Data Collation Coding Using Sampling-Range Average Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data collation techniques face challenges with high processing loads and long learning times due to large amounts of data required for image and sound recognition, making high-speed processing difficult.
Innovation Solution
A method of encoding target data by dividing it into sampling ranges, calculating average values and relative differences, and generating a code by concatenating numerical values as character string data, which reduces data size and facilitates efficient collation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image recognition technique uses large amount of reference data for comparison and collation, then identification accuracy is improved, but processing load increases and high-speed processing becomes difficult
Solution Approach 1:
The patent divides target data into multiple sampling ranges and generates separate codes for each range, then concatenates them to form a complete code. This segmentation approach reduces the complexity of comparing entire large datasets while maintaining identification accuracy through distributed feature representation across multiple smaller segments.
Solution Approach 2:
The patent extracts essential features from target data by calculating average values of data elements within each sampling range. This extraction process converts large amounts of raw data into compact numerical representations that retain the core characteristics needed for accurate identification while dramatically reducing processing requirements.
2Measurement precision
If image recognition technique processes large amount of target data, then identification accuracy is improved, but learning time becomes enormous
Solution Approach 1:
The patent performs preliminary encoding of target data into codes based on average values of data elements in sampling ranges before the actual recognition process. This pre-processing step creates compact representations that can be quickly compared and collated, eliminating the need for time-consuming processing of large raw datasets during the recognition phase.
Solution Approach 2:
The patent creates simplified copies of target data in the form of numerical codes that represent the essential characteristics of the original data. These codes serve as efficient proxies for the full datasets, enabling rapid comparison and identification without requiring access to or processing of the complete original data during the recognition process.
3Loss of information
If data encoding uses detailed data elements, then data fidelity is improved, but data size increases
Solution Approach 1:
The patent applies different encoding strategies to different sampling ranges of the target data, calculating average values locally within each range rather than applying a uniform encoding to the entire dataset. This local processing approach preserves important regional characteristics while reducing the overall data size through localized summarization.
Solution Approach 2:
The patent transforms detailed data elements into simplified numerical parameters representing average values within sampling ranges. This parameter transformation converts high-dimensional detailed data into low-dimensional compact representations, dramatically reducing data size while maintaining the essential information needed for accurate identification through the concatenation of these parameters.
Data Source
AI summary
A novel technology for encoding target data such as an image and an audio is provided. A code generation method for generating a code according to a content of target data using an information processing device is provided. The method includes a step of dividing the target data into a plurality of sampling ranges, a step of obtaining, for each of the sampling ranges, an average value of at least one data element among one or more types of data element included in each of the sampling ranges, each data element being represented by a numerical value, and a step of generating a reference code corresponding to the target data by concatenating, as character string data, the average values of the respective sampling ranges or numerals of a predetermined number of digits from a top digit of the average values.