Fall Detection Using Context-Aware Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fall detection systems often produce false positives, failing to distinguish between actual falls and other events like device drops or normal activities, due to reliance on accelerometers and other sensors that are computationally expensive and location-bound.
Innovation Solution
A system utilizing multiple sensors, including motion sensors and barometers, combined with machine learning algorithms that analyze user context and activity patterns to accurately identify falls, differentiate between falls and false positives, and determine the severity of falls.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If accelerometers are used to detect falls based on sudden acceleration changes, then fall detection capability is provided, but false positive identifications increase
Solution Approach 1:
The patent combines data from multiple sensors (accelerometer, barometer, gyroscope, magnetometer) to detect falls. By merging information from these different sensor types, the system achieves more reliable fall detection while reducing false positives, as no single sensor alone can reliably distinguish falls from other activities.
Solution Approach 2:
The system uses machine learning algorithms that analyze sensor data in context, incorporating feedback from multiple data sources to verify fall hypotheses. The system continuously refines its detection accuracy by evaluating patterns across multiple sensors and comparing against learned fall patterns, thereby improving reliability while maintaining precision.
2Reliability
If microphones and vibration detectors are used to identify falls, then detection capability is enhanced, but computational cost increases and location dependency is introduced
Solution Approach 1:
The patent extracts and utilizes data from commonly available mobile device sensors (accelerometer, barometer, gyroscope, magnetometer) that are already present in most smartphones and tablets. By taking out and leveraging these existing sensors rather than adding specialized hardware like microphones and vibration detectors, the system maintains high detection capability while avoiding the additional computational burden and location constraints of specialized sensors.
3Measurement precision
If multiple sensors and machine learning algorithms are used, then false positives are reduced, but device complexity increases
Solution Approach 1:
The system uses a multi-functional approach where a single machine learning processing unit handles analysis of data from multiple sensor types (accelerometer, barometer, gyroscope, magnetometer). This universal processing component reduces overall system complexity by consolidating the analytical functions that would otherwise require separate dedicated hardware for each sensor type.
Solution Approach 2:
The machine learning algorithms automatically learn and adapt to individual user patterns and contexts, performing self-service calibration and adaptation. This reduces the need for manual configuration and complex pre-programming, allowing the system to achieve high verification accuracy while maintaining relatively simple device architecture that adapts automatically to user behavior.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system effectively reduces false positive identifications by using sensor data and user context to verify fall hypotheses, providing accurate and location-independent fall detection with alerts for potential falls.
Implementation Method 1
Some prior art systems and devices employ accelerometers that measure sudden changes in acceleration that may indicate a fall
Implementation Method 2
A system utilizing multiple sensors, including motion sensors and barometers
Data Source
AI summary
A method of determining whether a user has fallen comprises detecting a potential fall using a motion sensing device, updating a probability of the potential fall being an actual fall based on an additional sensor, and updating the probability of the potential fall being an actual fall based on user context, the user context including an identified activity prior to the potential fall.