汽車音訊訊號處理系統之設計與實現

Abstract

本論文研究重點在於應用在汽車上之立體空間音效處理系統的設計與實現。共提出十個演算法以克服不同乘客及輸入訊號數目的播放模式。其中，兩個分別針對雙聲道立體聲及5.1聲道輸入所設計的聲道擴展/縮減演算法，目的在於平衡車內前後側揚聲器之聲音輸出。而另外八個基於反算濾波的演算法則可分成兩大類，第一類方法根據在汽車內所量測之頭部相關轉移函數(HRTF)設計，此類方法考慮到因人體頭部、耳朵及軀幹所產生繞射及遮蔽效應。而第二類方法則稱為點接收器模型，用一個位於乘客頭部中心的點接收器來模擬聲音的傳播。所提出之演算法在不同的聆聽模式下，透過一系列包含車內定位實驗的主、客觀實驗加以比較。實驗結果透過多變異量分析(MANOVA)及費雪最小顯著差異法(Fisher’s LSD)的事後檢定(post hoc test)檢視是否有統計上的顯著差異。實驗結果指出，在單一乘客的情況下，反算濾波法表現較佳，而在多位乘客的情況下，聲道擴展/縮減法的表現優於其他處理方法。

Design and implementation strategies of spatial sound processing are investigated in this paper for automotive scenarios. Ten design algorithms are implemented for various rendering modes with different number of passengers and input channels. Two up/downmixing algorithms aimed at balancing the front and rear reproduction are developed for the stereo two-channel input and the 5.1-channel input, respectively. Eight algorithms based on inverse filtering are implemented in two approaches. The first approach is based on binaural HRTFs (Head-Related Transfer Functions) measured in the car interior, which accounts for the diffraction and shadowing effect due to the head, ears and torso. In the second approach termed the point-receiver model, sound rendering is targeted at a point receiver positioned at the head center of the passenger. The proposed processing algorithms were compared via series of objective and subjective experiments under various listening conditions. In particular, a localization test was undertaken in the car interior to compare the proposed algorithms. Test data were processed by the multivariate analysis of variance (MANOVA) and the least significant difference method (Fisher’s LSD) as a post hoc test to justify the statistical significance. The results indicate that inverse filtering methods are preferred for the single passenger mode. For the multi-passenger mode, however, up/downmixing algorithms have attained better performance than the other processing techniques.