應用於3D音效之殘響器的合成與最佳化設計

Abstract

殘響音效為3D音效中之一重要部份，早期的殘響音效主要是透過設計一些無限脈衝響應(IIR)濾波器來產生，但使用這種方法在聆聽時會有金屬聲跟一些不自然的聲音產生。而後由於電腦運算速度的提升，現在的殘響器大部分都是用已量測的真實房間響應以有限脈衝響應(FIR)的型式結構對輸入音訊做摺積，使用此種方法不但耗時且耗經費。 本論文主要在發展一套經過參數最佳化設計之人工殘響器，為了兼顧到殘響器的音質和運算量的降低，我們結合兩種方法：以FIR為基礎的虛音源法(image-source method)來模擬空間的初期反射(early reflection)，以及以一系列的IIR濾波器來產生晚期混響(late reverberation)。 此外，我們建立了一套智慧型使用者介面，能讓使用者選擇所要聆聽的房間效果，利用模糊理論來估算各種不同房間對應系統所需要的參數,並可經由視窗上的響應圖形得到此房間的特性。

Reverberator is an important element in three-dimensional audio reproduction. The audio will be sounded more spaciousness when using a well designed reverberator. Improperly designed reverberators were implemented by using many IIR filters and sounds with metallic and ringing or other artifacts. Genetic algorithms (GA) are employed to optimize the parameters of reverberators with high sound quality. In view of the trade-off of complexity and performance, an artificial reverberator is designed to play input music with reverberation in real-time. The reverberator we proposed is combining FIR-based early reflection and IIR-based late reverberation. The complexity of computation for our system is between the FIR-based and IIR-based reverberators. Besides, we use the concept of fuzzy logic to develop an intelligent user interface. The system can determine all the system parameters according to user’s choice of room modes.