Slashdot Mirror
Your Computer As Your Singing Coach
Roland Piquepaille writes "Israeli researchers have developed an electronic ear to coach vibrato technique. Until now, the quality of a vibrato — the pulsating change of pitch in a singer's voice — could only be judged by voice experts. Now, a Tel Aviv University research team 'has successfully managed to train a computer to rate vibrato quality, and has created an application based on biofeedback to help singers improve their technique.' Interestingly, this research could be used for other applications, such as improving automated help centers, where computers could be trained 'to recognize a range of different emotions, such as anger and nervousness.'"
Interestingly, this research could be used for other applications, such as improving automated help centers, where computers could be trained 'to recognize a range of different emotions, such as anger and nervousness.
This task should be quite easy. Frustration and anger are the only two emotions I tend to experience when I get through to an automated help center. It would be a better investment of time to evaluate how long I spend interfacing with the system, how many times I have to re-navigate the menu hierarchy, how many times I have to call back and start over, how many actual people I end up being directed to, how many times I have to restate the same information and how long I spend talking to someone before I solve my problem, if I ever do. .. but I'm not bitter..
How is this technology new? I remember hearing many many years ago that they had developed gadgets that you could attach to your phone that could more or less sense if a person was nervous or not and could even function as a lie detector. These devices were probably pretty primitive, and their claims of being able to be used to spot when someone was lying to you were probably a little over the top, however, this technology doesn't strike me as new.
They could also just listen for:
"Goddamnit!"
"You piece of shit!"
or the always indicative "FUCK!"
Just by checking for those 3 phrases, they should be able to ID an angry caller with at least a 99% positive rate.
Give a man a fire and he'll be warm for a day. But light a man on fire and he'll be warm for the rest of his life.
I've said it once, I'll say it again...vibrato is just cheating when you can't sing the actual pitch. Seriously, just pick a note and sing it. What's so wrong about that?
Google's Super Secret Search Algorithm: SELECT @search_results FROM internet WHERE @search_results = 'good'
how pleasant someone's singing voice is, is a completely subjective thing that can only really be properly judged by other human beings.
They said nothing about the pleasantness of the singing voice. The system judges the quality of the vibrato.
While I believe the above was referring to "quality" in a scientific sense rather than how "good" or "bad" it was, the whole hypothesis of one's "vibrato" having all that much to do with whether one is a good singer or not is hogwash. There are many uses of vibrato from virtually none (listen to a good singer perform Handel) to a ton (listen to a different good singer perform Wagner), where the amount of vibrato in a given style changes over the course of a phrase...etc.
In the end, all this algorithm can probably do is determine how well a singer can constrain their voice into a certain vibrating range, having much more to do with muscular control than technique.
"In the end, we all fall back on fiction." -- Lonely Planet
OK Mr. Literal. But what they're working towards being able to quantify with a machine, is someone's singing voice.
I'm not sure that's necessarily true. The summary says they're using "biofeedback to help singers improve their technique". Based on that, it would seem they're more interested in it as an educational tool rather than a tool for critics. There are a number of other technologies to help musicians improve their technique, so it's not like this is the first. For instance, many wind musicians will practice playing long tones with a digital tuner to improve their overall intonation. There are also systems that schools use that track the notes that you've played, compare them with the notes you were supposed to have played, and tell you what you've done wrong.
ZuluPad, the wiki notepad on crack
...every programmer should work with something like this at least once. I did some audio programming work in college, and its a totally different world than the regular web dev stuff I have done, because you're working with the convergence of acoustics and physics with programming. In true signal processing apps, what you are doing has to happen FAST as well, which makes the guys who work in it true wizards, and that's without even considering the subjective recognition stuff that these researchers had to do. Kudos to them.
-The art of programming is the pursuit of absolute simplicity.
Yep. It's time for bed. I read "Your computer as your singing couch"
good night folks.
I have to say, you're stepping into a high-expertise field armed with a perilous lack of technical knowledge.
There are already numerous types of acoustical analysis and biofeedback in use in many places for the training of elite vocalists - by that I mean high-level classical singers. These include spectography which can be used to examine tonal balance factors, legato, vowel differentiation and modification, and so on; the electro-glottal graph which is a device that measures vocal fold closure and displays the individual cycles which can be used to evaluate pressed vs. breathy phonation; a device which measures the relative expansion and contraction of the chest and abdomen during breathing and singing and graphs them.
Contrary to your assertion, vibrato is a very important pedagogical tool. Vibrato rates that are too rapid (above 7.5 cycles per second or so), too slow (below 4 cycles), or too wide all indicate specific types of technical deficiency.
Vibrato is an important element of vocal technique as well, because the achievement of consistently vibrant sound through the range, and through different vowels, is an important goal in the training of singers. Vibrato is generally not related to muscle control factors except largely to the extent that through muscle tension or 'holding' the presence of vibrato can be reduced or eliminated. This is called "straight toning."
A tool that can help to measure quantitative vibrato factors: rate, consistency, pitch excursion, changes in dynamic, etc., could be very helpful in the training of singers. These are all subject to acoustical analysis and there's no reason to think that this machine wouldn't be able to do it.
As a matter of style, for both historical reasons and modern aesthetic reasons, I believe Handel should be sung with a fully vibrant sound. The tenor for whom Handel wrote Messiah and many of his other works was a full dramatic tenor whose large voice bore little resemblance to the light, lyric tenors who generally perform that music today for reasons of "historical accuracy."
I also find it somewhat odd that Shakira is held up as a model for good vibrato. She has a bleating vibrato which varies not only in pitch but in dynamic as well, which in another singer would be considered a serious technical deficiency.
I also find it somewhat odd that Shakira is held up as a model for good vibrato. She has a bleating vibrato which varies not only in pitch but in dynamic as well, which in another singer would be considered a serious technical deficiency.
It's not odd when you consider that most people can't tell when Auto-Tune has been used on a track.
It should be programmed to give a poor score for using a vibrato. I don't know about you but I can't stand it when I go to a ball game and have to tolerate a 40 minute version of the national anthem because the singer vibratos every line of the song for a full minute.
eg: "The land of the FRE-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E-E............."
Personally I consider it a mockery of the anthem.
I haven't RTFA (never do) but it would be interesting if the biofeedback somehow encouraged you in directions that you felt pleased with. That's the general kind of trajectory I think of when I think of "feedback", so it would make sense in that regards. I'm not sure how that would work, maybe encouraging you when you felt good about your results.
I just don't understand these singing competitions, their appraisals seem totally random. I've sang in a bunch of choirs and worked on the open vowels and proper articulation, but I prefer the Lou Reed/Leonard Cohen school of singing.
Actually, singing without vibrato is a singing style just as much as singing with vibrato. "Mixed" singing uses both, classical uses all vibrato, and a Capella uses almost none. Your statement indicates you prefer "mix" as a singing style.
Here's what I know, and forgive me if any of this seems rudimentary, but I think vibrato (like singing generally) is not well understood by most people:
Vibrato is a cyclic departure from and return to a pitch. When a cellist holds a note and wobbles her left hand without starting a new note, or when B.B. King does the same, that is vibrato. It is heard as a "throb" in the voice, especially in those voices where it coincides with a cycling of intensity as well. This pulsing quality is something that musical instruments can rarely capture.
Some things vibrato is not:
-Tremolo: the repetition of a note, usually rapid, despite the misuse of the term in electric guitar circles to mean pitch-bending equipment.
-Glissando: a change in pitch moving in one direction, like a slide whistle or a pianist running a finger across the keys.
-Trill: the rapid alternation of two distinct notes, though in some voices this can sound a lot like vibrato
-Melisma: in vocal music, the inclusion of many notes on one vowel -- think Mariah Carey
In singing, most or all of the excursion of a person's vibrato is below the note being held. The graph of a person's vibrato would rarely look like a perfect sine wave, but usually would have an element of saw wave mixed in. That is, during the 1/6th of a second of an average vibrato cycle, the pitch might drop fairly quickly to the bottom of the range of excursion (let's say 1/3 of a whole tone) and take the rest of that time to climb back to the "correct" pitch, and perhaps go sharp by a few cents briefly.
The rate, shape, dynamics and excursion of a singer's vibrato is something that a well-trained singer can tell with some accuracy after a few seconds of listening. "Eight beats per second, rather smooth, consistent dynamic, and shallow," for example. It is a an objective evaluation, and I'm not surprised a machine can do it too.
But it is terribly difficult to change one's natural vibrato. It takes months of practice and guidance for the typical voice student with a poor vibrato to improve it. Knowing that the end result (the voice) comes from a combination of physiology, psychology, and technique that involves muscles from the face to the feet, I don't see how this type of feedback will help them fix it.
Assuming, of course, that it needs fixing. The ideal of a moderate and inoffensive vibrato, while present in many successful singers' voices and most opera singers' voices, is also conspicuously absent from the voices of many well-loved singers and entertainers.
811.29.3.2
Yeah, but Shakira makes up in lack of singing talent with huge... tracts of land.
stuff |