Estimation of the sound source direction with artificial neural networks using experimental data on the limited area

In this study, it is aimed to find the direction of the sound source, according to certain distance which were determined previously at the normal conditions and taken from one square meter area by using the sound level values. A source of sound, at a constant amplitude and frequency, for a suitable environment arranged, and by recording the data, attempted to examine the relationship in the direction determination and "position-sound level relation" was established for releated area. The physical reality of the system is based on the decreasing the sound power level according to distance that the sound level meter from the source. As the angle of the sound source to the sound level meter changes, the relationship between the change in the received system data and the direction of the source was investigated. Five decibel meters were used in the system and some datas, such as distance, sound level, ambient noise, and sound source angle were taken. Using these data, linearity was examined at the various angles for the four defined regions in the system. The obtained linearity parameters and the values which were taken for another position in the environment, were investigated that whether the source angle was correct or not. The obtained datas were subjected to artificial neural network in MATLAB and the region and angle of the sound source were estimated. Moreover, the direction datas were obtained from the linearity analysis and the direction datas were taken from the artificial neural network. They were compared. For estimation, an artificial neural network structure which has 8 neurons were used in the feed-forward back propagation in its secret layer. Consequenlty, the results were expressed with the graphs.