Datasets:

davidscripka

/

MIT_environmental_impulse_responses

like 3

Hello David!

I've been working on a project similar to openWakeWord, and wanted to use this dataset for augmentation in the same manner your project did. However the unknown license bothered me, so I reached out to the original authors of the dataset and was able to get in contact with both James Traer and Josh McDermott.

They were able to clarify the data is released under CC-BY and provided the link to CC-BY-4.0, so that is the license I set on my copy of the dataset (in original 32khz) on huggingface.. They asked to cite their PNAS paper, so I think if you just include this citation you can change the license of this repository, too.

@article

	{
doi:10.1073/pnas.1612524113,
author = {James Traer and Josh H. McDermott},
title = {Statistics of natural reverberation enable perceptual separation of sound and space},
journal = {Proceedings of the National Academy of Sciences},
volume = {113},
number = {48},
pages = {E7856-E7865},
year = {2016},
doi = {10.1073/pnas.1612524113},
URL = {https://www.pnas.org/doi/abs/10.1073/pnas.1612524113},
eprint = {https://www.pnas.org/doi/pdf/10.1073/pnas.1612524113},
abstract = {Sounds produced in the world reflect off surrounding surfaces on their way to our ears. Known as reverberation, these reflections distort sound but provide information about the world around us. We asked whether reverberation exhibits statistical regularities that listeners use to separate its effects from those of a sound’s source. We conducted a large-scale statistical analysis of real-world acoustics, revealing strong regularities of reverberation in natural scenes. We found that human listeners can estimate the contributions of the source and the environment from reverberant sound, but that they depend critically on whether environmental acoustics conform to the observed statistical regularities. The results suggest a separation process constrained by knowledge of environmental acoustics that is internalized over development or evolution. In everyday listening, sound reaches our ears directly from a source as well as indirectly via reflections known as reverberation. Reverberation profoundly distorts the sound from a source, yet humans can both identify sound sources and distinguish environments from the resulting sound, via mechanisms that remain unclear. The core computational challenge is that the acoustic signatures of the source and environment are combined in a single signal received by the ear. Here we ask whether our recognition of sound sources and spaces reflects an ability to separate their effects and whether any such separation is enabled by statistical regularities of real-world reverberation. To first determine whether such statistical regularities exist, we measured impulse responses (IRs) of 271 spaces sampled from the distribution encountered by humans during daily life. The sampled spaces were diverse, but their IRs were tightly constrained, exhibiting exponential decay at frequency-dependent rates: Mid frequencies reverberated longest whereas higher and lower frequencies decayed more rapidly, presumably due to absorptive properties of materials and air. To test whether humans leverage these regularities, we manipulated IR decay characteristics in simulated reverberant audio. Listeners could discriminate sound sources and environments from these signals, but their abilities degraded when reverberation characteristics deviated from those of real-world environments. Subjectively, atypical IRs were mistaken for sound sources. The results suggest the brain separates sound into contributions from the source and the environment, constrained by a prior on natural reverberation. This separation process may contribute to robust recognition while providing information about spaces around us.}}

P.S. I will be open-sourcing my work of course, and will cite you when it is released!