Effect of Respiratory Effort Levels on SCG Signals

In this paper, an analysis of thoracic vibrations induced by cardiac activities is presented under two different breathing levels. These levels are associated to respiratory efforts in normal and stopped manner. Usually, the vibrations on the thoracic cavity are captured by the means of seismocardiogram (SCG) signal. Some morphological features are explored in this work towards the development of a SCG based discriminative method for characterizing these levels. The proposed method consists of five major steps: signal segmentation, heartbeat extraction with the help of annotations of aortic valve opening (AO) instants, synchronized beat averaging, feature-extraction, and featureselection scheme. Strengths of the proposed features kurtosis (K), energy of diastolic region (DE), amplitude of IM or IC instant (IA) and all their combinations are measured using two feature selection tests for better discrimination. Among all variations, the group of all three features reveals the strongest separability for these two classes. Whereas, an excellent separation is achieved alone with the DE as an individual feature. The quantitative analysis of the performance results reveals the optimum feature-set for breathing-level classification in three different dimensions.