Dynamic Gesture Recognition via Static Triggered Model Switching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dynamic gesture recognition technologies face challenges in accurately determining the start and end times of gestures due to complex transition actions and high computational burdens, leading to lower recognition accuracy and increased resource usage.
Innovation Solution
A method combining static and dynamic gesture recognition models, where a static gesture detection model is used to detect initial gestures and trigger the dynamic gesture detection model, allowing for efficient activation and deactivation to avoid continuous processing and reduce errors from mixed actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the dynamic gesture detection model is continuously executed to prevent missing gestures, then gesture detection completeness is improved, but computational burden and resource usage increase
Solution Approach 1:
The system performs gesture detection at specific intervals triggered by static gesture detection rather than continuously processing every frame. The dynamic gesture detection model is activated periodically when a static gesture is detected, reducing computational burden while maintaining detection completeness through strategic sampling of frames.
Solution Approach 2:
The static gesture detection model performs preliminary detection on individual frames to identify potential gesture start points before activating the more computationally intensive dynamic gesture detection model. This preliminary filtering action reduces the number of times the dynamic model needs to execute, lowering overall computational burden.
2Reliability
If the dynamic gesture detection model is continuously executed, then gesture detection completeness is improved, but recognition accuracy deteriorates due to mixed actions in frames
Solution Approach 1:
Instead of continuous execution that processes frames with mixed actions, the system uses periodic execution triggered by static gesture detection. This ensures the dynamic model only processes frames where a gesture is actually present, eliminating false detections from mixed actions while maintaining completeness through targeted frame selection.
Solution Approach 2:
The static gesture detection model serves as an intermediary that filters and selects appropriate frames for dynamic gesture detection. It mediates between continuous video input and the dynamic detection model, preventing mixed actions from being processed by reducing the input frames to only those containing actual gestures.
3Reliability
If both static and dynamic gesture recognition models are executed simultaneously, then comprehensive gesture detection is achieved, but device complexity and processing overhead increase
Solution Approach 1:
The gesture detection system is segmented into two distinct phases: static gesture detection on individual frames, followed by dynamic gesture detection on selected frame sequences. This segmentation allows each model to operate independently at appropriate times, achieving comprehensive detection while reducing processing overhead compared to simultaneous execution.
Solution Approach 2:
The system dynamically switches between static and dynamic detection modes based on gesture characteristics. Static detection handles individual frame analysis, while dynamic detection activates only when needed for continuous gesture sequences. This dynamic approach optimizes resource usage by applying the appropriate detection method rather than running both simultaneously.
Data Source
AI summary
A dynamic gesture recognition method includes: performing detection on each frame of image of a video stream using a preset static gesture detection model to obtain a static gesture in each frame of image of the video stream; in response to detection of a change of the static gesture from a preset first gesture to a second gesture, suspending the static gesture detection model and activating a preset dynamic gesture detection model; and performing detection on multiple frames of images that are pre-stored in a storage medium using the dynamic gesture detection model to obtain a dynamic gesture recognition result.