An ICA approach to reject noise from pressure changes of pillow

An approach based on independent component analysis (ICA) was presented to reject noise and extract respiration signals and pulse signals, which are contained in pulsatile signals measured from a pillow cuff installed on the occiput of a user. Because there are only two observations for a measuring system that is a linear mixture of three sources (pulse, respiration, and noise), usual ICA cannot separate them perfectly. To solve this problem, a two-step algorithm was proposed. The first step is a standard ICA and the second step is filter processing which uses two band pass filters. In the ICA step, noise can be removed from the mixture of pulse and respiration signals. The filter processing then further separates the pulse and respiration. In the primary research, data for 10 minutes of data were recorded from six subjects, and their verifications were presented as experimental results. The obtained results were compared with the reference pulse and respiration signals that were measured simultaneously with pillow cuff monitoring. Results showed that this approach was effective in separating pressure signals observed by the pillow system into vital signs and noise.