Waveform Segmentation for Sensor Noise and Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing waveform segmentation methods face challenges in accurately extracting feature values from sensor data due to spike-like noise and irregular amplitude fluctuations, which can lead to incorrect segmentation and compromised abnormality detection.
Innovation Solution
A waveform segmentation device with a state level estimation unit and segmentation identification unit that segments waveform data at multiple points based on estimated state levels, using conversion and adjustment units to refine segmentation points and feature value extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If waveform shaping is performed on each partial waveform to remove noise components, then noise removal is improved, but the difference from the normal waveform decreases making abnormality detection difficult
Solution Approach 1:
The waveform is automatically segmented into multiple partial waveforms based on detected segmentation points. By dividing the waveform at appropriate points, the system can process each segment separately while preserving the characteristic features needed for abnormality detection, thus avoiding the need for aggressive waveform shaping that would eliminate important differences from normal waveforms.
2Productivity
If automatic waveform segmentation is performed to extract feature values, then feature extraction efficiency is improved, but feature values may be erroneously divided into different partial waveforms
Solution Approach 1:
The system performs preliminary waveform shaping on the entire waveform before segmentation, and uses preliminary detection of segmentation points based on waveform characteristics. This preliminary action prepares the waveform data in advance, allowing subsequent segmentation to occur at appropriate points that preserve feature value integrity while maintaining extraction efficiency.
Solution Approach 2:
The system iteratively adjusts segmentation points based on feedback from waveform analysis. By continuously evaluating the waveform characteristics and adjusting segmentation points accordingly, the system ensures that feature values are not erroneously divided while maintaining automatic segmentation efficiency.
3Speed
If waveform data with spike-like noise and irregular fluctuations is processed directly, then data processing speed is improved, but feature value extraction accuracy deteriorates
Solution Approach 1:
The system performs preliminary waveform shaping on the entire waveform before segmentation to smooth out spike-like noise and irregular fluctuations. This preliminary action removes harmful noise components in advance, allowing subsequent processing to proceed efficiently with accurate feature value extraction from the cleaned waveform data.
Data Source
AI summary
A waveform segmentation device has a state level estimation unit that estimates a state level of input waveform data, and a segmentation identification unit that segments the waveform data at a plurality of segmentation points based on the state level estimated by the state level estimation unit. The segmentation identification unit may identify the plurality of segmentation points such that a feature value of the waveform data is included between two adjacent segmentation points among the segmentation points.