Wearable AR Video Generation for Real-Time Posture Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods fail to provide real-time feedback on the ideal posture of a presenter to an acquirer, limiting the acquirer's ability to correct their posture efficiently, as they cannot easily view the presenter's posture from different angles or maintain a fixed position while performing actions like dance, leading to increased errors and complexity in setup.
Innovation Solution
A video generation device and method using wearable devices with cameras and transparent displays, allowing the acquirer and presenter to capture images of each other, estimate corresponding points, and generate a semi-transparent video that matches the presenter's posture, enabling the acquirer to view the ideal posture from any angle without fixing their position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple cameras are set to capture images from different viewpoints, then the acquirer can view the presenter's posture from any angle, but the device configuration becomes complicated and the cost increases
Solution Approach 1:
The system divides the viewpoint provision into two separate components: the wearable device on the presenter captures images from a fixed position, while the AR display on the acquirer's head provides dynamic viewpoint switching through software processing. This segmentation avoids the need for multiple physical cameras while achieving viewpoint flexibility.
Solution Approach 2:
The system creates a virtual copy of the presenter's posture by processing the captured images through AR technology. The acquirer can view this virtual representation from any angle without needing multiple physical cameras, as the AR system generates different viewpoints computationally from the single captured image.
2Ease of operation
If the display is moved to follow the acquirer's position, then the acquirer can view their posture from any position, but the device configuration becomes complicated and the cost increases
Solution Approach 1:
The system makes the display dynamic by attaching it to the acquirer's head as a wearable AR device. This allows the display to automatically follow the acquirer's movements and maintain the correct viewing position without requiring external tracking or adjustment mechanisms, simplifying the overall configuration.
Solution Approach 2:
The wearable AR display on the acquirer's head automatically adjusts its position and orientation based on the acquirer's head movements. The system serves itself by using the acquirer's natural head movements to control the display position, eliminating the need for external control mechanisms.
3Device complexity
If a fixed-point camera is used to capture the acquirer's posture, then the configuration is simple, but the acquirer's position is fixed and big movements fall outside the angle of view
Solution Approach 1:
The system adds a new dimension to the feedback by providing a virtual overlay that can extend beyond the physical camera's field of view. The AR display presents the presenter's posture guidance in a separate visual layer that is not constrained by the camera's angle, allowing guidance for movements outside the original view.
Data Source
AI summary
To provide a video generation device or the like which generates a video allowing an acquirer to obtain an ideal posture of a presenter from any viewpoint in real-time without fixing a self-position, while at the same time, allowing confirmation of a self-posture. The presenter wearing a wearable device including a camera CP and the acquirer wearing a wearable device including a camera CL and a transparent display face each other and capture images of each other with each other's cameras, and the video generation device comprises: a corresponding point obtaining part which estimates a frame of the acquirer from a video obtained with the camera CP, obtains a set FL of one or more corresponding points representing the frame of the acquirer which is estimated, estimates a frame of the presenter from a video obtained with the camera CL, and obtains a set FP of one or more corresponding points representing the frame of the presenter which is estimated; and a video generation part which generates, based on the set FL and the set FP, a semi-transparent video representing a posture of the acquirer such that the semi-transparent video matches a posture of the presenter which is transparently seen through the display.