Kernel-Based Activity State Classification Using Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Determining the state of activity of a person based on sensor data is challenging due to the dynamic and multivariate nature of the data, making it difficult to develop reliable methods for classification, especially across varying personal characteristics.
Innovation Solution
The use of kernel-based modeling techniques, specifically similarity-based models, to process sensor data and classify activity states by creating models for each class of activity, which can be either inferential or autoassociative, allowing for real-time classification without requiring extensive computational power or expertise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If rules are applied to raw sensor data to determine activity state, then classification can be performed, but the method is difficult to apply across variations in personal characteristics such as weight and height
Solution Approach 1:
The patent transforms the classification approach by changing from fixed rules to parameter-based modeling. Each activity state is represented by a model with parameters (mean vector and covariance matrix) derived from training data. These parameters adapt to individual characteristics, allowing reliable classification across variations in weight, height, and other personal attributes while maintaining classification capability.
Solution Approach 2:
The patent applies preliminary action by pre-processing sensor data to extract features and pre-computing activity models from training data before actual classification. The system pre-calculates mean vectors and covariance matrices for each activity state, and pre-processes incoming sensor data into feature vectors, enabling reliable real-time classification without complex runtime computations.
2Measurement precision
If similarity-based modeling is used to classify activity states, then accurate predictions can be achieved, but computational power requirements increase
Solution Approach 1:
The patent extracts only the essential components needed for similarity-based classification: mean vectors and covariance matrices from training data, and key feature extractions from sensor data. By extracting only these critical elements rather than processing complete raw datasets, the system achieves accurate activity state predictions while significantly reducing computational power requirements for real-time operation.
3Measurement precision
If complex rules are applied to handle dynamic multivariate sensor data, then classification accuracy may improve, but the system becomes difficult to operate without expert knowledge
Solution Approach 1:
The patent implements self-service by automatically computing activity models from training data and autonomously performing classification without requiring user intervention or expert knowledge. The system self-adjusts to individual characteristics through the learned parameters and automatically handles the complex multivariate analysis, making it easy to operate while maintaining high classification accuracy.
Data Source
AI summary
The activity state classification method of the present invention employs a kernel-based modeling technique, and more specifically a set of similarity-based models, which have been created using example data, to process an input observation or set of input observations, each comprising a set of sensor readings or “features” derived there from or other data, to predict the activity state of a person from whom the sensor data was obtained. A model is created for each class of activity. The input data is processed by each model and the resulting predictions are combined to yield a final prediction of which state of activity is represented by the input data.