This paper proposes a new personalized head-related transfer function (HRTF) estimation method based on knowledge distillation (KD). The KD pipeline for generating personalized HRTFs comprises a teacher–student model for transferring well-trained knowledge. The teacher model is the expert that generates personalized HRTFs by representing extensive knowledge using all anthropometric data and ear image. In contrast, the student model is the mimic that attempts to learn from the expert using seven anthropometric data and ear image. The performance of the proposed personalized HRTF estimation approach is evaluated using the Center for Image Processing and Integrated Computing (CIPIC) database. The experiments reveal that the proposed method showed equivalent performance for the root mean squared error and log spectral distance measurements compared to the method using all anthropometric data with the ear images.

Authors: Lee, Geon Woo; Kim, Hong Kook; Chun, Chan Jun
Affiliations: Gwangju Institute of Science and Technology (GIST); Gwangju Institute of Science and Technology (GIST); Chosun University(See document for exact affiliation information.)
Express Paper 116; AES Convention 155; October 2023
Publication Date: October 25, 2023 Import into BibTeX
Subject: Immersive & Spatial Audio
Permalink: https://www.aes.org/e-lib/browse.cfm?elib=22270

Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!

This paper costs $33 for non-members and is free for AES members and E-Library subscribers.

Learn more about the AES E-Library

E-Library Location:   /conv/155/116.pdf

Start a discussion about this Immersive & Spatial Audio!