Inertial Sensor Worker Movement Risk Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning approaches require a large number of training samples to achieve accurate predictions, which is unethical or impractical in applications involving human injuries, leading to difficulties in creating sufficient datasets for supervised training.
Innovation Solution
A system utilizing inertial sensors attached to workers to generate movement data, which is analyzed by a machine learning model to identify predefined movement patterns, and then correlates these patterns with injury risks using a database, allowing for accurate injury prediction without the need for a large number of injured individuals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of training samples are used to train a machine learning model for injury prediction, then the prediction accuracy is improved, but the ethical and practical feasibility deteriorates
Solution Approach 1:
The patent introduces movement patterns as an intermediary representation between raw sensor data and injury outcomes. Instead of directly training on injury data, the system first identifies movement patterns from sensor data, then correlates these patterns with injury risks using a database. This intermediary layer enables accurate predictions without requiring direct training on large numbers of injured workers.
Solution Approach 2:
The patent segments the machine learning process into two distinct parts: (1) supervised learning to identify movement patterns from sensor data, and (2) expert system correlation to associate these patterns with injury risks. This segmentation allows the system to use small datasets for pattern identification while leveraging existing knowledge bases for injury correlation, resolving the contradiction between sample quantity and prediction accuracy.
2Reliability
If supervised training is used to predict injury outcomes directly, then the model can learn from real injury data, but the requirement for large datasets becomes problematic
Solution Approach 1:
The patent uses movement patterns as an intermediary that bridges sensor data and injury outcomes. The system identifies movement patterns from wearable sensor data and then correlates these patterns with injury risks using a database of established relationships. This approach maintains prediction reliability while eliminating the need for large datasets of injured workers, as the correlation step leverages existing expert knowledge.
Solution Approach 2:
The patent performs preliminary identification of movement patterns from sensor data before correlating with injury risks. By first establishing the movement patterns and their characteristics in advance, the system can then reliably associate these pre-identified patterns with injury outcomes using a database, avoiding the need to process and train on large quantities of injury data directly.
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
This approach enables accurate prediction of injury risks by splitting the machine learning process into supervised learning and expert system components, using a database to associate movement patterns with injuries, thus improving prediction accuracy without relying on extensive datasets of injured workers.
Implementation Method 1
multiple inertial sensors attachable to different body parts of the worker to generate inertial movement data
Data Source
AI summary
This disclosure relates to a sensor-based monitor for movements of workers. Multiple inertial sensors are attached to different body parts of the workers. A mobile device application receives the inertial movement data from the sensors and identify movement patterns by applying a trained machine learning model to the inertial movement data. The application then determine for each movement pattern an amount of time that movement pattern occurred and accessing a database to retrieve stored data on an association between the identified movement patterns and injuries. The application calculates a risk value indicative of a risk of injury of the worker as a result of performing the identified movement pattern for the determined amount of time and produces a report detailing the risk value for each of multiple risk categories.