ambisonictoolkit / atk-kernels Goto Github PK
View Code? Open in Web Editor NEWA set of impulse response files used for convolution-based encoders and decoders
License: Other
A set of impulse response files used for convolution-based encoders and decoders
License: Other
Current code lists the NFC filtering: r = 1.95
Need to check if this is correct, or not!
Update README to include more details on the various kernels.
Marco Del Piano reports that the spherical 192000 files don’t work, the files are silent. Investigating, it seems like the new 176.4 kHz kernels for the spherical decoder has the same problem too.
The new kernels for encoders at 176.4 kHz looks good, this only seems to be an issue with the binaural decoders.
The binaural decoders generated from measured HRTFs (Listen, CIPIC) currently encode without regard to the individual subject's "look direction". That is, it is assumed each subject is looking at [0, 0] during measurement. In practice, the true look direction varies somewhat with each subject.
Measuring the (band-limited, "pinked") normalized intensity vector will give a reasonable measure of the true "look" of each subject.
Correcting for "look" will bring greater consistency to measured binaural decoders.
@henk-spook questioned whether the binaural kernels can be used with Google's YouTube / Jump Inspector
https://storage.googleapis.com/jump-inspector/workflow/thrive.zip
Would be useful to find out what, if any, changes would be necessary to do so.
They are using AmbiX1 encoding (ACN-SN3D), and I recall seeing something that looked like a virtual speaker cube decoding strategy.
There are a few strategies I can think of:
The first of these would require generating new sets from the current ATK sets. Results would maintain all the desired qualities of ATK criteria.
The second is more efficient in terms of processing, and the current Jump Inspector developers may not be aware of this technique.
The current encoding and decoding kernels are linear phase, meaning they sound great ;), but as such there is a delay in the signal, which may not be ideal in some live situations. This issue is to make note of the discussion we've had about minimum phase or "mixed phase" filters which would minimize the delay. This implies a good amount of research and experimentation!
Just getting the thread going...
The release for kernels comes in a folder with the release number appended and "atk-" prepended. E.g.,
atk-kernels-1.1.0
atk-sc3 expects the kernel folder to just be named
kernels
As a result, dropping the unzipped and un-renamed folder into Platform.userExtensionDir
does not result in a working atk-sc3 installation. Here's my reply on the sc-users list to a new user attempting an installation.
The question, then, is what is the best way to resolve? Do we just give atk-sc3 users an instruction to rename the folder as I've done here (follow to my reply at the bottom)?
We'd like atk-sc3 and atk-reaper to both use the same assets to keep everything synchronized.
Related atk-sounds issue: Installation folder name conflict #3
Add externalisation to binaural decoders.
Question: Should the current HRTFs be replaced with the externalised versions? Or, should these be added as a new set.
README.md includes a section discussing Git LFS:
Working with this repository
The sound files in this repository are versioned using Git Large File Storage (LFS). You will need to install LFS in order to clone and work on this repository. The repository is set up to use LFS with all WAV files.
Is this up to date? If not, I'd expect the Working with this repository section should be modified or removed.
Responding to this question at atk-reaper:
Two commercial products include support for import of personalized HRTFs from SOFA/AES69 files:
BlueRipple describes the support as "experimental".
For ATK support, we'd either want to release a decoder design routine as a utility, or set up an online SOFA-file upload / ATK-decoder-kernel download mechanism...
This is one of the sample rates offered by most hi end audio cards as well as Reaper, and although it is probably less used than 196 kHz, one might imagine someone using it for oversampled processing targeting CD production (the sample rate is 4 x 44.1 kHz).
This issue relates to #1 .
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.