Feature Selection Algorithm for Noisy Biometric Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pattern recognition systems face challenges in dealing with noisy and sparse data, particularly in distinguishing between classes and reducing intra-class variances, which affects the accuracy of pattern recognition, especially in biometric applications.
Innovation Solution
The method involves applying inter-class and intra-class filtering to vector data based on distance metrics, specifically using Euclidean distance in all dimensions to reduce signal overlap between classes and within classes, thereby enhancing pattern recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional pattern recognition systems process noisy and sparse data, then they can handle real-world sensor data, but recognition accuracy deteriorates due to increased intra-class variances and inter-class similarities
Solution Approach 1:
The patent segments the data processing into distinct phases: feature extraction, inter-class filtering, and intra-class filtering. By dividing the complex task of handling noisy data into separate stages, each stage can focus on specific aspects (separating different classes, then separating instances within classes), thereby improving recognition accuracy without being overwhelmed by the complexity of the entire dataset.
Solution Approach 2:
The patent extracts and removes harmful factors (noise and redundancy) through filtering processes. Specifically, it extracts and eliminates features or data points that contribute to intra-class variance and inter-class similarity, retaining only the essential information needed for accurate recognition, thus improving reliability while reducing the harmful effects of noisy data.
2Reliability
If the system uses more data to improve recognition accuracy, then pattern recognition performance improves, but data collection time and computational requirements increase
Solution Approach 1:
The patent applies preliminary filtering actions during the training phase to pre-process and select the most informative features. By performing inter-class and intra-class filtering in advance, the system reduces the amount of data needed during actual recognition tasks, thereby improving accuracy without requiring extended data collection periods.
Solution Approach 2:
The patent changes parameters such as the number of features, dimensionality, and filtering thresholds to optimize the balance between accuracy and data requirements. By adjusting these parameters, the system achieves high recognition accuracy while minimizing the amount of data that needs to be collected and processed, thus reducing time loss.
3Reliability
If the system processes high-dimensional vector data, then pattern recognition capability improves, but computational complexity and processing time increase
Solution Approach 1:
The patent extracts and removes redundant dimensions and features from high-dimensional vector data through filtering processes. By identifying and eliminating features that contribute to intra-class variance and inter-class similarity, the system reduces computational complexity while maintaining pattern recognition capability, as only the essential features are processed thereafter.
Solution Approach 2:
The patent segments the high-dimensional data processing into manageable steps: extracting features, performing inter-class filtering to separate different classes, and intra-class filtering to separate instances within classes. This segmentation reduces the computational burden by processing data in structured phases rather than handling the entire high-dimensional dataset at once.
Data Source
AI summary
Application of inter-class and intra-class filtering, based on aggregate point-to-point distances, to vector data for purposes of filtering the vector data for purposes of pattern recognition. In some embodiments: (i) the inter-class filtering is based on Euclidean distance, in all dimensions, between vector data points in vector space; and/or (ii) the intra-class filtering is based on a distance, in all dimensions, between vector data points in vector space.