1. Introduction
Digital audio effects have afforded new levels of creativity for recorded and performed sound. Audio engineers have experimented with both analog and digital audio effects to enhance music while maintaining high fidelity. Artificial reverberation introduces a spatial dimension to a piece of recorded sound. Implementation in the digital domain gives us the flexibility of portable, repeatable and dependable algorithms on which to base reverberation effects.
The frequency coloration and delay properties of a particular reverberation effect can simulate general listening environments. For example, high frequencies give the impression of hard walls while their absence gives the impression of a soft-carpeted room.[9] Long reverberation times provide the feeling of a large hall, while short reverberation times give the impression of smaller rooms.
Artificial reverberation can also be used to model a specific acoustic environment in which to affect a dry unaltered signal. In the digital domain, processor speed is the limiting factor, thus the balance of fidelity, processing time and application becomes the crux of appropriate reverberation. If the application is a real-time system, fidelity has to be sacrificed so that the processing time can be maintained according to the application demands. However, if the application is strictly post-processing in nature, fidelity can be extremely high with processor-intensive calculations.
This thesis will try to destroy this dichotomy and investigate a new method by which a real-time system can achieve high fidelity with modest processor demands. This will be achieved through the merging of the two commonly used techniques: DSP filtering and acoustic impulse response convolution. While any acoustic environment can be modeled in the manner, the testing of this thesis was limited to concert halls, auditoriums, churches, etc. These acoustic spaces have predictable reverberation characteristics and high quality measured impulse responses are easily accessible. In this paper, the general term "room" will be used to represent these acoustic spaces.
Chapter 2 will introduce the major design tools, building blocks, cornerstone implementations and quality measurement systems that are common in current reverberation research. In order to realize a perceptually optimized artificial reverberation architecture, we must first be fluent in the ideology of current reverberation techniques. The utilities of convolution, comb filters and all-pass filters have been a staple ingredient in the pursuit of high quality reverberation. A hybrid architecture that marries the two typically independent reverberation techniques is described in Chapter 3. Methods of impulse response manipulation, frequency equalization and diffuse reverberation are also introduced. The specific design criterion is explored in Chapter 4. Practicality, necessity and fidelity are weighed against each other for finalizing the new algorithm. The new hybrid architecture is implemented and tested in Chapter 5 with a variety of listening material and using a PC based ABX software package. Finally, Chapter 6 presents applications, shortcomings and future development of this system.