ACOUSTIC SOURCE RECOGNITION BASED ON LARGE DISTANCE OBSERVATIONS, USING PHYSICAL PROPAGATION MODELS AND OPTIMIZATION TECHNIQUES

摘要

In this paper a non-linear optimisation technique is proposed for acoustic source recognition, based on the use of a database of known acoustic sources and a detailed physical model of the sea medium. The propagation of acoustic waves inside the sea is analysed by employing a horizontally stratified model for the sea medium with an arbitrary number of layers, characterised by individual sound speed and density. Through this physical model the acoustic field distribution, originating from an elementary source at an arbitrary point inside the sea medium is calculated, by implementing a Green's function approach. The proposed optimisation algorithm is based on the minimisation of the quadratic error function between estimated source signal values and signal values corresponding to known acoustical sources, for the frequency content of the measured signal, by employing the Downhill-Simplex minimisation technique.