Recognition of selected helicopter types based on the generated acoustic signal with application of artificial intelligence methods Wiesław. Wszołek 1 , Ryszard.Tadeusiewicz 1 , Andrzej.Chyla 2 1 University of Mining and Metallurgy, Al.Mickiewicza 30, 30-059 KRAKW, Poland 2 SVANTEK, 00-366 Warszawa ul. Foksal 12/14 Abstract In the paper selected parts of conducted research are presented, concerning the application of artificial intelligence methods, in particular the neural networks technique, in the task of helicopter type identification based on the generated acoustic signal. In the design of helicopters produced in recent years (passenger and transport as well as multipurpose ones) certain trends can be noticed, oriented towards increase of the take-off mass, decrease of the lift-off speed, definite increase of the climbing speed and achieved accelerations (steeper take-off profiles). These trends considerably affect the emitted outside noise level (during the initial lift-off phase the engines of the power unit work at their maximum power levels). In this presentation results are discussed, which has been obtained in the laboratories of University of Mining and Metallurgy by application of advanced acoustic signal analysis techniques and by referring to learning neural networks in the task of recognition of selected helicopter types based on their generated noise. 1. Introduction The tasks related to analysis and recognition of acoustic signals characterising selected technical objects (in particular vehicles), which are supposed to be recognised (or classified) using those signals [1], are specific by the fact that for such tasks it is very difficult to find a proper rule for the signal analysis or a proper algorithm for its recognition [2]. It follows from the fact that for identification of technical objects based on the emitted noise it is necessary to refer to rather atypical methods for the signal parameterisation as well as its categorisation and classification. In many cases it is impossible to implement the widely known methods as spectral analysis (alternatively carried out using the recently popular wavelet transform technique) or the typical recognition techniques, like discriminatory analysis, but it is often necessary to create provisional, special methods of analysis and recognition, adjusted to the specific features of the considered problem. Such situation is totally different from the one, which is encountered in processing of typical sounds, found for example in speech acoustics, or in analysis of vibroacoustic signals for the technological diagnosis purposes. It does not mean that the experience gathered during studies of recognition of other signal types is totally useless in the tasks of acoustic identification of vehicles. In particular many sources of information useful for methodology can be found in the well-known solutions to The 2001 International Congress and Exhibition on Noise Control Engineering The Hague, The Netherlands, 2001 August 27-30