Skip to Main Content
The present study is concerned with the use of neural networks as storage devices for the deployment of head-related transfer function (HRTF) in three-dimensional (3D) audio applications. There are two main advantages for this proposition: reduced storage cost and direct interpolation. The network structure, training procedure, and experimental results are provided to illustrate the effectiveness of our approach. It is shown that by using 35% of the original number of filter parameters, a relatively low average spectral distortion of 1.62 dB can be obtained.