Test of fire detection algorithms using artificially generated events

Abstract

The paper proposes a test method that is particularly suitable for the test of the false alarm rejection features of detection algorithms. It describes how to generate signal events artificially, which are similar to those occurring in practice, and thus may or may not cause false or unwanted alarms. The generated events are of random nature, i.e. the signal shape, amplitude and duration of the events is randomly varied and the points in time, where these events occur, are randomly selected according to a well defined signal model. A set of parameters controls the event density, the probability for the duration of events, the fluctuation and the full dynamic range. The whole signal model uses only one pseudo-random generator and generates a modified “random walk process”. Thus, it is either possible to reproduce with the same set of parameters for the signal model the same random sequence of events or, alternatively, to generate different random signal sequences using a different seed for the pseudo-random generator. The test procedure is carried out in the form of a computer simulation in time-lapse mode, and thus requires the detection algorithm to be available as a C-program.