Respiratory Pattern Detection via mmWave Radar & Classification

Batra, Rohan

Permalink

https://hdl.handle.net/2142/117039 Copy

Description

Title

Respiratory Pattern Detection via mmWave Radar & Classification

Author(s)

Batra, Rohan

Issue Date

2022-12-16

Keyword(s)

millimeter wave radar; abnormal respiration detection; classification; supervised learning; support vector machine; k-nearest neighbors

Abstract

Respiration is a vital physiological process that enables humans to obtain oxygen and generate energy while excreting waste gases. The regularity of respiration, characterized by respiratory rate and depth, can be assessed using various methods that track the rhythm of the lungs and heart. Abnormalities in respiratory patterns, such as frequency, rhythm, and depth, may indicate underlying diseases, particularly cardiac and pulmonary disorders. Therefore, detecting such respiratory anomalies can help mitigate the risks associated with these illnesses. In this regard, it would be advantageous to develop non-invasive methods for collecting respiration data, as this would eliminate the need for invasive techniques or cumbersome wearable devices. This paper presents a novel, non-invasive approach for obtaining respiratory waveforms using high-frequency (60 GHz) millimeter wave radar sensors, and a machine learning-based classifier for identifying respiratory patterns. The study was divided into two main sections: data collection, which yielded high-sample-rate respiratory frequency and depth data, and a supervised classification model that used extracted waveform features such as short-term energy intensity and peak-valley difference to identify six common respiratory patterns. The classifier, based on support vector machine (SVM) and k-nearest neighbors (KNN) algorithms, achieved an accuracy of 91.33% and 83.17%, respectively.

Type of Resource

text

Owning Collections