Noise Cause Identification via Spectrum Image Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The densification of electronic devices has made it difficult and costly to identify noise causes, increasing the time and effort required for effective electromagnetic interference (EMI) countermeasures, necessitating a more efficient method for noise cause identification.
Innovation Solution
An analysis system comprising an inference processing unit, a causal component identification unit, and a presentation information generation unit, which uses learned models to classify frequency spectrum data, identify causal components, and generate presentation information for users, thereby reducing the workload and improving the accuracy of noise cause identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional manual observation and analysis of frequency spectrum data is used, then worker expertise and experience can identify noise causes, but the process becomes increasingly time-consuming and costly due to device densification
Solution Approach 1:
The patent replaces the manual mechanical observation and analysis process with an automated image processing system. The frequency spectrum data is transformed into visual images, and machine learning models automatically identify noise sources, substituting human expert analysis with automated computational methods. This resolves the contradiction by maintaining identification accuracy while dramatically reducing the time required for noise cause analysis.
Solution Approach 2:
The patent creates visual copies (images) of the frequency spectrum data that can be processed and analyzed more efficiently than raw numerical data. By converting spectral information into visual representations, the system enables faster pattern recognition and noise source identification, reducing analysis time while preserving diagnostic accuracy.
2Ease of operation
If manual noise cause identification is performed, then flexible analysis is possible, but the workload and complexity increase with device densification
Solution Approach 1:
The patent replaces complex manual analysis procedures with automated image processing and machine learning algorithms. The system handles the complexity of device densification automatically through computational methods, making the noise identification process easier to operate while managing the increased system complexity through automation rather than manual procedures.
Solution Approach 2:
The patent introduces visual images as an intermediary representation between the raw frequency spectrum data and the noise cause identification process. This intermediate visual format simplifies the analysis by making patterns and noise sources more readily apparent, reducing the operational complexity while maintaining analysis effectiveness.
3Measurement precision
If more comprehensive analysis is performed to identify noise causes in dense device environments, then identification accuracy improves, but the workload and cost increase
Solution Approach 1:
The patent uses automated machine learning-based image processing to perform comprehensive analysis of frequency spectrum data. The system can analyze complex patterns in dense device environments automatically, maintaining high identification accuracy while improving productivity by eliminating manual analysis bottlenecks and reducing the time required to process comprehensive data sets.
Data Source
AI summary
An analysis system includes inference processer circuitry configured to infer a corresponding classification by inputting part of frequency spectrum data corresponding to reference measurement data to a learned model having learned a relation between part of frequency spectrum data corresponding to sample measurement data and a classification related to noise corresponding to the part, causal component identification processer circuitry configured to identify causal component data of noise from a component data list based on the inferred classification, and a presentation information generator configured to generate presentation information for a user based on the causal component data.