Julien Claassen wrote: Great link. Thanks ! -- Olivier

alex stone wrote: Does protools include some automatic speech noise detection and/or removal utilities? I know these of course, not as an advanced user though. It looks like you are talking about 100% manual noise removal. The users I'm dealing with already do that on protools on Mac. They lose a lot of time with this. I'm looking for a way to automate this, at least partly. I think they really need this, so, if needed, I suppose I could convince them to use Ardour for this specific task. In which case I'd try to code something to add this feature to Ardour. I suppose there is mainly two ways to do this: 1 - noise detection, manual removal (as mentionned by Dave on this thread). This is certainly the most flexible thing, safest, and also the most language independent. Because all selected regions could be removed at once it could also provide a high level of productivity 2 - complete automatic removal. It would be more language specific (just think about those tongue sounds in some african languages..), but may still work for a certain range of European languages (I'm primarily targetting french). The second solution has the advantage that it could be developed as a simple LV2 plugin or so, so that I wouldn't need to go and deal with the details of Ardour editing routines, as in the first solution. I can't exactly tell how much work I will be able to put into this, I might need to restrain my ambitions to stay within some realistic professional requirements (read: money). I like Rezound too, but the above remarks still apply. Thanks. But don't forget this is a LAD question, and that I'm not afraid to code (that's my job actually). -- Olivier

Julien Claassen wrote: Thanks about this idea, but I'm more familiar with C.. Plus I need something that integrates very well into a graphical daw, such as a plugin. -- Olivier

Julien Claassen wrote: I'll see what I can do about that, I can't promise anything though. -- Olivier

Julien Claassen wrote: After thinking about it, I think a plugin might be the way the go.. Not yet sure about it though. Btw, what are the available text-based hosts for LV2 plugins? Does ecasound support them? Julien, by curiosity, what would you use this speech filter for? -- Olivier

Julien Claassen wrote: Indeed.. Interesting point. What about vamp plugins? Any text-based hosts? -- Olivier

Hello Chris and Olivier! I don't know about these two. As a matter of fact, I never heard of Vamp plugins in detail before this discussion. I knew a tool called vamp or vamps, which had something to do with video. I'll have alook into the two mentioned bits of software. The sonic anotator sounds interesting. Kindest regards Julien -------- Music was my first love and it will be my last (John Miles) ======== FIND MY WEB-PROJECT AT: ======== http://ltsb.sourceforge.net the Linux TextBased Studio guide ======= AND MY PERSONAL PAGES AT: ======= http://www.juliencoder.de

alex stone wrote: That sounds encouraging. What to you mean by "pre and post" (sorry if that's an obvious question to you)? I think you're right. There's nothing universal about making noise and actually meaning something ;) -- Olivier

Alex! Why different "modules", just different configurations. It sounds to me, as if this software would be doing the same thing (the same algorythms and what not) over and over again. Just offer standard configs (how much and where to apply this and that piece of technique). just some thoughts... Warm regards Julien -------- Music was my first love and it will be my last (John Miles) ======== FIND MY WEB-PROJECT AT: ======== http://ltsb.sourceforge.net the Linux TextBased Studio guide ======= AND MY PERSONAL PAGES AT: ======= http://www.juliencoder.de

alex stone wrote: [...] Alright, got that. We've done some experimentations. On a 3 minutes of speech recording, we got 53 noises, with the following repartition: inspire: 37 expiration: 2 lips: 6 nose: 3 glottal: 1 inspire+lips: 4 That means "inspiring" (breathing in, between two phrases or groups of words) noises makes 70% of the noises. These can be silenced, no need for frequency filtering, because they always happen "pre and post" as you say, and they're apparently always preceded and followed by small silences. Here's the spectrogram+waveform of two inspire noises (the cursor, a white vertical line, is on the noise). On each view appears 1 noise, surrounded by speech: http://www.samalyse.com/code/speechfilter/inspire1.png http://www.samalyse.com/code/speechfilter/inspire2.png We've also measured the duration of 14 inspire noises. Except for 1, all of them are under 1 second. The durations range from from 256ms to 1024ms, with an average of 529ms. An automatic way of removing these inspire noises may largely satisfy the users I'm dealing with. Saving 70% of manual editing time is everything but marginal. So I'm going to concentrate on that at first, leaving all lips, nose, ... noises for later. Correct me if I'm wrong, but if I code this as a plugin which exposes parameters, I think that presets should be handled by the host, not the plugin. Anyway, I'm not sure I could code this as a plugin, because, for detecting the inspire noise, I would need to buffer something like 2 seconds of signal. I suppose that might not be such a problem though, there's already plenty of non RT-capable plugins... Plus, before removal, visual/auditive review of the detected noises (in some sort of audio editor) sounds quite important: there's alway a risk to confuse a noise with the end of a phrase or an other element of speech. So I might need to craft a little gui, or manage to integrate this detection into rezound, ardour, etc.. Anyway, before this happens I need to find a way to detect the noises. A colleague has told me that the best technology in this field currently involves using a database of noise recordings. You then try to find these noises in the signal by doing a more-or-less tolerant comparison in frequency domain. However, looking at the above spectrograms and waveforms, I think there could be a more algorithmic way of detecting these noises, thus avoiding the need for such a database, given the following facts: 1 - on the waveform: their amplitude is much lower than speech 2 - on the spectrogram: the frequencies in the noise seem to spread rather homogeneously (maybe a bit like white noise) where the speech contain noticeable peaks under 1000Hz or so. Do you think I could use these characteristics to detect the noises? PS: Alex, maybe that you could try and improve the way you handle citations when posting? That's not essential, but would make your replies more readable... -- Olivier

Chris Cannam wrote: That's a very interesting idea. I don't think splitting would do the job. Any chance that the noise regions reported by the vamp plugin could turn into selected regions in Audacity? So that one can edit them, for example to ignore a false positive, before silencing all remaining regions at once? What about vamp and ardour? Any other hosts for vamp plugins? -- Olivier

On Thu, Mar 12, 2009 at 3:21 PM, Paul Davis <paul(a)linuxaudiosystems.com> wrote: Sorry, I didn't mention Ardour just because it doesn't currently allow you to take an arbitrary plugin and run it on an arbitrary bit of audio (not least because there is no single action that would be appropriate to take, to handle the results output by any possible plugin). But yes, it has all the code in place. A generic Vamp plugin output -> automation curve adapter within Ardour might be an interesting thing to have... Chris

On Thu, Mar 12, 2009 at 11:43 AM, Olivier Guilyardi <list(a)samalyse.com>wrote;wrote: currently VAMP is used within the context of "rhythm ferret" - the analysis done by the VAMP plugin becomes a set of lines drawn on the region(s) and you can then optionally split the region(s) at the lines. at the moment, we use the note onset and percussive onset detection plugins.

On Thu, Mar 12, 2009 at 2:19 PM, Olivier Guilyardi <list(a)samalyse.com>wrote;wrote: no. doing this within the context of rhythm ferret (which would be odd) would take less than a day. doing it without the GUI part would take me no more than a few hours. doing a custom GUI ... hard to say.