The Pressure Bed Sensor (PBS), which is presented as a contactless sensor for physiological signals recording, allows the acquisition of respiration movements signal. The aim of the present study is to identify spectral parameters from the PBS respiratory signal that allow the discrimination between normal and abnormal breathing epochs.
The nasal airflow and the PBS respiratory signal acquired on 19 subjects were pre-processed in order to obtain their positive envelope signals. Both of them were analyzed by means of an optimized Time-Variant Autoregressive Model (TVAM). Total sleep time was divided into consecutive epochs of 60 s classified as normal and abnormal (at least one apnea or hypopnea). The mean Power Spectral Density (PSD) for each sleep epoch was estimated from the averaged TVAM coefficients. Spectral features were extracted from both the nasal airflow and the PBS respiratory signal. A statistically significant difference (p-value<0.01) between normal and abnormal breathing epochs has been found in all the considered spectral features for both the nasal airflow and the PBS respiratory signal. These results suggest that the discrimination between normal and abnormal breathing epochs is thus possible by using parameters obtained from an easy-to-use, comfortable and non-obtrusive system for sleep monitoring, such as the Pressure Bed Sensor.