Hand Sliding Direction Detection Using Dynamic Vision Sensor Events
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing short-range gesture identification schemes face issues such as high computational burden, power consumption, and accuracy problems due to motion blur and sparse data sources, limiting their ability to accurately identify subtle hand gestures.
Innovation Solution
A method utilizing a Dynamic Vision Sensor (DVS) to generate time plane images from event data, process these images to identify gestures, and determine the sliding direction of a hand by analyzing active pixels, variance, and deviation directions, reducing computational burden and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image sensor is used for gesture identification, then key points can be detected on images, but computational burden and power consumption increase significantly
Solution Approach 1:
The patent replaces traditional image sensors with a Dynamic Vision Sensor (DVS) that outputs event data instead of continuous images. This substitution changes the fundamental data representation from dense image frames to sparse event streams, reducing computational burden and power consumption while maintaining gesture identification accuracy.
Solution Approach 2:
The patent extracts only the essential motion information from the visual data by using event-driven detection. Instead of processing entire images, the system extracts and processes only the event data that represents actual changes in the scene, significantly reducing computational requirements.
2Measurement precision
If traditional image sensor is used for gesture identification, then key points can be detected, but system response becomes slow
Solution Approach 1:
The patent transitions from periodic frame-based processing to event-triggered processing. Instead of analyzing every image frame at fixed intervals, the system processes data only when changes occur (when events are triggered), enabling faster response to actual gesture movements while reducing unnecessary computational cycles.
Solution Approach 2:
The replacement of traditional image sensors with DVS enables asynchronous event-driven processing, eliminating the frame rate bottleneck and allowing the system to respond immediately to gesture movements regardless of frame timing.
3Speed
If hand moves too fast with traditional image sensor, then motion blur occurs and key points fail to be detected
Solution Approach 1:
The patent replaces traditional image sensors with DVS, which does not suffer from motion blur. The event-driven architecture captures instantaneous changes in light intensity at each pixel, allowing accurate detection of fast-moving hands without the temporal averaging that causes blur in frame-based systems.
Solution Approach 2:
The patent implements continuous event-driven monitoring where the sensor continuously tracks changes in the visual scene. This continuous detection approach ensures that fast movements are captured as a sequence of events rather than being averaged out in discrete frames, maintaining detection accuracy at high speeds.
4Adaptability or versatility
If millimeter-wave radar is used for gesture identification, then simple movements can be identified, but data sources are sparse and subtle gestures cannot be detected
Solution Approach 1:
The patent replaces millimeter-wave radar with a Dynamic Vision Sensor that provides rich spatial and temporal information. The DVS captures detailed visual events including position, intensity change, and timing, enabling detection of subtle gestures that radar cannot detect due to its sparse data nature.
Solution Approach 2:
The patent changes the fundamental measurement parameters from radar's distance and velocity measurements to DVS's pixel-level intensity change events. This parameter change provides much richer spatial resolution and temporal detail, enabling detection of subtle hand gestures while maintaining the ability to identify broader movement patterns.
Data Source
AI summary
The present disclosure provides a method and a system for identifying a sliding direction of a hand, a computing device and an intelligent device. The method includes: generating at least one time plane image in accordance with a series of event data from a dynamic vision sensor, each event being triggered in accordance with movement of an object relative to the dynamic vision sensor in a field of view; identifying a gesture in the time plane image; in the case that the identified gesture is a predetermined gesture, entering a hand sliding identification state; determining active pixels indicating the sliding of the hand in a corresponding time plane image in the hand sliding identification state; and determining the sliding direction of the hand in accordance with the active pixels.