Recognition Dictionary Generating Device for Pattern Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pattern recognition techniques are inefficient due to the processing of redundant reference vectors, which consume significant time without improving recognition accuracy.
Innovation Solution
A recognition dictionary generating device that selects and corrects reference vectors using attraction and repulsion forces, and determines the importance of each vector through an offset value, excluding redundant vectors to enhance processing speed without compromising accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all reference vectors are processed during pattern recognition, then recognition accuracy is maintained, but processing time increases significantly
Solution Approach 1:
The patent extracts and removes redundant reference vectors from the recognition dictionary that do not contribute to recognition accuracy. By identifying and eliminating these unnecessary vectors, the system reduces the number of vectors requiring processing during pattern recognition, thereby decreasing processing time while maintaining recognition accuracy.
Solution Approach 2:
The patent introduces an offset value parameter for each reference vector to indicate its importance. By changing the parameter set to include this offset value and using it as a criterion for selecting reference vectors, the system can prioritize important vectors and exclude redundant ones, resolving the contradiction between processing all vectors for accuracy and processing fewer vectors for speed.
2Productivity
If redundant reference vectors are excluded from the recognition dictionary, then processing speed increases, but recognition accuracy may deteriorate
Solution Approach 1:
The patent uses the offset value as a feedback mechanism to guide the selection of reference vectors. During the learning process, the offset values are updated based on the importance of each reference vector, and this feedback is used to determine which vectors to exclude. This ensures that only truly redundant vectors are removed, maintaining recognition accuracy while improving processing speed.
Solution Approach 2:
The patent performs preliminary analysis during the learning phase to identify and mark redundant reference vectors using offset values. This preliminary action allows the system to prepare an optimized recognition dictionary in advance, so that during actual pattern recognition, only important vectors need to be processed, achieving both speed and accuracy.
Data Source
AI summary
A recognition dictionary generating device includes a unit that acquires plural reference vectors each containing an offset value indicating a degree of importance; a unit that selects a first reference vector belonging to the class same as an input vector and having the minimum distance from the input vector, and a second reference vector belonging to a class different from the input vector and having the minimum distance from the input vector; a unit that acquires a first distance value indicating a distance between the input vector and the first reference vector and a second distance value indicating a distance between the input vector and the second reference vector; a unit that corrects the first reference vector and the second reference vector using a coefficient changing in accordance with a relationship between the first distance value and the second distance value, the first distance value, and the second distance value; and a determining unit that determines a reference vector to be excluded from a recognition dictionary in accordance with the offset value of the corrected first reference vector and second reference vector.