Vision-Based Robot Pose Control for User Movement Imitation

The system segments the interaction into distinct modules: image capture module, neural network processing module, and actuator control module. This segmentation simplifies each individual component while maintaining overall system reliability, allowing the complex task of user-robot interaction to be divided into manageable, independent functions that can be optimized separately.

The patent replaces complex mechanical control systems with a vision-based neural network system. Instead of using sophisticated mechanical sensors and actuators to detect and replicate human movement, the system uses image capture devices and neural networks to detect user pose and generate robot movement commands, significantly reducing mechanical complexity while improving reliability through software-based control.

The neural network model serves multiple functions: it detects various user poses, adapts to different users automatically, and works across different interaction scenarios. This universal model eliminates the need for separate configuration systems for different users or contexts, achieving high adaptability without increasing system complexity. The same core neural network handles diverse interaction requirements through its inherent generalization capabilities.

The system performs self-adaptation through the neural network's automatic learning and generalization capabilities. When a new user interacts with the system, the neural network automatically adjusts to recognize and track that user's pose without requiring manual calibration or configuration. This self-service adaptability reduces the complexity of system setup and makes the system universally applicable to different users.

The system replaces expensive, precision mechanical components with more affordable electronic and software-based solutions. Instead of using costly mechanical encoders, force sensors, and precision actuators, the system uses standard image capture devices and neural network processing to achieve movement accuracy, significantly reducing manufacturing costs while maintaining acceptable precision for interactive applications.

The patent substitutes mechanical precision measurement systems with vision-based detection. Rather than relying on expensive mechanical sensors to measure user movement with high precision, the system uses image processing and neural networks to detect pose and calculate movement, achieving comparable accuracy at a fraction of the cost through non-contact optical measurement.