A robust signal processing method for extraction of respiratory activity from artifact corrupted PPG signal

Continuous monitoring of respiratory activity is required in many clinical settings from emergency departments to operating theatres and high dependency intensive care units. Current techniques such as nasal thermistors, capnography and monitoring of transthoracic impedance are prone to movement artifact or difficult to use in a continuously ventilating patient. To help overcome some of these problems, it was reported in literature that it is possible to extract respiratory activity from the Photoplethysmogram (PPG) using wavelet transform based methods. In this paper, we present a robust yet simple ICA based signal processing method to extract the surrogate respiratory activity from non-invasive recordings of artifact corrupted PPG signals. To prove the efficacy of proposed method, experimental results are carried out considering two commonly encountered motion artifacts of PPG signals. Simulation results and statistical calculations in terms of Magnitude Squared Coherence (MSC), Relative Correlation Co-efficient (RCC) and Accuracy Rate (AR) indicated a strong correlation between the extracted respiratory signal with that of original respiratory signal.