Online multitechnology sensors explosive recognition

A chemometric method based on an hypothesis test approach built upon RLS (Recursive Least Square) algorithm is presented. This method offers an alternative for online recognition: each analyzed sample leads to the convergence of the system to a state characteristic of the presence of a chemical compound. We present how hypothesis tests are used to exploit this state. In particular we detail some adaptations of RLS used in order to apply our method to a sensor matrix of several technologies in optimal conditions. We show the efficiency of the approach on a real dataset reinforced by statistic results based on simulations.