Almost every person wears an activity
tracker, smart watch or GPS-enabled sports watch nowadays. These devices
provide information on how much a person has moved such as distance covered,
speed and heart rate.
But this information does not give an insight
as to the nature or quality of the movement, neither does it aid in minimising
an injury nor in training a new skill.
According to the report
released by Curtin
University, a PhD student from the Curtin School of Physiotherapy
and Exercise Science aims to better measure the training volume and specific
musculoskeletal loads in a cohort of female pre-professional ballet dancers by
building an automated human activity recognition system.
Ms Danica Hendry studies the contributing
factors towards pain and disability in dancers. Professional and
pre-professional ballet dancers have an intense physical training regime, which
can eventually lead to fatigue and overload injuries.
Recording and managing their physical
workload is completely subjective. This may document hours spent in training,
but do not take into account the frequency of specific movements and
musculoskeletal loads that may lead to injury.
Because of this, her project became a
collaboration that includes physiotherapists, biomechanists, and computer
scientists from Curtin University and Edith Cowan University (ECU).
The Curtin Institute for Computation, the
Curtin School of Physiotherapy and Exercise Science, the Civil and Mechanical
Engineering at Curtin, and the ECU’s Western Australian Academy of Performing
Arts (WAAPA) are all taking part in the project.
Since existing activity trackers cannot
distinguish a jeté (jump) or an arabesque (leg lift) from a plié (bending at
the knees), and do not record much when dancers train on one spot at the barre,
the automated human activity recognition system had to be built.
Ms Hendry explained that sensors can be
placed on the dancers but if the movement is not specifically identified, it is
considered useless data. To address this, the dancers were videotaped so that
it can be correlated against the sensor data.
Six sensors per dancer were used. Each
sensor incorporated an accelerometer, a gyroscope and a magnetometer. The
sensors were placed on the left and right shins, left and right thighs, sacrum
and thoracic spine to document movement as each dancer worked through specific
movements.
The continuous signals were then segmented
and manually cross-referenced against the video footage in order to connect specific
signal segments to individual dance movements.
Because each dancer is different, the research
team had to record 23 dancers from WAAPA as they worked through a sequence of
dance movements. The study focused on jumps as the force exerted on the body
during landing are implicated for lower limb injury. Leg lifts were also
observed as they are implicated for hip and lower back pain.
In order to make sense of large data sets,
Ms Hendry turned to machine learning. CIC specialist Dr Kevin Chai led the team
that built a convolutional neural network. The network was trained by using Ms
Hendry’s library of manually-classified movement data.
Training let the network identify patterns
and diagnostic features in the mass of sensor data that had been correlated with
different jumps and leg lifts through the video.
Using data gathered from all six sensors,
the network could identify target movements with 80% accuracy, which was enough
to assess training load.
During the process, the team learned that
with data coming from only one sensor, the one placed on the sacrum, the neural
network still had over 75% accuracy. Having only one sensor, which can be
hidden under a costume, opens up avenues to studying performance, not just
training.
52 dancers are now being recorded over an
entire day of training, four times across a semester. The trained neural
network is then used to convert the data into a quantitative measure of jumping
and leg-lifting training volume for each.
During each data collection day, the
dancers complete a survey that assesses a range of emotional, cognitive and
lifestyle factors, pain experienced, and limitations faced during training.
The data will then be used to look at the
trajectory of each dancer across the semester and explore the various factors
that correlate with pain and disability.
WAAPA Biomechanist Dr Luke Hopper shared that
they want to assist the pre-professional dancers to reach the challenging
heights of being professional dancers. A tool like this can be calibrated to
focus on training, and measure outputs comparable to live performances.
Curtin Biomechanist and Ms Hendry’s
supervisor, Dr Amity Campbell explained that field-based analysis is the new
way of doing biomechanical research. The dancers can be captured in their
normal environment
Bringing them, she added, in laboratories with
cameras and sensors will not capture normal performance pressures. Capturing
their activity in real conditions will be so much more useful for injury
prevention, performance development, and high-performance training.
