System monitoring and diagnosis is a research areathat has received a growing interest from both thesignal processing community and the control commu-nity in the recent years. Indeed, the development ofnew sensors makes it possible to record variables on asystem that were not acquired previously. These newdata provide useful information on the inner evolutionof the system. Moreover, complex numerical treatmentcan be easily performed due to the computationalcapabilities and storage capacities of today processorsand computers. Fed by this information, supervisionsystems aim at detecting abnormal situations so asto increase the system safety. Several applicationsareas, such as sewer networks [1], mining industriesand mineral flotation plants [2] or aeronautics [3],have been benefited from the recent advances in thediagnosis field.In medicine, the system monitored is a patient,which makes the notion of safety all the more impor-tant. Therefore, reliability is a strong requirement fora medical monitoring system. It must detect promptlyany degradation in the patient’s state and avoid falsealarms. Note that missed detections obviously decreasethe physician’s reliance, whereas false alarms decrease