Signal Feature Extraction Using Wavelet Transform and Gaussian Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting signal features from biosignals, such as photoplethysmogram (PPG) signals, face challenges in accurately distinguishing feature points like the dicrotic notch, leading to incorrect identification of propagation and reflection waves, which affects the estimation of cardiovascular information like blood pressure and vascular stiffness.
Innovation Solution
A method and apparatus that estimate element signals from an input signal using a signal model, determining parameters based on derivative signals and feature points, and iteratively extracting signal features like maximum, minimum, and peak points by overlapping Gaussian waveforms, allowing for accurate extraction of features even in unclear waveforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing methods are used to extract signal features from biosignals, then the process is simple, but the accuracy of distinguishing feature points like dicrotic notch deteriorates
Solution Approach 1:
The patent segments the complex biosignal into multiple element signals (propagation wave, reflection wave, and other component waves) and processes each separately. By dividing the signal decomposition into sequential steps where each element signal is estimated and eliminated in turn, the method achieves accurate feature point identification while managing computational complexity through structured segmentation of the analysis process.
Solution Approach 2:
The patent transforms the signal analysis from time-domain only to include frequency-domain analysis by applying continuous wavelet transform. This dimensional change allows the method to identify maxima ridges in the scalogram at scales with characteristic frequencies, providing additional information for accurate feature point detection that cannot be obtained from time-domain analysis alone.
2Reliability
If simple signal processing methods are used, then the computational load is low, but the reliability of cardiovascular information estimation deteriorates
Solution Approach 1:
The patent applies preliminary filtering and preprocessing steps before the main signal decomposition process. By pre-processing the biosignal to remove noise and prepare it for analysis, the method ensures more reliable feature extraction while reducing the computational energy needed in subsequent processing steps, as the data is already optimized for analysis.
Solution Approach 2:
The patent replaces traditional mechanical signal processing methods with wavelet transform-based analysis. This substitution allows for more accurate identification of signal features through mathematical transformation in the frequency domain, improving reliability of cardiovascular estimation while the efficient algorithms keep computational energy consumption manageable.
3Measurement precision
If continuous wavelet transform is applied to identify maxima ridge, then frequency resolution is improved, but processing time increases
Solution Approach 1:
The patent applies wavelet transform selectively to identify only the critical maxima ridges corresponding to the propagation and reflection waves, rather than analyzing the entire frequency spectrum in detail. This partial action approach achieves sufficient frequency resolution for the specific cardiovascular features of interest while reducing overall processing time by focusing computational resources on the most relevant signal components.
Data Source
Figure 1A
Figure 1B~1E
Figure 2
AI summary
A signal feature extracting method and apparatus is disclosed. The signal feature extracting apparatus estimates element signals forming an input signal using a signal model to be determined by parameters, and extracts signal features using the estimated element signals. The method of extracting a signal feature including estimating element signals from an input signal, and extracting a signal feature using the estimated element signals, wherein the estimating of the element signals comprises estimating a first element signal of the input signal, and estimating a second element signal based on a waveform of a first intermediate signal, the first intermediate signal being a signal derived from the first element signal eliminated from the input signal