Modeling head-related impulse response based on adaptive Fourier decomposition
Chenguang Ke, Qinghua Huang, Lin Zhang, Yong Fang · 2017
Three-dimensional (3D) audio rendering techniques employing head-related impulse responses (HRIRs) are currently used in many fields such as virtual reality (VR) devices and 3D games. The large data of HRIRs makes a high computation cost for the real-time spatial audio applications. Efficient methods for HRIR modeling are necessary for the real-time 3D audio rendering. In this paper, a low computational HRIR model based on adaptive Fourier decomposition (AFD) is presented. In the model, the HRIRs are represented as the sum of mono-components which determined by a small number of the selected atoms in the disc. However, we must sift a substantial of atoms to get the desired atom each decomposition. To further reduce the computation load, we propose an improved method for the AFD model. This method can make the procedure of atoms selection faster by reducing the undesired atoms to some extent. The experimental results show that the improved AFD model requires much less time consuming than the original AFD model.