Objective monitoring of swimming technique is crucial for effective improvement. Existing studies primarily address the detection of stroke phases using multiple inertial measurement units. These studies use underwater video of the swimmer as a reference.
In our study, we implemented automatic detection of stroke phases, but unlike other studies, we only used a single inertial measurement unit attached to the palm. For a reference system, we had a system of cameras and markers on dry land. Also, unlike other studies, the swimmers were able to breathe while swimming. In addition to the phases, we also detected some of the most common errors in swimmers (unnecessary wrist rotation, arm deviation from the direction of swimming, flicking, and shortening the stroke). We defined and compared metrics for these errors after providing feedback and found improvement in all swimmers.
Orientation errors obtained with the IMU relative to the optical reference system are in the range of ±3.77 °. The average relative temporal errors for determining individual stroke phases as a percentage of the total stroke are: entry 3.3%, pull 2.8%, push 2.3%, and recovery 2.0%.
Eight swimmers participated in the study. Three of them were professional swimmers, while the rest were recreational swimmers with at least 7 years of experience. We conducted two series of measurements. Between the two, we gave the swimmers feedback on the errors we had detected. All swimmers were able to reduce the errors we pointed out to them. Two swimmers reduced flicking at the back of the stroke, with one of them eliminating it completely. One swimmer corrected their shortened stroke. Four swimmers reduced the rotation of their wrist and palm during the pull phase, and one swimmer successfully reduced palm rotation during the recovery phase.
|
