US20260102658
2026-04-16
Human necessities
A63B24/0075
The patent application outlines a system that provides real-time coaching for fitness exercises and yoga using artificial intelligence. The system utilizes a client application on a device to capture images, video, and audio, which are then analyzed by an AI agent. This agent estimates poses, evaluates movement sequences, and analyzes breathing patterns to classify deviations from reference models. The system offers a dynamic, multi-segment response that includes brief commands for immediate corrections, detailed instructions for learning, and additional information.
The client application operates on devices such as mobile phones or tablets, equipped with cameras and microphones. It interacts with an AI agent that processes captured data to provide personalized feedback. Optional wearable sensors and multiple cameras can be integrated to enhance the accuracy of the analysis. The system employs a fusion module for data integration and features both on-device and cloud-based processing to optimize performance while ensuring user privacy through encryption and anonymization.
The system's Dynamic Response Selector (DRS) chooses appropriate response segments based on the application's state and user actions. These segments are presented via device displays or augmented reality headsets, offering overlays that guide the user. The system is designed to provide immediate safety-critical corrections during practice, enhance learning through post-session instructions, and manage data efficiently to reduce latency and bandwidth usage.
Privacy and security are key considerations in the system's design. Data processing is divided between the device and the cloud, with sensitive information being anonymized and encrypted. Caching mechanisms are employed to further protect user data. The system ensures that all operations comply with privacy standards, safeguarding user information while delivering effective coaching.
The system is adaptable, allowing for variations in implementation. It can accommodate different modalities such as images, video, and audio, and supports various model families for analysis. Users can customize their experience by creating or selecting reference items, which can be stored and used for personalized guidance. The system's architecture supports distributed processing, enabling it to function efficiently across different environments and devices.