A Novel Method for Sound Localization Based on Improved Correlation algorithms & Spatial Gradients Approach

Abstract

Sound localization has been essential in the overall scheme of simulating the placement of an auditory cue. In our work, we have proposed improved correlation algorithms, spatial gradients approach, and inverse algorithm over cross channel algorithms based on time queues comparing them with biological inspired methods. The correlation algorithms were tested with sources in the plane in front of them and were able to determine the position of a source within less than a degree of error. The spatial gradients approach was also tested in the same experiment but in a spherical/3d nvironment and had the same accuracy, moreover but the spread of the estimate for a given source position was much more reliable. A related technique, transfer function based ones were found to have cross channel algorithms with an error of about 2 degrees having an edge over ordinary human hearing with a chance of around 5 degrees of error in the best scenario; whereas our proposed methodology has given the more accurate results with even less than 1 degree of error. Finally, we gave a comparison based on complexity limitations of each of these and accuracy of our practiced work.