RF Sensor Activity Recognition on Work Surfaces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solutions for recognizing user activities above a desk or counter face privacy concerns, sensitivity to illumination changes, occlusion issues, and inability to detect objects relevant to activities, particularly in environments where camera-based systems are intrusive and wearable sensors lack object detection capabilities.
Innovation Solution
A system utilizing RF sensors operating in the 3.3-10.3 GHz spectrum, placed unobtrusively behind light construction materials, combined with wearable sensors, to recognize activities through machine learning algorithms and user-driven data labeling, enabling accurate detection of user actions and object presence without line-of-sight requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If camera-based solutions are used to recognize user activities, then activity recognition capability is improved, but privacy concerns increase and the system becomes sensitive to illumination changes and occlusion
Solution Approach 1:
The patent replaces camera-based optical sensing with RF sensor arrays that emit and detect radio frequency signals. This substitution eliminates dependency on visible light conditions and removes the privacy concerns associated with visual recording, while maintaining the ability to detect user activities through RF signal interactions with objects and users on the work surface.
Solution Approach 2:
The system changes the detection parameter from optical reflection (camera) to RF signal modulation (RF sensors). By operating in the radio frequency domain rather than the visible spectrum, the system achieves illumination independence and reduced privacy concerns while preserving activity recognition capabilities through detection of RF signal variations caused by user interactions.
2Measurement precision
If wearable sensors are used to track user activities, then activity tracking is improved, but the system cannot detect objects present and relevant to the activity
Solution Approach 1:
The patent introduces RF sensors as an intermediary detection mechanism that bridges the gap between wearable sensor data and object detection. The RF sensors detect objects on the work surface by measuring their impact on RF signal propagation, providing object presence information that complements the motion data from wearable sensors, thereby recovering the lost object detection capability.
Solution Approach 2:
The system merges data from wearable sensors and RF sensors into a unified activity recognition framework. The wearable sensors provide user motion information while the RF sensors provide object presence and location information, and their combination enables comprehensive activity recognition that includes both user actions and relevant objects, resolving the limitation of wearable-only systems.
3Measurement precision
If IMU-based solutions are used, then user motion tracking is improved, but the system cannot determine where an activity is occurring on the work surface
Solution Approach 1:
The patent adds a spatial dimension to motion tracking by deploying multiple RF sensors across the work surface area. While wearable IMUs track user motion, the distributed RF sensors provide spatial localization information by detecting which specific sensors are affected by user interactions, thereby determining the location of activities on the work surface in addition to the motion data.
Data Source
AI summary
Systems and methods of the present disclosure can involve several work spaces wherein each of the work spaces are associated with a set of activities and wherein each of the work spaces are coupled to one or more radio frequency (RF) sensors. Through RF sensor data detected from interactions with the work surface, example implementations described herein can determine which activity from the set of activities is being conducted through the application of a recognition algorithm that is generated from a machine learning algorithm.