Extraction of Respiratory Rate from Impedance Signal Measured on Arm: A Portable Respiratory Rate Measurement Device

In this paper, respiratory rate is extracted using signal processing and machine learning methods from electrical impedance, measured across arm. Two pairs of electrodes have been used along the arm, one for injecting the current, and one for sensing the voltage. After filtering, the frequency components and other signal features have been extracted using Short Time Fourier Transform (STFT). Then a Support Vector Machine (SVM) model is trained to detect the breath-holding state. Frequency components and signal features of the parts of the signal that are detected to be representing the breathing state are then fed into another SVM model that extracts the respiratory rate and reduces the effect of motion artifacts. A similar method has been applied to the signal taken from end-tidal CO2 respiratory measurement device as the reference signal. This signal has been used as the ground truth for training of the SVM model and for validation of the method. The results are validated using 5-fold cross-validation method. Statistical analysis confirms the significance of the introduced features.