Adaptive Interference Suppression for Artifact-Resistant Measurement Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated evaluation of physical measurement data is hindered by noise and interference, which can be amplified during analysis, leading to misclassifications and artifacts, and existing preprocessing methods may exacerbate these issues.
Innovation Solution
A trainable module is used to select and apply a combination of interference suppression modules from a predefined catalog, optimized through a cost function to adaptively suppress noise and interference based on the specific disturbances present in the measurement signal, decoupling the selection process from the denoising method to enhance flexibility and convergence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If preprocessing is applied to measurement data to reduce noise and interference, then the quality of measurement data is improved, but artifacts are generated and interference influence is exacerbated
Solution Approach 1:
The patent applies dynamics by making the preprocessing approach adaptive rather than fixed. The system dynamically selects and adjusts filtering parameters based on the detected type and characteristics of interference in the measurement data. This allows the preprocessing to adapt to different noise patterns and interference types, reducing artifacts while maintaining noise reduction effectiveness.
Solution Approach 2:
The patent utilizes parameter changes by modifying filtering parameters such as kernel size, threshold values, and filter types based on the detected interference characteristics. The system changes these parameters adaptively to match the specific interference pattern, thereby optimizing noise reduction while minimizing artifact generation.
2Device complexity
If fixed preprocessing methods are used for all measurement data, then the processing pipeline is simple, but the suppression of interference is insufficient for diverse noise types
Solution Approach 1:
The system transitions from fixed preprocessing to dynamic, adaptive preprocessing. The trainable module continuously monitors the measurement data, detects interference types, and automatically adjusts the preprocessing parameters and filter selection accordingly. This dynamic approach maintains effectiveness across diverse noise types while keeping the user interface simple.
Solution Approach 2:
The system implements self-service by automatically detecting interference types and selecting appropriate preprocessing methods without requiring manual intervention. The trainable module self-adjusts based on the input data characteristics, eliminating the need for user configuration while maintaining high interference suppression effectiveness.
3Adaptability or versatility
If multiple interference suppression modules are applied to handle different noise types, then the adaptability is improved, but the processing time and computational load increase
Solution Approach 1:
The patent applies segmentation by dividing the interference suppression task into separate, specialized modules, each designed for specific noise types (e.g., Gaussian noise filter, salt-pepper noise filter, motion blur filter). The trainable module then selects only the relevant segments needed for each specific case, avoiding unnecessary processing and reducing overall time while maintaining high adaptability.
Solution Approach 2:
The system implements partial action by applying only the necessary interference suppression modules based on the detected noise type. Rather than always applying all possible filters, the trainable module selects the minimal effective set, reducing computational load and processing time while maintaining sufficient adaptability to handle various noise patterns.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Method (100) for training a trainable module (3) for processing recordings (11) of a measurement signal (10) containing interference (11b) comprising the steps: • a trainable module (3) is provided (110); • with the trainable module (3), a noise reduction module (2a-2d) or a combination of noise reduction modules (2a-2d) is selected for at least one training recording (11a) of the measurement signal (10) (120); • the selected noise reduction module (2a-2d), orThe selected combination is applied to the training recording (11a) (130) to obtain a noise reduction result (12); • using a predefined cost function (13), an evaluation (14) is determined (140) to what extent the noise (11b) is suppressed in the noise reduction result (12); • a parameter set (33) that characterizes at least the behavior of the trainable module (3) is optimized (160) with the aim of improving the evaluation (14) of the noise reduction result (12) obtained by the cost function (13) when the same training recording (11a) and/or further training recordings (11a) are processed again.