Partitioned convolution algorithms for real-time auralization
Author | : Frank Wefers |
Publisher | : Logos Verlag Berlin GmbH |
Total Pages | : 278 |
Release | : 2015-05-11 |
ISBN-10 | : 9783832539436 |
ISBN-13 | : 3832539433 |
Rating | : 4/5 (36 Downloads) |
Download or read book Partitioned convolution algorithms for real-time auralization written by Frank Wefers and published by Logos Verlag Berlin GmbH. This book was released on 2015-05-11 with total page 278 pages. Available in PDF, EPUB and Kindle. Book excerpt: This work discusses methods for efficient audio processing with finite impulse response (FIR) filters. Such filters are widely used for high-quality acoustic signal processing, e.g. for headphone or loudspeaker equalization, in binaural synthesis, in spatial sound reproduction techniques and for the auralization of reverberant environments. This work focuses on real-time applications, where the audio processing is subject to minimal delays (latencies). Different fast convolution concepts (transform-based, interpolation-based and number-theoretic), which are used to implement FIR filters efficiently, are examined regarding their applicability in real-time. These fast, elementary techniques can be further improved by the concept of partitioned convolution. This work introduces a classification and a general framework for partitioned convolution algorithms and analyzes the algorithmic classes which are relevant for real-time filtering: Elementary concepts which do not partition the filter impulse response (e.g. regular Overlap-Add and Overlap-Save convolution) and advanced techniques, which partition filters uniformly and non-uniformly. The algorithms are thereby regarded in their analytic complexity, their performance on target hardware, the optimal choice of parameters, assemblies of multiple filters, multi-channel processing and the exchange of filter impulse responses without audible artifacts. Suitable convolution techniques are identified for different types of audio applications, ranging from resource-aware auralizations on mobile devices to extensive room acoustics audio rendering using dedicated multi-processor systems.