The concepts of human seat occupancy detection and driver’s drowsiness monitoring require a sophisticated, sensing technology capable of capturing human vital signs in a reliable manner. The concept discussed in this paper may help enable the development of future systems capable of detecting an occupant in a seat.

The present study explores the feasibility of detecting humans based on a polymer sensor fitted into the seat cushion and capable of capturing human vital signs. A bulk, polypropylene ferroelectric film has been charged and polarized in a strong external electric field prior to the sensor assembly. The resulting 323 sq cm sensors displayed a high piezoelectric d33 coefficient of approximately 200 pC/N, considerably higher than vibration sensors made of PVDF or PVDF-TR piezoelectric films. This type of electroresponsive polymer has been used for medical respiration, heartbeat and epileptic seizure monitors.

We employed dedicated, microprocessor-based electronics including charge and variable gain amplifiers and 4th-order anti-aliasing filter for data collection. Three different types of algorithms have been fitted or developed and tested:​ i) a commercial medical monitor with estimation of respiratory and heart beat rates, ii) a signal extraction, filtering and matching wavelet-based algorithm for vital sign detection and (iii) a frequency domain, 2nd-order classifier for humans/objects, using knowledgebased discrimination.

Experimental data involved a minimum of 20 human subjects ranging from a 5-month old infant in a child restraint to a 95th%ile male, both in fully static (sleeping like) and non-static scenarios. Recordings using test loads and a pack of water bottles were also collected as the counterpart to the passengers.

Human-specific presence detection and discrimination from objects by detection of vital signs was achieved within a relatively short detection time in this conceptual study. Infants and small children were placed in dedicated child restraint seats (CRS) and not moved during the data collection, thus simulating sleeping children. All subjects were detected typically within a 20 seconds sampling interval. In a few cases and with additional time, their respective signals could be extracted from collected data as confirmed by a medical monitor used in parallel.