|
We have investigated sonic cues for sound localization. Our goal was to investigate signal processing methods for synthesizing spatialized sound.
Two of the primary cues we use for localization are Interaural Time Differences (ITD) and Interaural Level Differences (ILD) created between the two ears by sounds from a particular point in space. However, these are ambiguous cues and they do not lend any information about the elevation of a sound source. However the torso, head, and outer ear do perform elevation-dependent spectral filtering of sounds before they reach the inner ear. This affect is commonly called a Head Related Transfer Function (HRTF). This aids in resolving the ambiguity from the ITDs and ILDs alone and helps us localize sounds in free space.
We have investigated several different approaches to modeling these HRTFs generating spatialized sound for presentation over headphones. We have implemented two methods in MATLAB.
One method is to use measured HRTFs from a mannequin to filter sound presented to the listener over headphones.
The other method is to implement a signal processing model which tries to mimic the physical effects of the sounds interacting with the body.
We tested the effectiveness of these two methods by synthesizing spatial audio for use in listening tests with human subjects.
This report presents our test results and our findings concerning how well the subjects were able to localize the sounds samples generated with the two methods.
|
