Design of Carryable Respiratory Training System Using Learning-Based Waveform Classification

Spirometry is a crucial assessment for detecting respiratory obstructive diseases such as chronic obstructive pulmonary disease (COPD). While there are several respiratory trainers and spirometers available on the market, incorrect blowing techniques may render key metrics like forced vital capacity (FVC), forced expiratory volume in first second (FEV<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf>), and Peak Expiratory Flow (PEF) unreliable, necessitating supervised testing by healthcare professionals in hospital settings. Our system integrates the functionalities of a spirometer and respiratory trainer. After conducting pulmonary function tests using the App, users' blowing waveforms are measured with an airflow sensor. The App provides guidance for lung capacity tests and a gaming interface for respiratory training, storing all data in a MySQL database. This study categorizes common blowing waveforms into six types using machine learning models and dynamic time warping (DTW) distance calculations. Based on the classification results, users are directed to appropriate respiratory training sessions to help them master correct testing techniques. Analysis of the results indicates that using Mel Frequency Cepstral Coefficients (MFCC) feature extraction with Random Forest for classification yields optimal outcomes.