The Head-Related Transfer Function (HRTF) is a measurement of how sound waves transform as they bounce around and reverberate through the human head, pinna (external ear), and torso. Each person has a unique HRTF due caused by the specific geometries of their upper body and ear. It plays a crucial role in producing the cues in sound localization, which allow the brain to locate sounds in space. HRTFs allow us to recreate a more realistic binaural/virtual surround sound.

​

We can also generalize the Head-Related Transfer Function to work on the "average" person, by analyzing the key components that determine most of sound localization. This method should get us close enough to an approximate HRTF for sound localization, although it may not sound as fully realistic as a personalized one would. For our purposes, this will suffice. Below, we will detail the two main approaches we plan to implement HRTF with.

​

​

Approach 1 (add-on): Computational Model of Vertical Head Shadow and Pinna Effect

This approach adds onto what we have been working so far (horizontal plane ILD and ITD values through empirical data collection) by modifying and perfecting certain features that are missing. Specifically, vertical sound localization cues, especially of the pinnae region, as as the vertical elevation's effect on the head-shadow attenuation. More on this will be updated later.

​

 

Approach 2: Generalized Empirical Model of HRTF from 3D3A Lab Dataset from Princeton University

This approach would "replace" the previous work we have done so far with far more precise data obtained by laboratory experts on state-of-the-art equipment. It entails the use of the 3D3A Lab Dataset [3] which contains HRTF measurements (full head, torso, pinnae measurements) of 38 subjects, including human and non-human physical models. We plan to read in this data and attempt to create a generalized HRTF based on some of the key components of the various HRTFs through some analysis. More on this will be updated later, although it can be said that we plan to use this as a target goal for what our final System should sound like.
 

 

​