Design of a Wearable Dynamic Respiratory Monitoring System Based on a Distributed Inertial Measurement Unit

Inertial measurement units (IMUs)-based respiration signal detection provides a low-cost, noninvasive approach to measure the volumetric changes of the thorax and abdomen as a consequence of respiration. In view of the challenges of body motion interference to respiration detection in dynamic scenarios, this article proposes a distributed IMU respiratory detection method that allocates two IMUs onto the lower back and chest, respectively, to concurrently record the trunk motions and the respiration activities. Different positions of the IMUs are evaluated to optimize the effective capturing of the respiration signal and the suppression of motion interference. The principal component analysis (PCA) is applied to dynamically select the components under different motion modes to extract the respiratory signal, combining with a joint discrete Fourier transform (DFT) and empirical mode decomposition (EMD) denoising processing. Finally, the Savitzky-Golay filter (SGF) and the Butterworth bandpass filter (BPF) are used to optimize the respiratory signal quality. The proposed method is evaluated in standing, sitting, walking, and squatting motions under different subjects using the Xsens DOT IMU sensors for measurements, and the TI ADS1298R evaluation kit as the reference respiration signal recording by means of impedance pneumography. The experimental results show that the root mean square error (RMSE) between the measurements and the references under different activity states is less than 0.8, and their correlation coefficient is larger than 0.7, showing excellent measurement accuracy and robustness. The results show the potential of distributed IMU respiration measuring in dynamic scenarios and are expected to overcome complex motion interferences in various motion modes.