Orientation-Independent Movement Recognition via Difference Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional movement recognition systems require precise placement and orientation of inertial sensors, which can be disrupted by user movement, necessitating recalibration and limiting their functionality in applications like healthcare rehabilitation and consumer electronics.
Innovation Solution
A context-aware movement recognition system utilizing an orientation-independent feature extraction circuit and classification circuit, which receives measurements from 3D accelerometers and gyroscopes to classify movements irrespective of sensor orientation, allowing operation with arbitrary sensor placement on body joints or segments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If precise placement and orientation of inertial sensors is required, then measurement precision is improved, but device complexity and ease of operation deteriorate due to recalibration requirements
Solution Approach 1:
The patent transforms the measurement parameters from absolute orientation-dependent values to relative difference-based features. By computing differences between sensor measurements taken at different times or positions, the system eliminates the need for precise initial sensor orientation while maintaining measurement accuracy for movement recognition.
Solution Approach 2:
The patent introduces an intermediary processing layer that computes differences between sensor measurements. This intermediary transformation converts orientation-dependent raw data into orientation-independent difference features, allowing the system to function without precise sensor placement while maintaining measurement precision.
2Measurement precision
If sensor placement orientation is fixed, then measurement precision is improved, but adaptability deteriorates
Solution Approach 1:
The patent creates a universal movement recognition system that works with sensors placed in any orientation on different body segments. The difference-based feature extraction method is universally applicable regardless of sensor location or initial orientation, enabling the same system to function across multiple applications and sensor configurations.
Solution Approach 2:
By changing from absolute orientation parameters to relative difference parameters, the system achieves adaptability to various sensor placements while maintaining measurement precision. The parameter transformation makes the system versatile across different body segments and sensor orientations.
3Adaptability or versatility
If sensor displacement occurs, then adaptability is improved, but measurement precision deteriorates without recalibration
Solution Approach 1:
The patent employs a feedback mechanism where the system continuously computes differences between current and reference sensor measurements. This feedback-based difference calculation automatically compensates for sensor displacement without requiring recalibration, maintaining measurement precision while adapting to changed sensor positions.
Solution Approach 2:
The system transitions from a static calibration-based approach to a dynamic difference-based approach. By continuously calculating measurement differences rather than relying on fixed calibration data, the system adapts to sensor displacement in real-time while maintaining precision.
Data Source
AI summary
A movement recognition system includes a feature extraction circuit and a classification circuit that is orientation independent. The feature extraction circuit is configured to receive a plurality of acceleration or angular velocity measurements from a sensor or multiple sensors that include a three dimensional (3D) accelerometer and 3D gyroscope. Each of the measurements is taken at different times. The feature extraction circuit is also configured to determine a difference between a first of the measurements and each other of the measurements. The classification circuit is configured to classify a movement of an object attached to the sensor irrespective of an orientation of the sensor on the object utilizing a signal recognition technique based on the difference between the first of the measurements and each other of the measurements.