Self-Referenced Gesture Recognition Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current gesture recognition systems face challenges in accurately and efficiently differentiating between various forms of human body movements, particularly in real-time applications such as entertainment and security, due to the subjective nature of gesture definition and the complexity of human gestures.
Innovation Solution
A machine learning-based system that uses self-referential gesture data, where the positioning and movement of body parts are referenced to a specific point on the body, such as the waist, to classify and recognize gestures through a process that involves data collection from multiple sources, including gaming platforms and public venues, and utilizes artificial intelligence to predict future gestures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional shape descriptors and various processes are used for gesture recognition, then gesture classification can be achieved, but the system complexity and difficulty in accurately differentiating between various forms of human body movements increase
Solution Approach 1:
The patent replaces traditional mechanical shape descriptor-based gesture recognition systems with an artificial intelligence/machine learning system. The AI system automatically learns and extracts gesture features from data, eliminating the need for manual shape descriptor design and complex traditional processing pipelines, thereby maintaining high recognition accuracy while reducing system complexity.
2Measurement precision
If more detailed gesture data is collected to improve recognition accuracy, then gesture classification precision improves, but data processing time and computational resources increase
Solution Approach 1:
The patent implements preliminary action by pre-collecting and storing gesture data from multiple sources including gaming platforms and public venues before actual recognition is needed. This pre-processing and data accumulation phase allows the AI system to have training data ready in advance, reducing real-time processing requirements and enabling faster gesture recognition when actually deployed.
3Productivity
If traditional gesture recognition methods are used, then implementation is feasible with current technology, but the speed and efficiency of gesture classification are insufficient for real-time applications
Solution Approach 1:
The patent applies parameter changes by transitioning from traditional gesture recognition parameters to AI-based recognition parameters. The machine learning model processes gesture data with optimized parameters that enable both high-speed classification and high accuracy, making real-time applications reliable. The system achieves this by learning optimal feature extraction and classification parameters during training phases.
Data Source
AI summary
The disclosure describes systems and methods of detecting body movements using gesture data. The gesture data may be self-referenced and may be comprised by frames which may identify locations or positions of body parts of a subject with respect to a particular reference point within the frame. A classifier may process frames to learn body movements and store the frames of gesture data in a database. Data comprising frames of self-referenced gesture data may be received by a recognizer which recognizes movements of the subject identified by the frames by matching gesture data of the incoming frames to the classified self-referenced gesture data stored in the database.