AI Posture Identification for Home Exercise Guidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Home training exercises often lack effective guidance for ensuring correct postures, which can impact exercise effectiveness and safety.
Innovation Solution
A display apparatus equipped with a camera, display, and artificial intelligence model that identifies user postures by analyzing location data from images, providing real-time feedback on posture matching or not matching a training image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If home training is performed without a trainer, then exercise accessibility and convenience are improved, but exercise effectiveness and safety deteriorate due to lack of posture guidance
Solution Approach 1:
The system captures images of the user's posture through a camera, compares it with the training image using an AI model, and provides real-time feedback by displaying whether the posture matches or does not match the correct form. This closed-loop feedback mechanism enables self-guided exercise with automated correction, resolving the contradiction between exercise accessibility and effectiveness.
2Device complexity
If posture identification is performed using basic location data only, then system complexity is reduced, but measurement precision of user posture deteriorates
Solution Approach 1:
The AI model transforms 2D location data from images into 3D posture information by inferring depth and spatial relationships. This dimensional transformation enables accurate 3D posture reconstruction and matching without requiring complex 3D sensors, resolving the contradiction between system simplicity and measurement precision.
3Quantity of substance
If the AI model is trained only with original location data, then training data quantity is limited, but model adaptability to various shooting angles deteriorates
Solution Approach 1:
The training process performs preliminary transformations on the location data by generating rotated versions of the original data before model training. This preprocessing step creates a diverse set of training samples that simulate various shooting angles and positions, enabling the model to adapt to different exercise environments without requiring extensive real-world data collection.
Data Source
AI summary
A display apparatus is disclosed. The display apparatus includes: a camera, a display, a memory storing an artificial intelligence model trained to identify a posture of a user based on location data with respect to a plurality of body parts of a user included in images and additional location data acquired based on the location data, and a processor configured to: control the display to display a training image and images photographed by the camera, identify the posture of the user included in the photographed images by inputting the location data with respect to the plurality of body parts of the user included in the photographed images, and control the display to display a training guide based on whether the posture of the user matches a posture corresponding to the training image.