Decomposition and Modeling of Head-Related Impulse Responses for Customized Spatial Audio

In order to achieve full binaural sound, audio signals must be synthesized using specialized digital filters characterized by their so-called Head-Related Impulse Responses (HRIRs). They represent the modification to sound due to the listener’s torso, shoulders, head and pinnae, or outer ears. Conventional methods of measuring individual HRIRs require the use of expensive and cumbersome equipment to obtain the HRIRs of every prospective user of the spatial audio system. Alternatively, a manikin may be used to create generic HRIRs. Unfortunately, the access to the necessary equipment is very limited and generic HRIRs do not provide as high spatialization fidelity as individually measured HRIRs. Hence, the creation of a method to create customizable HRIRs without the use of the specialized equipment is extremely desirable. In previous research on this topic, Prony’s modeling method was used to obtain an appropriate set of time delays and a resonant frequency to approximate measured HRIRs. In an effort to expand upon this research the Prony method was substituted for the Steiglitz-McBride iterative approximation method and a noticeable improvement in the approximation was achieved. This paper shows that the improvement achieved is statistically significant for most HRIRs and, therefore, may be advantageous for HRIR decomposition. Key-Words: Head-Related Impulse Responses (HRIR), Prony modeling method, Steiglitz-McBride iterative approximation method, customizable spatial audio.