Google patents multi-effects processor

[Blackaddr]

Came across this:
https://patentimages.storage.googleapis.com/c5/a4/33/4c4318a762e875/US20180357992A1.pdf

If you're not used to reading patents, it can seem very long-winded to say very little. The pictures through out will help you get the gist of it. They have patented taking an input audio signal, sending it to a bunch of signal analyzers, then applying a bunch of effects (delay, reverb, etc.) before it goes on to a speaker.

Now as you're going through this large document, you're probably looking for the meat. The actual algorithms for those fancy signal analyzers and effects that Google dreamed up. That's probably the real new, novel invention.

But no, the patent doesn't cover any algorithms whatsoever. All it does is patent the "idea" of a realtime system with signal analyzers and effects. That's it.

How the hell does something with no novel ideas or content at all get awarded a patent (*edit* this is a published application it hasn't been awarded *yet*) for something so obviously common in the industry? Drives me nuts because this is actually an extremely common occurrence in US Patent cases.

[EBK]

The linked doc is not an issued patent.  It is a published patent application.

[Blackaddr]

True, but this isn't some random company. I'm pretty sure Google knows what they will and will not be able to get away with on a patent.

[EBK]

If I told you that I know from my professional experience that large companies have zero interest in getting away with something in a patent, would you believe me?

[octfrank]

I am not a lawyer but my reading of claim 1 makes it appear very weak. It seem to me things like expanders, compressors, automatic gain controls, auto-wah, etc. would all be prior art and invalidate claim 1. If you invalidate claim 1 you also invalidate claims 2 - 11 since they are all dependent claims. Claim 12 seems to be the same as 1 except you connect an instrument and a speaker, seems same prior art would invalidate it.

I do not see Google listed as the inventor or applicant so what causes you to believe Google/Alphabet is behind it? Most big companies try to do as many independent claims and as few dependent claims as possible so if one claim is found invalid it does not affect other claims.

[EBK]

If you invalidate claim 1 you also invalidate claims 2 - 11 since they are all dependent claims.

Not so.  Not in the US, anyway.

Dependent claims are narrower in scope and may be novel and non-obvious even when an independent claim is not. That's actually the main point of having dependent claims.

By the way, the USPTO issued an Office action rejecting all of the claims on August 3rd.  You can find it here: https://portal.uspto.gov/pair/PublicPair

[Blackaddr]

I do not see Google listed as the inventor or applicant so what causes you to believe Google/Alphabet is behind it?

It's stored on Google's patent server. Patents are filed by the inventor, not their employer so you wouldn't expect to see Google in the application. Beyond that, you're 100%, I'm speculating the patent is on behalf of Google, they might store un-affiliated patents on their patent storage as well.

Regarding "getting away" with something in an application, perhaps I should clarify what I meant. Large companies have legal teams that know whether a patent application is likely to succeed or not. Entrepeneurs not familiar with the system are far more likely to file an application they think has merit and have it rejected, which incurs quite a bit of cost to a small business entrepeneur.

If a large company files a patent they know will be rejected, it's usually to take advantage of the FIRST-TO-FILE system like in the US. Somebody else files a more detailed patent application later and the former can claim it's derivative from their earlier application. Countries like Canada have a FIRST-TO-INVENT system where you must prove you invented it first, you can literally take away someone's patent if you can prove you were already doing it but kept it as a trade secret.

That said, I've seen many similar types of patents in the video industry where I work where lego-patents cause a lot of problems. By lego-patent, I mean someone obtains a patent for the concept of putting blocks together like lego, in reality, there was nothing novel about it but they get awarded anyway and that FIRST-TO-FILE is so very valuable in the US.

[Ben N]

What's this mean: "Application ready for PDX access by participating foreign offices"?

[EBK]

What's this mean: "Application ready for PDX access by participating foreign offices"?

I think it means they've converted it into the agreed-upon format for viewing through the Global Dossier initiative.
See https://globaldossier.uspto.gov/#/

[Digital Larry]

Well, this thing says it analyzes attributes of the input signal to decide what to do.  Most processors I'm aware of do the same thing to all signals coming in.  Now OK, if we're talking about a compressor, you could say:

a) The processor reduces the gain by 1 dB for every 2 dB in signal increase above a certain threshold (fixed processing)
b) The processor analyzes the input and if it's over the threshold, it reduces the gain 1 dB for every 2 dB increase of the input level (analytical).

I think what they might be getting at which is more novel is something like "if the user plays Bb staccato at a rate exceeding eighth notes at 115 BPM, then all C notes will be pitch shifted by 3 semitones and subjected to 150 msec delay".  Let me tell you, blues in E never sounded so... whatever.

[audioartillery]

I'm speculating the patent is on behalf of Google

Most likely the inventor is this Denis Cronin who does not work for Google.  He appears to be a software engineer with DSP experience.  If I had to guess I'd say he has some novel guitar effect which is very adaptive and he's patenting it before bringing it to market.

Wish we had more people on this forum, it always seems a little quiet -- maybe we can get him to join us.

[potul]

I guess this is what he's trying to patent, basically a system that detects certain aspects of the signal and decides what effects or parameters to use. Not something really new as a concept, as we've been using things like envelope detectors for a long time, but there are some interesting innovations in it. I like the idea of detecting if you are playing mono or polyphonic and adapt the effects to it. So when playing chords you have a set of effects, if you switch to a melodic or solo playing you get something different.

We should ping the guy and ask him to join the forum (maybe he is already in)


http://morfx.net/

[audioartillery]

Cool, good job stalking him down on youtube.  Next I figure we find his address and see if we can watch him working on effects from the street at night 

I sent him an email, maybe he'll come chat with us.

I'm curious what he's using to adapt the ARM eval board to guitar -- is there some codec under there or is he just using onboard DAC/ADC?

[confundido]

Ok, here I am. Thanks for the invite.  I just used the onboard 12 bit ADC/DAC. It's kinda crummy but it was good enough to play with some stuff.

I swore I'd looked at you guys before to see if you had a platform.  When I'm not at work I'll spend a bit more time perusing...

Dennis aka confundido (Spanish and Portuguese for "confused")

[Blackaddr]

Ok, here I am. Thanks for the invite.  I just used the onboard 12 bit ADC/DAC. It's kinda crummy but it was good enough to play with some stuff.

I swore I'd looked at you guys before to see if you had a platform.  When I'm not at work I'll spend a bit more time perusing...

Dennis aka confundido (Spanish and Portuguese for "confused")

Welcome! Sorry for accusing you of being a Googler! I jumped to the wrong conclusion since I found it on their patent server. This thread started regarding a discussion over your patent application. The application was published in Dec and another forum member indicated it had been rejected, but I had trouble checking the status myself. Can you tell us a bit about your patent application process and it's current status. I'm sure some readers here would be interested in what that process looks like.

[confundido]

I've abandoned the patent.  The patent examiner turned up this previously granted patent which seemed to box me out:

https://patents.google.com/patent/US8748724B1

The lawyers wanted to go ahead and patent just the adaptive delay concept but that seemed pretty minor.

I also talked to the Fractal guys and they said they'd been doing input adaptation for years altho I've gone thru a Fractal pedal board pretty thoroughly and didn't see anything like what I was envisioning.

Maybe there's still something patentable, I'm frankly amazed nobody (that I'm aware of) has product, but in the grand scheme of busy stuff I need to do I decided to cut my losses and set this one aside (after spending piles of money on patent lawyers btw)

[Rob Strand]

How the hell does something with no novel ideas or content at all get awarded a patent (*edit* this is a published application it hasn't been awarded *yet*) for something so obviously common in the industry? Drives me nuts because this is actually an extremely common occurrence in US Patent cases.

I'm over it as well.

The way I've seen it work:
- Make a whole lot of claims which try to own a much space as possible (even stuff that's already common knowledge or possibly already patented.)
- Submit patent
- *Some* items are rejected but not all the ones that should be.   The case reviewer(s) is human and he only has so much time he can spend reviewing it.
- Argue about claims you want to keep but were rejected.  Often a different spin on a point gets it through even though that spin doesn't end-up getting mentioned in the final document.
- Update document
- Submit again
- If it passes, pay the money and run.
- After that it's up to the courts.

EDIT:
FWIW, The meat is in the claims.   For that patent there are claims beyond a generic DSP box.

[confundido]

Yeah, that guy's patent is super broad in some ways (talking about automatic lighting changes) but then he's kind of fixated on FFT freq analysis. but i didn't think i had a great chance of dodging that, maybe i should have.  but i have the same kind of dim view of patents.

and i hadn't really had any decent nibbles on the idea.

it's way fun to play thru tho. seemed like something barroom guitar gods would enjoy.

[Blackaddr]

FWIW, The meat is in the claims.   For that patent there are claims beyond a generic DSP box.

The meat I was thinking of is actual algorithms, rather than detailed descriptions of things you could do. Novel algorithms are very patentable. A perfect example is the work of Julius O. Smith III. Anybody who does sound algorithm development should know who this guy is. He invented the concept of digital waveguides, patented it, then licensed it to companies like Roland who used it in their keyboards. What Woz was to early computer development, Julius is to early audio processing.

Pitch detection and pitch shifting is a very active area of research. Basic approaches use fourier synthesis which has a lot of drawbacks. Novel approaches probably still use FFTs (that's not novel) but apply some novel enhancements that are patentable. I think Julius also patented the Sines + Transients + Noise method of breaking down polyphonic audio in harmonic and non-harmonic components. The harmonic components are shifted conventionally, the transients are shifted with very small FFTs, and the noise passed as unity.

This is getting in algorithm territoriy and is far more patentable, but even then, exactly how you determine what is transients is an algorithm, and all someone needs to do is make an enhancement over your algorithm and they now have a valid patentable material.

Thus in summary, IMO:
1) companies patent ideas all the time because they can afford to waste some money, and want to take advantage of FIRST-TO-FILE in hopes of blocking others. For individuals, ideas are rarely profitable to patent.
2) Novel ways to solve a problem (like Sines + Transients modelling) is patentable, but is easily superceded by a better implementation, and is great for companies/professors who want a fancy score card about how many patents they have, but aren't that profitable.
3) Actual specific algorithms are highly patentable, and if they are good enough may represent the state-of-the-art long enough to pick up some licensing from companies who want to use them. Julius O. Smith III's digtial waveguides are a perfect example of this. They are still pretty much state-of-the-art (even though patent is so old it's public domain now)

[confundido]

Btw very funny comment from audioartillery about watching me from the street at night!