Wearable Fall Detection Using Compressed Local Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for detecting falls in senior and disabled care facilities lack efficient and accurate methods to promptly respond to falls, often resulting in delayed or incorrect responses due to limitations in wearable device power consumption and processing capabilities.
Innovation Solution
A method utilizing a wearable device with a compressed fall detection model that locally detects falls and intermittently communicates with a remote computer system executing a complete fall detection model, optimizing power usage while maintaining high accuracy through selective data upload and processing, thereby reducing false positives and negatives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a complete fall detection model is executed continuously on the wearable device, then fall detection accuracy is improved, but power consumption increases
Solution Approach 1:
The fall detection system is segmented into two parts: a compressed model running continuously on the wearable device for initial detection, and a complete model running on the remote computer system for confirmation. This segmentation allows the wearable device to use minimal processing while maintaining detection capability, and the complete model provides accuracy when needed without continuous power consumption.
Solution Approach 2:
Instead of continuous execution of the complete model, the system uses periodic action by only uploading sensor data for complete model analysis when the compressed model detects a potential fall event. This periodic processing significantly reduces power consumption while maintaining high detection accuracy for actual fall events.
2Measurement precision
If sensor data is uploaded continuously to the remote computer system, then fall detection accuracy is improved, but data transmission power consumption increases
Solution Approach 1:
The system applies partial action by uploading only the necessary subset of sensor data (corpus of sensor data corresponding to a duration terminating at approximately the first time) to the remote computer system when a fall event is detected, rather than continuously uploading all sensor data. This reduces transmission power consumption while providing sufficient data for accurate fall confirmation.
3Use of energy by moving object
If a compressed fall detection model is used on the wearable device, then power consumption is reduced, but false positive rate increases
Solution Approach 1:
The system uses feedback by having the remote computer system receive the corpus of sensor data and cue for fall event confirmation, then apply the complete fall detection model to verify whether the fall event actually occurred. This feedback loop allows the compressed model to operate with lower thresholds for initial detection while the complete model filters out false positives, maintaining both low power consumption and high reliability.
Data Source
AI summary
One variation of a method for detecting and responding to falls by residents within a facility includes: at a wearable device worn by a resident, writing sensor data from a sensor integrated into the wearable device to a buffer, inputting sensor data into a compressed fall detection model—defining a compressed form of a complete fall detection model and stored locally on the wearable device—to detect a fall event at a first time, and transmitting a corpus of sensor data from the buffer and a cue for confirmation of the fall event to a local wireless hub in response to detecting the fall event; and, remotely from the wearable device, inputting the corpus of sensor data into the complete fall detection model—stored remotely from the wearable device—to confirm the fall event and dispatching a care provider to assist the resident in response to confirming the fall event.