Late Reverberation Synthesis using Interleaved Velvet-Noise Sequences
Vesa Välimäki and Karolina Prawda
Companion page for a paper in the IEEE/ACM Transactions on Audio, Speech and Language Processing
Abstract
The paper proposes a novel algorithm for simulating the late part of room reverberation. It is well known that a room impulse response sounds similar to exponentially decaying filtered noise after some time from the beginning. The algorithm proposed here employs several velvet-noise sequences in parallel and combines them so that their non-zero samples never occur at the same time. Each velvet-noise sequence is driven by the same input signal but is filtered with its own feedback filter which has the same delay-line length as the velvet-noise sequence. The resulting response is sparse and consists of filtered noise which decays approximately exponentially with a given frequency-dependent reverberation time profile. It is shown by a formal listening test that four interleaved branches are sufficient for producing a smooth high-quality response. The outputs of the branches can be combined in different permutations to produce decorrelated output signals for multichannel reproduction. The proposed method is compared with a state-of-the-art delay-based reverberation method and advantages are pointed out. The computational load of the method is 60% smaller than that of a comparable existing method, the feedback delay network. The proposed method is well suited to the synthesis of diffuse late reverberation in audio and music production.
Examples of velvet-noise stimuli used in the listening test
Reference - infinite velvet noise
Anchor - one sequence, length: 80 x 89 samples
Two sequences, lengths: 80 x 101, 80 x 103 samples
Three sequences, lengths: 80 x 107, 80 x 109, 80 x 113 samples
Four sequences, lengths: 80 x 83, 80 x 89, 80 x 97, 80 x 101 samples
Five sequences, lengths: 80 x 103, 80 x 107, 80 x 109, 80 x 113, 80 x 127 samples
Examples of IRs with original (measured) and synthesized (with IVN) late reverberation
"World's longest echo" - measured IR, sample taken from [1]
"World's longest echo" - late part synthesized using IVN reverberator (first 100 ms taken from the original sample)
Concert hall in Pori, Finland - measured IR
Concert hall in Pori, Finland - late part synthesized using IVN reverberator
Lecture room - measured IR, sample taken from [2]
Lecture room - late part synthesized using IVN reverberator
Segmentation- examples
No segmentation
3 segments
Exponantial decay
Examples of sounds reverberated using the proposed method (IVN)
Female voice - anechoic sample from [3]
Guitar music - anechoic sample from [4]
Handclap - anechoic sample from [5]
[1] T. Cox. World's 'longest-echo' fifth impulse, [Online]. Available:http://freesound.org/people/acs272/sounds/214221/
[2] M. Jeub, M. Schäfer, and P. Vary, A binaural Room Impulse Response database for the evaluation of dereverberation algorithms, inProc. Int. Conf. Digital SignalProcess. (DSP), Santorini, Greece, Jul. 2009
[3] P. Kabal, TSP speech database, Department of Electrical and Computer Engineering, McGill University, Montreal, Canada, 2018
[4] B. Bernschütz, Anechoic recordings , Cologne University of Applied Sciences, Institute of Communication Systems, Cologne, Germany, 2013
[5] A. Woldhek, Handclaps, [Online], Avialable: https://freesound.org/people/Anton/sounds/345/