Biomedical signals are difficult to detach from noise during daily activities. Classical linear techniques attenuate noise, but frequently are based on strict assumptions.
Impedance plethysmogram (IPG) and dual-wavelength reflection photoplethysmogram (PPG) sensing units were developed and inconspicuously embedded in a wheelchair to avoid further diminution of the subject’s autonomy and mobility. It is of interest to gather these signals, since often wheelchair users need permanent monitoring of their cardiopulmonary system.
However, IPG and PPG are sensitive to the wheelchair user’s movements, which produce significant baseline fluctuations. To lessen such problems, real-time signal processing techniques have to be hastily applied before any cardiovascular evaluation to ensure its accuracy.
This paper presents a study on the combination of several procedures (PCA, SWT, and EMD) to enable adaptive selection of signal-related components, for a described hardware implementation on a manually powered wheelchair.