1 Introduction

A number of various methods of processing the RICH (Ring Imaging CHerenkov) detector data for particle identification (PID) are already known (see, e.g., refs. [1, 2, 3]). Some of these methods are based on the maximum likelihood approach [1], or on comparing probability distributions [2] of Cherenkov photons for different hypotheses and, therefore, assume an extended preliminary preprocessing of the bulky array of many thousand pads from which the raw RICH data are consisting. Other methods, like ref. [3], use the pad information, but only to count the number of pads in fiducial areas calculated for alternative particles. All PID methods demand the knowledge of Cherenkov ring centers and corresponding particle momenta with a high level of accuracy.
This paper is motivated by the following considerations:
(i) There is a need for a substantial speeding up of processing of RICH measurements.
(ii) This can be achieved, in particular, by eliminating such a time-consuming stage as locating photon hits in a sea of pads.
(iii) Successful application of robust statistical methods in our previous work [4] inspired us to apply those methods to the raw RICH data.
It allows to compute Cherenkov ring parameters with a better accuracy and improve the particle identification (PID) procedure by choosing the most likely radius corresponding to a pad sample surrounding a Cherenkov ring center. Numerical tests show considerable improvement in the accuracy over traditional algorithms. We do not discuss here other interesting but more time consuming methods of RICH raw data processing such as the MCMC (Markov chain Monte-Carlo) approach studied in ref. [5].