Challenges in Aircraft Noise Research

Abstract

Aircraft external noise is due to several parameters which are classified in: engine noise and aerodynamic noise. Each parameter plays a key role for the robust and distinguishable acoustic feature vectors useful for the aircraft classification and identification. There are various techniques to classify noise source by analyzing the sound emission from an aircraft for generating the acoustic signature recognition. The task of the signature classification process is the feature extraction: the assignment of noise histories to classes which have strictly defined properties. This Note highlights a technique in acoustic identification of airplane through different flight conditions. Aircraft identification phase is further carried out by using the trained net to classify any identified aircraft noise signal, not included in the neural network training phase. The results show that this procedure provides good identification of unknown noise sources when proper wavelet decomposition (type and number of levels), a good design of the neural network (number of layers, neurons, activation functions) and a proper set of training feature vectors are employed. Keywords: acoustic vectors, orthogonal wavelets, aircraft noise.