Automated Design of Digital Filters for 3-D Sound Localization in Embedded Applications

In this paper, an automated design method of digital filters for 3-D sound localization is proposed, which approximates head-related transfer function (HRTF). Target digital filter organization is dedicated for embedded applications so that ROM capacity and computational load are reduced with only a slight degradation of 3-D sound effect. The proposed design method consists of two steps. In the first step, the least-squares method is adopted to acquire initial design result quickly, while the second step utilizes the gradient search algorithm for precise optimization. In an objective evaluation, the proposed two-step method is proved to be effective for improving design results. Results of subjective listening tests also show that the proposed method is comparable to manual design in the approximation of reference HRTF. In these subjective tests, our automated design achieved better results per ROM capacity than simple FIR filter did