Sensor Data Recognition With Cross-Correlation Feature Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recognition technologies fail to accurately identify specific types of sensor data due to shared contribution values across different sensor data, leading to inefficiencies in class identification.
Innovation Solution
A recognition apparatus that generates a first feature quantity from sensor data, converts it into a second feature quantity contributing to class identification, calculates a significant feature quantity based on cross-correlation, and integrates these features to enhance class identification precision using neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the same contribution values are shared among entire sensor data, then the processing is simplified, but the precision of class identification deteriorates because unique sensor features are not reflected
Solution Approach 1:
The patent segments the contribution values by creating separate attention mechanisms for different sensor types. Each sensor's contribution is evaluated independently through dedicated attention layers, allowing unique features of each sensor type to be captured without being averaged or shared across all sensors, thus improving class identification precision while maintaining manageable processing complexity
Solution Approach 2:
The patent applies local quality by making the attention mechanism sensor-specific rather than global. Each sensor type has its own attention weights and contribution values that are locally optimized for that sensor's characteristics, enabling the system to capture sensor-specific features while keeping the overall architecture organized and computationally efficient
2Power
If feature quantities are processed with shared attention information, then the computation is reduced, but the ability to capture sensor-specific features is lost
Solution Approach 1:
The computational process is segmented into sensor-specific attention layers that process each sensor type independently. This segmentation allows the system to capture reliable sensor-specific features while maintaining computational efficiency by avoiding redundant cross-sensor attention computations and focusing processing power where it is most needed for each sensor type
3Device complexity
If a single attention mechanism is used for all sensor data, then the model structure is simplified, but the adaptability to different sensor configurations deteriorates
Solution Approach 1:
The patent implements a universal attention mechanism framework that can adapt to different sensor configurations. The same basic attention layer structure is universally applied across multiple sensor types, but each layer is configured to handle the specific characteristics of its assigned sensor type, providing both structural simplicity and configuration adaptability
Solution Approach 2:
The attention mechanism is designed to be dynamic and configurable based on the specific sensor configuration being used. The system can dynamically adjust which attention layers are active and how they process different sensor types, allowing the model to adapt to various sensor configurations without requiring a completely different architecture for each case
Data Source
AI summary
According to one embodiment, a recognition apparatus includes processing circuitry. The processing circuitry generates a first feature quantity exhibiting a feature of sensor data based on the sensor data, converts the first feature quantity into a second feature quantity exhibiting a feature contributing to identification of a class of the sensor data, generates a significant feature quantity exhibiting a feature that is significant in the identification of the class based on a cross-correlation between the first feature quantity and the second feature quantity, generates an integrated feature quantity considering features of the first feature quantity and the second feature quantity, based on the second feature quantity and the significant feature quantity, and identifies the class based on the integrated feature quantity.