Staying in shape is important for physical and mental health. Adding squats
to a regular workout routine is beneficial, but performing them correctly and
determining the right number of repetitions can be challenging. A feedback
system utilizing body movement identification methods can assist in performing squats correctly and efficiently. The system we have developed
incorporates velocity-based training principles and posture estimation using Mediapipe’s pose landmarks. It provides feedback on squat depth, neck
angle, and improvement techniques.
Our system functions as a visual feedback application. It was tested for
velocity-based training, posture estimation, and feedback and interface functionality. A velocity loss of 10-20% indicates the need for a break during
squats, where comparing current squat velocity to the initial squat in a set
notifies the user of the velocity loss. The maximum velocity of the upward
motion of the squat was calculated with a standard deviation which translates to a potential error of 0.008% in velocity comparison. The depth of
the squat is determined by the femur’s parallelism to the ground, with a
deviation corresponding to a 10% error. A neck angle within a determined
range indicates good posture.
The application improves squatting performance, advises on breaks, and
offers user-friendly features like gesture control, however, the gesture control
does not work as intended and requires further work. Future work can
expand posture feedback to include additional requirements like stance, knee
position, back angle, and bar placement. If adapted as a phone application,
it can serve as a versatile and user-friendly exercise aid, providing valuable
information on correct exercise performance.