usability / ergonomics, and data
reliability were the main concerns. The sensors are mounted away from the
areas of interaction with the cup, and can only be placed in the mount in one
orientation, ensuring the sensors don’t harm the usability nor change an
individual’s interactions with the cup. Sensor platform used is the Texas
Instruments Sensor Tag.
Sensor Selection
The side mounted accelerometer proved to be the most feature rich as such
was explored in terms of pre-processing options and classification. The
accelerometer also provides the best power consumption range at 10-135
μA, compared to the gyroscope at 5900 μA , and the magnetometer at 8.6 -
900 μA. Low battery consumption is a great benefit as the longer batteries in
the sensor tag can last the less human input the system needs.
Pre-processing
The performance of the pre-processing techniques was judged on the ability
to remove the high frequency noise / smooth the raw accelerometer data
from the sensor tag, and the ability to preserve the magnitude of peaks in
the data. A combination of these two properties will provide data which is
less affected by noise while still providing clear features in the form of the
magnitudes of peaks. Having evaluated moving average filter, Butterworth
filter and down sampling the data, down sampling was chosen, due to
minimal computational requirement and simplicity. In addition to reduced
power consumption.
Classification
Thresholding was the simplest technic devised in order to classify the cup
state; a simple method was desired to be applied to the real time system.
Having analysed the data it is clear there are
far more prominent peaks when the
cup is being drank from as opposed to,
lifted. 100% accuracy across all tested
was achieved. Graph to the left shows
experimental data with thresholds.
Feasibility of Tremor Detection
The majority of tremors have frequencies which are detectable with the
standard sensor tag and its max accelerometer sampling of 10 Hz. It was also
shown it would be possible to track tremor decline / improvement using the
accelerometer data collected by passing it through a Fourier Transform, and
looking at the frequency content of the signal.
Fourier Transform of
accelerometer data
without tremor (left)
and with (right)
Real Time Classification and System
Node Red was used to implement the real system which used the threshold
method for Classification.
Core System Functionality Requirements:
•Read and record sensor tag readings
•Reliably classify cup state in real time
•Record data relating to cup interaction
•Make recorded data viewable to a carer or other authorised party
•Publish classified cup state to other systems
•Be able to subscribe to updates on other
individuals cup state
•Trigger actuation based on received cup state
Additional Functionality:
•Threshold creation assistance
•SOS / panic button functionality
•Remind individual of drink if untouched
•Display dashboard type interface
In order to avoid any usability issues the only interface the user will have
with the system is via the physical artefacts / dashboard used, and sensor tag
readings.
In summary the system met all the requirements both core and additional
that were set. Resulting in a system capable of passively monitoring an
individual in their home, providing data to carers, and providing the
individual social support through ambient social communication. Lastly the
system has the potential to be deployed on a large scale given that
throughout development this has been considered.
Project summary
The aim of this project was to conduct research into
intelligent assistive technology with a view to
developing "smart" products, which might be useful
for diagnosing progressing age-related disabilities.
For this project we have designed and developed a
“smart cup” instrumented with sensors to monitor
aspects such as tremors and frequency and level of
fluid intake. The potential for social integration with
others using the smart cup has also been considered
to address the issues of social isolation.
Project Objectives
• Perform Participant studies to collected motion
data from natural cup interaction.
• Analyze data to determine best sensor and mount
position in addition to preprocessing and
classification technique.
• Reliably identify states cup when used by different
individuals, testing interpersonal, intrapersonal,
and context reliability.
• Implement real time system capable of monitoring
an individual in there own home in addition to
providing social support and information to carers.
• Finally the aim would be to work towards creating
two complete systems for a proof of concept and
deployment in the Ambient Assisted Living Lab.
Project Conclusion
The project has been a success having achieved the
goals set at the start. The amended project plan was
also kept to with work being completed to schedule.
In summary the system that has been created is of
genuine value, combining areas from the existing
types of assistive systems with social support. Two
systems will be deployed for demonstration within
the newly created Ambient Assisted Living area
within the Bristol Robotics Lab. Furthermore it shows
the power of not just the internet of things but
technology in general to help people, not just in a
medical sense but on an emotional / social level also.