3D Motion Recognition Using Depth Sensor and Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 3D user motion recognition methods require expensive high-speed cameras or motion sensors, along with complex synchronization and calculation processes, making them costly and inefficient.
Innovation Solution
A system and method utilizing low-cost depth sensors and deep learning models to convert 3D low-quality depth data into high-quality image data, allowing for precise automatic recognition of user motion without the need for expensive equipment or complex calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple high-speed cameras are used for 3D user motion recognition, then measurement precision is improved, but device cost and system complexity increase
Solution Approach 1:
The patent replaces the mechanical/optical system of multiple high-speed cameras with a single depth sensor combined with deep learning algorithms. The depth sensor captures 3D spatial information, and the neural network processes this data to achieve accurate motion recognition, substituting complex hardware synchronization with software-based processing.
Solution Approach 2:
The patent introduces depth data as an intermediary representation between the physical motion and the recognition system. The depth sensor captures spatial information that serves as a bridge, allowing the neural network to process 3D motion data without requiring multiple camera views and complex geometric computations.
2Measurement precision
If multiple motion sensors are attached to each joint area, then measurement precision is improved, but ease of operation and device cost worsen
Solution Approach 1:
The patent replaces the mechanical attachment of multiple motion sensors to body joints with a non-contact depth sensing system. The depth sensor captures 3D spatial information optically, and the neural network extracts joint positions and motions from this data, eliminating the need for physical sensor attachment to each joint.
Solution Approach 2:
The patent creates a digital copy of the user's 3D body structure and motion through depth sensing. Instead of physically attaching sensors to joints, the system captures a volumetric representation of the user and uses neural networks to extract joint information from this digital model, achieving the same measurement goal without physical intervention.
3Measurement precision
If conventional 3D motion recognition methods are used, then measurement precision is improved, but loss of time increases due to complex calculation processing
Solution Approach 1:
The patent pre-trains the neural network on large datasets of depth images and corresponding 3D motion labels. This preliminary learning phase enables the model to quickly infer joint positions and motions from new depth data without requiring complex real-time calculations, significantly reducing processing time during actual use.
Solution Approach 2:
The patent replaces computationally intensive geometric calculations and coordinate transformations with a trained neural network that performs inference. The deep learning model has learned the complex mappings from depth images to 3D motion data, substituting manual calculation processes with automated pattern recognition that is both accurate and efficient.
Data Source
AI summary
Provided is a system and method for automatically recognizing user motion. The system for automatically recognizing user motion includes an input unit configured to receive three-dimensional (3D) measurement data, a memory which stores a program for performing automatic recognition on 3D user motion using 3D low-quality depth data and a deep learning model, and a processor configured to execute the program, wherein the processor converts the 3D low-quality depth data into 3D high-quality image data.