Unsupervised statistical method for multivariate identification of anomalous sensors.

Overview

The present invention relates to a method for identifying an anomalous sensor for measuring a characteristic of an individual, the method comprising:

• Collecting a curve of the characteristic of the individual measured by each sensor;

• For a given sensor and a reference curve, a processing step of calculating a difference index between this curve and each of the other curves of this sensor;

• A first iteration step in which, for each curve obtained from the same sensor, the processing step is iteratively repeated to determine a difference index for each curve;

• A second iterative step of determining a table of difference indices by performing previous steps for another sensor;

• Calculating an abnormality index for each individual from the multivariate statistical processing of the table; identifying abnormal individuals;

• Step of identifying an abnormal sensor.