> Any links towards that? Because, frankly, I've yet to see a *single* > audio-DSP application which crucially relies on deep learning.
There are a bunch! I personally work for Unity developing an audio-DSP application which crucially relies on deep learning! It's part of the OTO team <https://venturebeat.com/games/unity-acquires-ai-chat-analysis-platform-oto-launches-toxicity-in-gaming-report/>. The product is in alpha state, but we expect to have an open beta by Q3, I can share more details with you then. In any case, there are a ton of DL audio-DSP applications out there in the wild (youtube's automatic subtitles is the first one that comes to mind). Maybe you are thinking of DSP exclusively as modifying an audio to sound in a certain way? I gave a talk for AES in 2019 about ML for audio that includes some commercial examples like Izotope's Ozone and Neutron plugins. Those are even in real-time. I included the slides in case you want to skim through some names of things I considered back then to be ML for audio DSP. There are even more products out there by now. Sorry that the slides are in (broken) german. Before I go, be sure to check out magenta's ddsp project <https://magenta.tensorflow.org/ddsp>. Tons of cool things mixing classic DSP with ML. Best, Andrés Marafioti AES_Talk.pdf <https://drive.google.com/file/d/1LFpGZT-oIWmWQ4hyjJrzaLou8oJgJidR/view?usp=drive_web> On Fri, Jan 20, 2023 at 7:49 PM Andy Farnell <padawa...@obiwannabe.co.uk> wrote: > On Thu, Jan 19, 2023 at 05:54:40AM +0200, Sampo Syreeni wrote: > > > Any links towards that? Because, frankly, I've yet to see a *single* > > audio-DSP application which crucially relies on deep learning. > > https://www.youtube.com/watch?v=duWX4RAFWIo > > > machine and deep learning could be utililized for spotting point sources > > there, even their distance, and quite certainly recurrent neural > networks in > > time would be well suited towards such a task. In particular if you do > > Indeed. We have very good understanding of what to expect in acoustic > signals and lots of opprtunities to extract environmental features in > real-time. > > Automatic auditory scene analysis is a natural evolution from > basic "identifcation" machine listening. > > Although even that reached some impressive goals the last few years > as you can see from Dan Stowell's work on identifying birds [1], > and the NN recogniser they built at Cornell [2]. > > [1] machine-listening.eecs.qmul.ac.uk/bird-audio-detection-challenge/ > > [2] birdnet.cornell.edu > > > > >