Mobile Activity Model Calibration with Personalized Sensor Weights
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user activity models on mobile devices are not customized to individual users, leading to inaccuracies in activity recognition, particularly for minority groups such as children, elderly, or people with disabilities, due to variations in movement patterns.
Innovation Solution
A method to calibrate user activity models on mobile devices by applying personalized weights to sensor data using a gradient boosting machine (GBM) model, without retraining, through a transfer learning approach, utilizing a smaller, individualized data set to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a general user activity model is used for all users, then the device complexity is reduced and ease of manufacture is improved, but the measurement precision and reliability of activity recognition deteriorate for individual users
Solution Approach 1:
The system performs preliminary calibration by collecting sensor data during normal device usage and automatically generating personalized weights before actual activity recognition begins. This preliminary action creates a customized model specific to each user without requiring manual intervention or complex setup procedures.
Solution Approach 2:
The calibration process is fully automated and self-service oriented. The system automatically collects sensor data, processes it through the gradient boosting machine model, generates personalized weights, and applies them to improve activity recognition accuracy without requiring user intervention or expert involvement.
2Measurement precision
If personalized calibration is applied to each user, then the activity recognition accuracy is improved, but the loss of time for data collection and processing increases
Solution Approach 1:
The system collects calibration data during normal device usage periods, performing the calibration action preliminarily before actual activity recognition begins. This allows the system to gather necessary sensor data without dedicating separate calibration sessions, thereby reducing the perceived calibration time for users.
Solution Approach 2:
The calibration process operates continuously in the background during normal device usage. Instead of interrupting users for separate calibration sessions, the system continuously collects sensor data and processes it through the gradient boosting machine model, maintaining useful action continuity while performing calibration.
3Reliability
If standard sensor data weighting is used, then the device complexity is minimized, but the reliability of activity recognition for diverse user groups deteriorates
Solution Approach 1:
The system applies local quality by generating personalized weights specific to each user's movement patterns and characteristics. Instead of using uniform weighting for all users, the gradient boosting machine model analyzes individual sensor data and creates localized weight adjustments that optimize activity recognition reliability for each specific user group and individual.
Solution Approach 2:
The system changes the weighting parameters of sensor data based on individual user characteristics. The gradient boosting machine model processes sensor data and dynamically adjusts the weight parameters assigned to different sensors and sensor readings, transforming the standard uniform weighting into customized parameter sets that improve reliability for diverse user groups.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems, computer-implemented methods, and computer program products that can facilitate calibrating a user activity model of a user device nodes are described. According to an embodiment, a method for calibrating a user activity model used by a mobile device can comprise receiving sensor data from a sensor of the mobile device. Further, applying a first weight to a first a first likelihood of a first occurrence of a first activity, wherein the first likelihood is determined by a first estimator of the user activity model by applying preconfigured criteria to the sensor data. The method can further comprise performing an action based on a determination of the first occurrence of the first activity, the determination being based on the first weight and the first likelihood of the first occurrence of the first activity.