以田口實驗法及基因演算法對強健性陣列喇叭交越干擾消除系統作最佳化設計

Abstract

本文提出以陣列喇叭為基礎之交越干擾消除系統(CCS)最佳化設計用於個人3Ｄ音訊系統。本系統使用到的技術包含頭部轉移函數(HRTF)和交越干擾消除系統。使用Tikhonov正規化(Regularization)逆濾波的方法設計多通道數位濾波器來消除人耳和喇叭間的交越干擾。以由陣列喇叭到一組設計控制點的逆空間運算子為基礎將聲束聚焦之技術以改善交越干擾消除系統之強健性。另一種稱為線性限制最小變異聲束聚焦法(LCMV)一同於本文中介紹。為了實現具備良好性能之強健性交越干擾消除系統，本文以田口實驗計劃法及基因演算法(GA)對於此系統作最佳化設計。文中以數值模擬之聲壓分佈輪廓圖來比較不同系統結構之系統性能和系統強健性。文中之實驗結果分為物理性測試及人體聆聽測試兩部份。由數值模擬和實驗的結果顯示，本文提出的系統能有效地改善系統對於頭部位移強健性的目的。

A optimal design technique of the cross-talk cancellation system for personal three-dimensional (3-D) audio system based on a speaker array is presented in this paper. The proposed system mainly consists of the head related transfer function (HRTF) and the cross-talk cancellation system (CCS). A Tikhonov regularization inversing technique is employed in designing multi-channel digital filters to cancel the cross-talks between the speakers and the listener’s ears. A focusing technique based on the inversion of the propagation operator relating an array of transducers to a set of control points is employed to improve the robustness of cross-talk cancellation system. Another focusing technique called linearly constrained minimum variance (LCMV) beamforming method is introduced in this paper. The Taguchi method and the genetic algorithm (GA) are used to optimize this system in order to achieve a robust CCS with good performance. Numerical simulations of contour plots are presented in comparing performance and robustness in different configurations. The experimental results are divided into two parts, physical tests and subjective listening tests. Numerical simulations and experimental investigations indicated that the proposed system is effective in improving robustness to head misalignment.