Respiratory Status Classifier Using Extracted Frequency Components
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for diagnosing sleep apnea and snoring do not effectively generate a respiratory status classifier using respiratory sound signals, lacking a method to accurately determine respiratory status through a trained classifier.
Innovation Solution
A method is developed to segment respiratory signals into cycles, select specific frequency components using power spectrums, and train a classifier to differentiate between normal, snoring, and apnea states, reducing computational load while maintaining high accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If respiratory signals are fully processed using all frequency components, then classification accuracy may be improved, but computational load increases significantly
Solution Approach 1:
The patent extracts only the essential frequency components from the respiratory signal spectrum. By identifying and removing redundant frequency components that do not contribute to respiratory status classification, the system maintains high classification accuracy while significantly reducing the computational load required for processing.
Solution Approach 2:
The patent creates a simplified spectral representation that copies only the critical features needed for classification. Instead of processing the complete frequency spectrum, the system generates a reduced spectral model containing only the essential frequency components, achieving accurate classification with minimal computation.
2Reliability
If a comprehensive respiratory status classifier is developed using all available signal data, then diagnostic reliability is improved, but device complexity increases
Solution Approach 1:
The patent segments the respiratory signal analysis into distinct frequency components and identifies which segments are essential for reliable classification. By dividing the spectral analysis into manageable frequency bands and selecting only the critical ones, the system achieves high diagnostic reliability while keeping the classifier structure simple and manageable.
Solution Approach 2:
The patent applies local quality by assigning different importance weights to different frequency components. Instead of treating all frequency components equally, the system identifies specific frequency regions that locally contribute most to respiratory status differentiation, thereby simplifying the overall classifier while maintaining high reliability.
3Measurement precision
If real-time respiratory status determination is implemented with full spectral analysis, then measurement accuracy is improved, but processing speed decreases
Solution Approach 1:
The patent extracts only the critical frequency components needed for real-time respiratory status determination. By removing unnecessary spectral calculations and retaining only the essential frequency bands, the system achieves both high measurement accuracy and fast processing speed suitable for real-time monitoring.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are a method of generating a respiratory status classifier and a method of determining respiratory status. The method of generating a respiratory status classifier includes segmenting collected respiratory signals according to respiratory cycles, acquiring data pairs of a calculated power spectrum of frequency components and a designated respiratory status indication value for a plurality of cycle-specific respiratory signals selected from among the segmented cycle-specific respiratory signals, selecting specific frequency components from among the frequency components, inputting data pairs of a power spectrum and a designated respiratory status indication value corresponding to the specific frequency components to a respiratory status classifier, and training the respiratory status classifier according to output values from the classifier.