Abstract:
In the digital era, the confluence of technology and fitness has taken home
workouts to the next level, making it more accessible but often less supervised. This,
however, portends the worry of the risk of injury through wrong exercise postures in the
absence of professional guidance. Our project, entitled "Enabling Workout Precision:
AI and Trigonometric Ratio-Based Pose Correction," solves this problem by introducing
a first-of-its-kind system that integrates Human Pose Estimation with both
Trigonometric Ratios and deep learning for real-time and precise posture correction
while exercising.
Two approaches have been implemented in our project for pose estimation
correction. The first method uses trigonometric ratios to calculate angles and alignments
directly, offering a mathematical and deterministic basis for evaluating posture. The
second method employs a deep learning model, specifically a convolutional neural
network (CNN), which classifies key exercises like planks, squats, and shoulder presses.
This model is trained to identify correct and incorrect workout poses and predicts
specific deviations defined as labels in the dataset for each of the exercises.
The real-time feedback mechanism helps with the maintenance of the right form,
and it further significantly minimizes the risk of injury; hence, it assures a secure and
effective home workout experience. This system provides a practical solution to the
challenge of supervision in home fitness routines, leveraging great practical application
of trigonometry and AI to personal fitness.
