Data Conversion Device Synchronizing Motion Frames via Angular Embedding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in synchronizing motion data from moving images across different backgrounds, leading to reduced accuracy in motion recognition and extraction, as they are affected by background variations and are not applicable to non-moving image data.
Innovation Solution
A data conversion device that normalizes posture data from moving image frames into an angular representation using a graph convolutional network, calculates feature amounts, and synchronizes target motion data with reference data by aligning optimal paths based on calculated distances, effectively isolating background influences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the nearest frames in the learned embedded space are associated to synchronize common motions, then the learning of the generative adversarial model can be stabilized, but the accuracy of synchronization decreases when the background is greatly different in different moving images
Solution Approach 1:
The patent segments the feature extraction process by using separate encoders for reference moving image data and synchronization target moving image data. Each encoder independently extracts features while being trained to map to a common embedded space, allowing the system to handle background differences while maintaining synchronization accuracy.
Solution Approach 2:
The patent introduces an embedded space as an intermediary representation between the reference and target moving images. By mapping features from both sources into this common embedded space through trained encoders, the system can compare and synchronize motions while being invariant to background differences.
2Reliability
If the nearest frames in the learned embedded space are associated to synchronize common motions, then the learning of the generative adversarial model can be stabilized, but the method is not applicable to data that is not in the form of a moving image
Solution Approach 1:
The patent creates a universal feature extraction framework using encoders that can process different types of input data (moving images, depth images, infrared images, or their combinations) and map them to a common embedded space. This allows the same synchronization approach to be applied across multiple data types while maintaining learning stability.
3Adaptability or versatility
If motion data is extracted using moving image data, then the same motion can be extracted from moving image data having different backgrounds, but the accuracy decreases when backgrounds are greatly different
Solution Approach 1:
The patent extracts motion information by taking out only the relevant features through trained encoders that map to an embedded space. This extraction process separates the motion features from background information, allowing accurate motion extraction even when backgrounds differ significantly between reference and target images.
Data Source
AI summary
A data conversion device including a feature amount calculation unit that normalizes posture data estimated in each frame constituting moving image data including a synchronization target motion into an angular representation, and calculates a feature amount in an embedded space by inputting the posture data normalized into the angular representation to an encoder, a distance calculation unit that calculates a distance between a feature amount calculated in each frame constituting reference moving image data and a feature amount calculated in each frame constituting synchronization target moving image data, a synchronization processing unit that calculates an optimal path for each frame based on the calculated distance and synchronizes the synchronization target moving image data with the reference moving image data by aligning timings of frames connected by the optimal path, and an output unit that outputs the synchronization target moving image data synchronized with the reference moving image data.