5 ms·
An Introduction to Sine-Wave Speech (2007)
- smashah 6y agoWhoa. Spooky. I wonder if this works with a text spoiler also.
- Evidlo 6y agoThere's the classic chainmail that claims you can scramble the inner letters of all words in a paragraph and it still remain legible. Turns out some of the words are cherry picked, but it still works somewhat. Aoccdrnig to a rscheearch at Cmabrigde Uinervtisy, it deosn't mttaer in waht oredr the ltteers in a wrod are, the olny iprmoetnt tihng is taht the frist and lsat ltteer be at the rghit pclae. The rset can be a toatl mses and you can sitll raed it wouthit porbelm. Tihs is bcuseae the huamn mnid deos not raed ervey lteter by istlef, but the wrod as a wlohe. http://www.mrc-cbu.cam.ac.uk/people/matt-davis/cmabridge/ http://www.mrc-cbu.cam.ac.uk/people/matt-davis/cmabridge/
- smashah 6y agoNo I mean send someone a sine wave speech file and via another method the transcription of the speech. Would you hear it after reading it?
- tomxor 6y agoAfter hearing the first two examples with clear speech I could identify most of the last three without, took a few listens before the last one clicked.
- Tomte 6y agoI saw the other submission yesterday and thought it was very cool. Today it's more meh. The priming didn't hold for me from yesterday. I clicked the first of the four extra examples and didn't understand anything. After listening to the clear version and trying again I heard it. But again priming did not hold for me (unlike what's described on the page): each of the other three examples were unintelligible to me (not even recognizable as speech sounds). Again, after listening to the clear version, it became easy, but only for the corresponding sine-wave encoding. I still like that idea, if only for nostalgic reasons. We did some formant analysis at university, up to trying to "read" a sonagram, i.e. deducing fron the sonagram as shown on the page what was being said.
- hprotagonist 6y agoever played with praat? it’s the pinnacle of highly specialized totally non-ux-designed software.
- Tomte 6y agoI've had it installed, but didn't do much with it. We mostly used Wavesurfer. But yes, Praat is fantastic if you're really deep into that stuff.
- yorwba 6y agoIts source code is also highly ideosyncratic with lots of macro use, including definitions like "our" for "this ->" https://github.com/praat/praat/blob/9d20024e2f68f84bf23ff49bf313b2e7ba24fb42/melder/melder.h#L41 https://github.com/praat/praat/blob/9d20024e2f68f84bf23ff49b...
- hprotagonist 6y agoweirdo academic code at its “finest”
- mhandley 6y agoI first saw this on Twitter four days ago, but only one example was given there. Turns out that priming from a single example four days ago did hold for me - I got the second half of all the sentences today without any further priming, and the first half of several of them. Seems like it has quite different effects on different people.
- andrelaszlo 6y agoI listened to only the "degraded" sample, and I could understand it on the second listen. English is not my native language and I have pretty severe hearing loss. I suspect the latter makes it easier for me in some situations to decode language with a lot of noise - I have a lot of practice guessing, I suppose. It would be interesting seeing some research on this.
- tricolon 6y agoThe last time I listened to these was ten years ago. I understood all of them just now.
- bzb3 6y agoThis reminds me of voice inversion, a scrambling method used by some analogue portable radios. https://en.m.wikipedia.org/wiki/Voice_inversion https://en.m.wikipedia.org/wiki/Voice_inversion
- carapace 6y agoDIBS! I called it! Dibs on this for robot voices! H.E.L.P.eR-style - - - - Unrelated, there are at least thirty-five Venture Bro.'s (so far.) I don't want to show my work here because it would be spoiler. - - - - Seriously though: this is how "droids"(tm)(R) should talk, no? - - - - Also unrelated, the only speech act of H.E.L.P.eR's that I can understand in the entire series is when he and Brock are discussing poetry and he asks Brock, "Maya Angelou?". That's it. :-)
- kragen 6y agoAwesome :) You could probably turn up the comprehensibility knob by adding a fourth formant or a little bit of vocoded white noise. Or a sawtooth or FM or AM tone or something.
- kragen 6y agohttp://www.haskins.yale.edu/featured/sws/sws.html http://www.haskins.yale.edu/featured/sws/sws.html is a page the Haskins Lab folks have put up about this; unfortunately at some point they switched it to use Flash or something, so now it's unusable. The archived version at https://web.archive.org/web/20060828113107/www.haskins.yale.edu/featured/sws/sws.html https://web.archive.org/web/20060828113107/www.haskins.yale.... directly links to AIFF files; by telling Firefox to open them with /usr/bin/play or /usr/bin/vlc I can hear them. The original experiments were done in the 1970s, I think, and the papers on the page are I think from 1981–1994. It would have been interesting to see how things would have developed if this formant-centric viewpoint had become central to speech codec research in the 80s and 90s. In a sense LPC codecs like GSM RTE-LTP are formant-centric, since the linear prediction coefficients in some sense encode the formant structure of the speech, but (and I'm getting out my depth here) as I understand it the impulse response of the linear prediction filter is pretty broadband. We really need a better medium than the current WWW for this kind of archival, because the researchers are going to retire or die, and then they will probably stop updating their pages, which will then apparently become inaccessible with modern browsers.
- unlikelymordant 6y agoIt is my understanding that most voice codecs are still essentially lpc based, it is just that lpcs are unstable when quantised, so they get converted to things like line spectral pairs, then there is all sorts of different excitations and codebooks used to eek out every bit of savings. But the core idea of lpcs are still in there. Happy to be corrected by someone who knows better!
- kragen 6y agoI would be too! I think "LPCs are unstable when quantised" is a little overgeneralized — a more precise statement is that linear predictors, being all-pole filters, can be stable or unstable, and that quantization can convert a marginally stable or even absolutely stable filter into an unstable one. But I could easily be misunderstanding something basic! I've never understood the LSP thing. (Yet!)
- 6y ago
- tauntz 6y agoRandom data point but as a non-native speaker I could understand about 80% of the words on the first go. Listening to the clear audio didn't really make me "hear" the correct word on the 2nd try. It just made me realize that "yes, that word that I know is correct here, might indeed sound like this" IDK how to explain it. Take for example, the "The camel was kept in a cage at the zoo" sentence. I heard "The owl was kept in a cage at the zoo" on the first go and after listening to the clear version, I still hear it's an "owl". I now know that "camel" is also reasonable there and makes sense, but I still hear "owl". Am I wired incorrectly? :)
- swixmix 6y ago> Am I wired incorrectly? :) Doubtful. Reminds me of when Grover was accused of saying a bad word. https://youtu.be/HfwF5cuAMsk https://youtu.be/HfwF5cuAMsk
- crazygringo 6y agoWhat I find interesting is that I only need priming on one, then I can understand all of them once I realize I'm supposed to be listening for speech that's behind some kind of high-pitched distortion. It reminds me somewhat of when I'm in a bilingual conversation and somebody unexpectedly switches to the other language for just a quick word or two when you don't expect it. Your brain tries to interpret the sounds in the first language and they're just gibberish, even though you speak the second language. Then if the speaker continues in the second language quickly, you "retroactively" re-interpret what you just heard and understand it.