Wearable Activity Recognition with Earbud Orientation Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fusing data from wrist-worn and head-worn sensors for activity recognition is challenged by earbud misalignment and the scarcity of misalignment training data, leading to performance issues and inefficient processes.
Innovation Solution
A method involving orientation transformation matrices to correct earbud misalignment using Euler angles, combined with modality-aware dynamic fusion and conformal prediction to enhance activity detection accuracy, utilizing a pair of earbuds and a smartwatch.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data from multiple wearable devices is fused for activity recognition, then detection accuracy is improved, but device complexity and data processing requirements increase
Solution Approach 1:
The system segments the activity recognition task into multiple stages: (1) individual device activity detection, (2) activity transition detection, and (3) coordinated activity recognition. This segmentation allows complex multi-device fusion to be broken down into manageable components, reducing overall system complexity while maintaining high accuracy.
Solution Approach 2:
The system performs preliminary activity detection and classification on each individual device before integrating results. By pre-processing and segmenting the detection task across multiple devices independently, the system reduces the computational burden on the final fusion stage and simplifies the overall architecture.
2Measurement precision
If earbud orientation variations are accounted for in activity recognition, then measurement accuracy is improved, but training data requirements increase
Solution Approach 1:
The system automatically estimates and corrects earbud orientation using sensor data from the wearables themselves, without requiring external calibration equipment or extensive manual alignment data. The orientation estimation leverages the devices' own sensors to infer spatial relationships, reducing the need for large annotated training datasets.
Solution Approach 2:
The system replaces physical calibration procedures with computational orientation estimation algorithms. Instead of requiring mechanical alignment or physical calibration data, the system uses sensor fusion and mathematical models to infer orientation, significantly reducing training data requirements while improving measurement precision.
3Measurement precision
If multi-stage fusion processing is implemented, then activity recognition accuracy is improved, but processing time increases
Solution Approach 1:
The fusion processing is segmented into distinct stages that operate in parallel where possible: individual device processing, transition detection, and coordinated activity recognition. This segmentation enables efficient processing by allowing independent computation on each stage, reducing overall processing time while maintaining accuracy.
Solution Approach 2:
The system maintains continuous activity detection and transition monitoring across all stages, ensuring that processing is ongoing rather than batched. This continuous processing approach reduces total processing time by eliminating idle periods between stages while preserving the accuracy benefits of multi-stage fusion.
Data Source
AI summary
In one embodiment, a method includes accessing a current sensor signal from an inertial measurement unit (IMU) of an earbud in a current earbud orientation worn by a user and accessing a baseline-orientation IMU sensor signal for the earbud in a baseline earbud orientation. The method further includes estimating, for the earbud and based on the current sensor signal and the baseline-orientation IMU signal, an orientation transformation matrix that transforms the current sensor signal from the current earbud orientation to the baseline earbud orientation; and transforming the current sensor signal from the current earbud orientation to the baseline earbud orientation using the orientation transformation matrix.