Mengfan ZhangFollow

Abstract

When sound reaches human ears, sound waves do not arrive at each ear at the same time. Rather, depending on the direction of the sound source, sound waves take different paths to each ear. These paths will likely have different lengths, so the sound waves will reach the closest ear first and the farthest ear shortly after. The exact paths sound waves take to reach a person’s ears are dictated by the person’s head and ear anatomy. To reproduce sounds accurately for immersive experiences, these anatomical differences, which are scientifically modeled by Head-Related Transfer Functions (HRTFs), must be accounted for.

Typically, personalized HRTFs are developed using anthropometric measurements, 3D scanning approaches, partial HRTF measurements, or tuning approaches, etc. However, most of these methods are not practical for production, as they require significant user effort or specialized equipment. Although 3D scanning and anthropometric techniques are more practical, they still require individual data collection. In theory, a user’s unique HRTF could be compared to a database of existing HRTFs, which might provide the user with a personalized sonic experience. However, obtaining a user's unique HRTF would require extensive and impractical measurements, which can be avoided. Through experimentation, it was found that a single universal HRTF derived from a data-driven clustering process minimizes user effort while delivering an immersive user experience. This offers a practical and scalable solution for delivering high-fidelity immersive audio to a mass market. The core innovation involves eliminating collecting any types of individual features, such as anthropometric parameters, 3D scans, or even performing a whole set of HRTF measurements.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS