UTO-LAB Model: USRP Based Touchless Lung Anomaly Detection Model with Optimized Machine Learning Classifier

In the light of the rising incidence of breathing illnesses and the significance of early detection for effective treatment. In modern healthcare, noninvasive and touchless clinical diagnostic methods are becoming increasingly important for detecting lung anomalies. In this work, a novel Universal software radio peripherals (USRP) based TOuchless Lung ABnormality (UTO-LAB) detection framework is introduced that harnesses using a machine learning (ML) algorithm. This UTO-LAB framework employs a blood pressure meter and Red Green BlueDepth (RGB-D) sensing camera for capturing information of an individual without any tactile interaction. The heart rate (HR) is analyzed by converting the images into imaging photoplethysmography (iPPG) signal and blood pressure (BP) from the blood pressure meter after the Analog conversion. This touchless arrangement enables the extraction of the most relevant signal features vital for the analysis of respiratory patterns and abnormalities. The acquired data is processed using advanced computer vision techniques such as Updated Melfrequency cepstral coefficients (UMFCC) and Principle component analysis (PCA) to extract the most relevant features focusing on breathing anomalies. These extracted features are fused with a multiplier and given to the Aquila optimized multiclass support vector machine (OMSVM) to analyse breathing activities into normal, shallow, and elevated breathing. The efficacy of this UTO-LAB framework is assessed with the real-time and simulated data. Based on the experimental results, the UTO-LAB framework achieves accuracy of 96.23 % for real-time data and 99.62 % for simulated data when analyzing the respiratory anomalies.