多聲道音響重現之分析與實現

Abstract

當使用多聲道揚聲器重現立體空間聲場時，交越干擾消除系統扮演了非常重要的角色。然而此技術因為有限的有效區域與大量的計算量，沒有普遍的應用到一般的系統上。在所有的參數中，揚聲器間的夾角是影響分離效果和有效區域最重要的因素。本篇論文有以下幾項重點：首先，針對兩聲道揚聲器規劃了一套完整的研究來探討聆聽角度對於交越干擾消除效果的影響。目的是要找到一個兼具效能與強健性的配置。文中使用兩種有效區域的定義來評估強健性。在數值模擬階段，除了採用理想的點聲源，為了更接近實際的情況，採用頭部相關轉移函數模擬如頭部在高頻的遮蔽效應。針對三個聆聽角度(10、60、120度)作主觀與客觀的實驗。第二，提出以次頻帶濾波為基礎之限頻寬交越干擾消除器降低運算的負擔。由於人類對於低頻的訊號較敏感，我們將頻帶限制到6千赫茲以下。為了驗證此系統，在無響室內實行包含方位測試和音質測試的主觀聆聽實驗。實驗結果使用變異數分析判斷處理前後是否有統計上顯著差異。第三，發展以免感測器之振膜速度估測器為基礎的重低音加強系統，讓整個3D立體空間重現系統更加完整。最後將此系統擴展到多聲道反算濾波，應用在汽車音響上。由於小空間造成的反射、揚聲器與聆聽者沒有位於理想位置和環境噪音等問題，造成車內並不是一個好的聆聽環境。有必要發展一套系統讓聲音在這種環境下能夠正確的傳送。本文針對兩聲道與5.1聲道輸入，提出四種方法。針對兩聲道輸入，提出兩種方式，一為以空間響應合成器為基礎之聲道擴展技術加上反算濾波。另一為聲道擴展技術加上聲道縮減技術和權重與延遲。聲道擴展技術是將兩聲道轉換成5.1聲道，而聲道縮減技術則相反。反算濾波目的是定位出正確的5.1聲道音像。針對5.1聲道輸入也提出兩種方式：一為聲道縮減加上反算濾波；另一為聲道縮減加上權重與延遲。用模擬和實驗來驗證這些演算法且實現於一般的轎車上。主觀聆聽實驗用來比較每個方法並且使用多變量分析結果。

Crosstalk cancellation system (CCS) plays a vital role in spatial sound reproduction using multi-channel loudspeakers. However, this technique is still not of full-blown use in practical applications due to small sweet spot and heavy computation loading. Among the parameters of loudspeaker deployment, span angle is a crucial factor that has a profound impact on the separation performance and sweet spot robustness achievable by the CCS. First, a comprehensive study was conducted to explore the effects of listening angle on crosstalk cancellation in spatial sound reproduction using two-channel stereo systems. The intention is to establish a sustainable configuration of CCS that best reconciles the separation performance and the robustness against lateral head movement. Two kinds of definition of sweet spot are employed for assessment of robustness. In addition to the point source model, HRTF are employed as the plant models in the simulation to emulate more practical localization scenarios such as the high-frequency head shadowing effect. Three span angles including 10 degrees, 60 degrees, and 120 degrees are then compared via objective and subjective experiments. Second, a bandlimited CCS based on subband filtering approach is presented to reduce the computation loading. A pseudo Quadrature Mirror Filter (QMF) bank is employed in the implementation of CCS filters which are bandlimited to 6 kHz, where human’s localization is the most sensitive. To justify the proposed system, subjective listening experiments were undertaken in an anechoic room. The experiments include two parts: the source localization test and the sound quality test. Analysis of Variance (ANOVA) is applied to process the data and assess statistical significance of subjective experiments. Third, a bass enhancement system based on a sensorless cone velocity observer is developed to construct a complete spatial audio reproduction system. At last, this technique is extended to multi-channel inverse filtering for automotive virtual surround audio system. The interior of a car is known as a notorious listening environment due to reflections in a confined space, non-ideal user/loudspeaker positions, and ambient noise, etc. It is then desirable to develop audio systems that are capable of rendering quality spatial sound fields in harsh car environments. Four design approaches are proposed for 2-channel input and 5.1-channel input, respectively. For 2-channel input, a method of reverberation-based upmixing with inverse filtering and another method of up/down mixing with weighting and delay are presented. The upmixing algorithm is used to convert two-channel signals to four-channel signals, while the downmixing algorithm does just the opposite. Inverse filters are employed to position the virtual sound images according to the 5.1 configuration. For 5.1-channel input, a method of downmixing with inverse filtering and another method of downmixing with weighting and delay are presented. These processing algorithms have been practically implemented on a car. Simulations and experiments were conducted for validating the proposed spatial audio systems. Subjective listening tests were also conducted to compare these methods, with the data processed by multivariate analysis of variance (MANOVA).