Feature Extraction Device for User State Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating a user's state based on application usage history in mobile terminals rely on basic statistics like the number of app starts, lacking information on sequence and correlation between applications, resulting in low accuracy.
Innovation Solution
A feature extraction device that acquires and processes activity history information to extract co-occurrence relationship information and generate activity history feature quantities, which are then used to improve state estimation accuracy by reflecting the user's state more accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If basic statistics such as the number of starts of an application are used for estimation, then the complexity of data processing is reduced, but the accuracy of estimating a user's state deteriorates
Solution Approach 1:
The patent segments the activity history information into multiple dimensions: basic statistics (number of starts), sequence information (order of application execution), and co-occurrence relationships (correlations between applications). Each dimension is processed separately to extract specific feature quantities, allowing the system to handle complex information in a structured manner while improving estimation accuracy.
Solution Approach 2:
The patent transitions from one-dimensional basic statistics to multi-dimensional feature extraction by adding sequence dimension (temporal order of activities) and correlation dimension (co-occurrence relationships between activities). This dimensional expansion enables the system to capture richer user behavior patterns without overwhelming computational complexity.
2Measurement precision
If sequence and correlation information of applications are incorporated, then the accuracy of estimating a user's state is improved, but the complexity of data processing increases
Solution Approach 1:
The patent divides the complex processing task into three separate extraction units: one for basic statistics, one for sequence information, and one for co-occurrence relationships. Each unit independently processes its specific aspect of the data, extracting relevant feature quantities without interfering with other processing streams, thereby managing complexity systematically.
Solution Approach 2:
The patent merges the output feature quantities from the three separate extraction units (basic statistics features, sequence features, and co-occurrence features) into a unified feature set that is input to the state estimation model. This combination integrates multiple dimensions of information while leveraging the modular architecture to maintain processing efficiency.
3Ease of manufacture
If only the number of starts of an application is used, then the information processing is simplified, but information such as the sequence of using applications and the correlation between applications is lost
Solution Approach 1:
The patent segments information extraction into distinct categories: basic statistics (simple counting), sequence information (temporal ordering), and co-occurrence relationships (correlation analysis). Each segment processes its specific type of information using appropriate methods, preserving sequence and correlation data that would otherwise be lost in simplified processing.
Solution Approach 2:
The patent adds temporal and relational dimensions to the basic counting approach. Instead of merely counting application starts, the system tracks the order of execution (temporal dimension) and the relationships between co-occurring applications (relational dimension), thereby preserving critical behavioral information while maintaining processing feasibility.
Data Source
AI summary
A feature extraction device is a device that extracts a feature quantity to be input to a state estimation model for estimating a state related to a terminal where the activity is executed or a user of the terminal, which includes an acquisition unit that acquires activity history information indicating an execution history of the activity in a specified period of time and including time-series activity logs, an extraction unit that extracts co-occurrence relationship information indicating a co-occurrence relationship of activities in a specified period of time based on the activity history information, a generation unit that aggregates activity feature quantities indicating a feature of each activity based on the co-occurrence relationship information and generates an activity history feature quantity indicating a feature of the activity history information, and an output unit that outputs the activity history feature quantity as a feature quantity to be input to the state estimation model.