Motion Sensor Activity Detection with Dynamic Reference Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion sensor activity detection methods require prior calibration and are sensitive to the orientation and positioning of the sensor, making them less robust to changes in the observed system or arrangement, and assume invariant values across individuals and experiments, which is a restrictive approximation.
Innovation Solution
A method that calculates a likelihood value of the observation sequence using a statistical model and compares it to a threshold, allowing for the estimation of a new reference value for the sensor's axis, enabling calibration at any time and making the system more robust to changes in sensor placement and orientation, with optional steps for automatic calibration and invalidation of activity determinations based on the likelihood value.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If prior calibration is performed to establish reference values for sensor orientation, then measurement precision is improved, but the system becomes sensitive to changes in sensor placement and orientation, reducing reliability
Solution Approach 1:
The patent transforms the static calibration approach into a dynamic one by continuously estimating reference values during operation. Instead of using fixed calibration values obtained during initial setup, the system adapts reference values in real-time based on current sensor readings and statistical models, allowing the system to accommodate changes in sensor placement and orientation while maintaining measurement precision
Solution Approach 2:
The system performs self-calibration by automatically estimating reference values for sensor orientation without requiring external calibration tools or manual intervention. The statistical model uses the sensor's own measurement data to infer and update reference values, enabling the system to self-correct for placement variations and maintain reliability
2Reliability
If calibration is performed during operation to improve robustness, then reliability is improved, but device complexity increases due to additional computational requirements
Solution Approach 1:
The patent replaces complex mechanical calibration procedures with statistical computation. Instead of using physical calibration tools or manual adjustment mechanisms, the system uses statistical models and probability theory to estimate reference values computationally, reducing mechanical complexity while improving reliability through continuous adaptation
Solution Approach 2:
The system changes the approach from fixed parameters (calibration values) to dynamic parameters (estimated reference values). By using statistical distributions and probability models to represent sensor characteristics, the system can adapt parameters in real-time without increasing hardware complexity, managing computational load through efficient statistical methods
Data Source
Figure 1~2
Figure 3~4
Figure 5~7
AI summary
This method for detecting the activity (A) of a physical system having a motion sensor comprises the following steps: extracting (100) a sequence (M) of measurements provided by the motion sensor; deducing (102) therefrom an observation sequence (O) calculated using the sequence of measurements (M); determining (104) the activity (A) of the physical system in the form of a sequence of states corresponding to the observation sequence, using a statistical model of components of the observation sequence in view of a plurality of possible predetermined states in which the physical system may be located. At least one component of the observation sequence (O) is calculated from a reference value ( ? ) of a vector representative of a reference axis. Moreover, the method comprises the following calibration steps: calculating (106) a likelihood value of the observation sequence (O) in view of said statistical model; comparing (106) this likelihood value to a predetermined threshold; depending on this comparison, triggering an estimation (108) of a new reference value ( ? ) of the vector representative of the reference axis.