Individualizing spatial audio is of crucial importance for high-quality virtual and augmented reality audio. In this paper we propose a method for individualizing spatial audio by combining the recently proposed ear shape modeling technique with computer vision and machine learning. We use a convolutional neural network to obtain estimates of the ear shape model parameters from stereo photographs of the user ear. The individualized ear shape and its associated individualized head-related transfer function (HRTF) can be calculated from the obtained parameters based on the ear shape model and numerical acoustic simulations. Preliminary experiments, evaluating the shapes of the estimated individual ears, proved the effect of individualization.
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.