Where is your outrage over [negligent] deaths? [...] should we bring back prohibition for the safety of the children?
The argument was for banning things with a primary purpose to kill, so you're not really providing an effective rebuttal by talking about deaths due to negligence or related to alcohol
How about instead of banning things, we focus our resources on figuring out why people go nuts and try to kill children? Why don't we try to help the nutters before they kill our children?
There I agree with you. But I still ask the question if it is worth taking other measures to reduce homicide rates in the meantime, because surely you're not suggesting that we can solve everyone's psychiatric issues (just as I would not suggest that we could eliminate homicides or even gun-related homicides with gun control)
If someone wants to kill people, they don't need guns.
No, but in many cases access to guns does dramatically increase the damage they can do. (Yes, some people can make bombs...how many?)
That's a good point; in addition to tracking the hum from the signal to be modified, you could use other recordings and power plant data to get information about the hum, line it up to your recording, and create an inverse filter.
(Inverting the hum signal in the time domain as I think you were getting at (were you?) wouldn't work well because the each recording will have different spectral and phase responses, so the signal well never cancel out completely and in fact in some cases may amplify. Like how if were to record a clean signal with the exact same model mic at two different spots in a room, those two signals will not cancel each other just by subtraction. However, once you know the time-varying frequency content of the hum signal, you could knock out those spectro-temporal regions. Of course the resulting notches in the spectrogram is its own signature that screams "this recording was tampered with" but there are probably reasonable ways of then filling in the gaps to obscure this. Then you could superimpose a different hum signal.)
I wouldn't worry too much about this being used as evidence, or i guess I should say I hope i shouldn't have to worry; as others have said elsewhere, this method is good for disproving authenticity, not proving it. But, I suppose, never underestimate what bullshit lawyers can work with.
1. Pet peeve: DFT-based, or freq. domain, filter. FFT, being the algorithm to compute the DFT, has little to do with the actual filtering being done. Anyway, not important.
2. You still can't eliminate the hum that specifically quite like you say, digital or no. Yes it's a lot easier to implement sharper filters in the digital domain, but that has little to do with whether the filtering is done in the freq. or time domain and a lot to do with the ability to run long for loops. Keep in mind that the DFT is only a discrete sampling of the DTFT. Zeroing a point tells you little about the energy immediately bordering that point (between DFT samples). Zeroing out several contiguous points does not remove all energy between those points. The hum varies in frequency, and the analysis window imposes further spectral spread, so just trying to zero out some points of the DFT (if that's what you were implying) will probably not be the best approach.
Even if you, using a well-designed filter, knock out the range of frequencies that you expect the hum to be in, you still have the harmonics of the hum, and eliminating those is going to start eliminating the signal. Maybe if you could track the hum as TFA suggests and implement a time-varying notch filter at all the harmonics...well someone must have tried this, right? I wonder how well it works.
First, if you can find significantly better deals online, then the store is overpriced, period. You have a choice when you run a store: sell cheap and make it up in volume, or sell expensive and lose the sale.
I agree with some of your points, but I object to this simplification as it does accommodate my shopping considerations (and it's fine, of course, that mine are different than yours). I like to support small business that seem to be doing a decent job even if they don't have the volume that allows lower prices (not necessarily "underperforming" in my book). I feel that if nobody did that, markets would more quickly become dominated by a few huge companies, and I believe that would stifle competition and ultimately result in the downfall of the quality of services that those companies would be motivated to provide (e.g. cable providers in many US locations). Perhaps this argument is better suited toward producers and not retailers, but I haven't thought out that distinction too thoroughly yet and I'm still willing to spend a little extra money on that belief.
I'd be curious to hear other opinions regarding the value or folly of supporting small business competitors.
I can't really tell if the authors were in fact more careful than this article suggests, but the following demonstrates your point quite clearly:
Falk and her colleagues also noticed an unusual feature in the right somatosensory cortex, which receives sensory information from the body. In this part of Einstein’s brain, the region corresponding to the left hand is expanded, and the researchers suggest that this may have contributed to his accomplished violin playing.
It is already quite known that experience can cause expanded representation in various cortical areas, so failing to address that this "unusual" feature might have been caused by practicing the violin, rather than being the cause of violin skill, does little to boost the credibility of this article.
I'm not an expert on the subject, but I know there are at least several landmarks in the "nooks and crannies" (search for: gryi and sulci) that are shared across different people, and different areas of the brain are very typically found by looking relative to these landmarks (though often some amount of individual mapping is required in brain imaging studies).
If using a significance level of 0.05, then if you have 20 independent parameters and the null hypothesis is true for all of them, then the probability of all 20 statistical tests showing no difference is 0.95^20 = 0.36. Therefore the probability of getting at least one false positive is 64%. (I think I'm doing that right, anyway. Feel free to correct me)
Of course not all the measurement are independent, etc, and perhaps the authors already corrected for multiple comparisons. I don't really know, I'm not that interested to find and read through the full manuscript. I also can't tell if the unfounded jumps to causal relationships came from the study authors or the summary author.
If these are all true, they should have had some sense of the beating in the dissonance, and been able to at least detect with accuracy greater than chance dissonant notes.
Unless the dissonant notes were played using pure tones and intervals wider than auditory filters, in which case dissonance can exist without beating, no?. I'm sure other methods might have be used as well to tease them apart, this is just what I'm thinking of without having immediate access to the article.
Or maybe the idea that beating and dissonance are related is incorrect.[...] if there was no preference for harmonic tones with amusia, the study cannot exclude beating while including harmonicity as a foundation of musical preference [...] If anything, I would have concluded that beating is not the foundation of dissonance.
I'm a little confused by your objections here. The authors' conclusion is exactly that the idea that beating and dissonance are (perceptually) related is incorrect, as you just stated. The study shows no difference in preference for beating, but differences in dissonance aversion, and presents this as evidence that it is therefore unlikely that beating aversion underlies dissonance aversion.
What I'd question is how trustworthy the measure of "preference" is; such things can be very sketchy in a lab experiment setting.
I believe you missed my point. I admitted that there's likely some reason for an upper limit (and no, IANAP), that's not what at issue with violation of Shannon's limit as per OP, which is what I was addressing.
I'm glad you can sort-of derive that upper limit. Good for you. You know far more physics than I.
Maybe you and I just have different interpretations of what "in principle" means in this context. I'm only referring to the point at hand, which was violation of a the mathematical principle. How we realize the application of that mathematical principle in the physical world is subject to whatever limitations due to our choice of medium. That the medium of EMR has the limitations you point out is wonderful, but is irrelevant to the principle of shannon's limit.
I simply wanted to point out how that mathematical principle was not violated. Is that not clear from my statements? If you think my statements imply in any way that I think there's a potential practical realization of infinite channel capacity, I have clearly failed to convey my meaning.
I believe you missed my point. I admitted that there's likely some reason for an upper limit (and no, IANAP), that's not what at issue with violation of Shannon's limit as per OP, which is what I was addressing. Even still, doesn't your argument about infinite energy still leave us unbounded? It does not provide a strict upper limit to what frequencies we can manipulate, and therefore, still infinite in principle. In practice, obviously, we can't expect to keep upping the energy without bound, I don't and didn't argue against that obvious point.
An aside; seriously, why is everyone so fucking combative in internet posts?
*In principle* is a key modifier, so let's think this through. Doesn't the EMR spectrum *in principle* have infinite capacity if we could use arbitrarily high frequencies to transmit information? I suppose there may be some modern-physics-predicted upper limit to what frequencies of radiation can exist, just as I suppose there may be a limit also to the orbital angular momentum states, but that's not what's at issue with violation of Shannon's limit. Shannon's limit applies to one-dimensional signal. I'm sure for a given bandwidth and a given resolvable orbital angular momentum state, Shannon's limit still applies. But if we have a theoretically unbound number of accessible, resolvable, states to use in each frequency bandwidth, just as there are a theoretically unbound number of finite-frequency bands in the entire EMR spectrum, would it not be appropriate to state that *in principle* the technology would allow for infinite capacity?
If the input waveform would have been unknown, the reconstruction would not aid at all in knowing what word the patient listened to.
Why do you say "not aid at all"? Audio can be reconstructed from spectrograms, and computer matching algorithms can match spectrograms to words (that's basically how they usually do speech recognition). So while accuracy might not have been quite 90% on an open set of unknown words, the procedure would still aid at least a little.
The electrical signals transmitted from the eardrum to the brain would naturally have a pretty tight correspondence with the sound waves received
Near the ear (i.e., auditory nerve), that would be mostly correct. For low frequencies (<2 kHz), the summed impulses on the auditory nerve look very much like the acoustic signal. For high frequencies, they look only like the overall envelope of the acoustic signal, but that's still pretty good and good enough to figure out most speech.
and I would naturally expect that electrical activity in the brain corresponding to regions associated with hearing would be similarly correspondent.
Mostly incorrect. Neurons in cortex fire very slowly (on the order of Hz), often eve nonce per "signal" (loosely defined, discrete chunk of audio). The pathway goes auditory nerve > cochlear nucleus > superior olive > lateral lemniscus > inferior colliculus > medial geniculate > auditory cortex. Very cool, complex computations are being performed at every stage that transform a time waveform received at the eardrum to a perception of sound and the world around you. To understand what the signals at the cortex mean is a huge insight toward reverse-engineering the brain, so don't knock it before you understand it.
Always love a good DIY project, but in this case at least consult with an audiologist first--even if you may never get their approval, you may save your remaining hearing. Depending on the nature of your hearing loss, you can further damage your hearing this way. For example, if you have a problem affecting the mechanics of your inner ear at high frequencies, high-frequency energy that normally goes to driving that portion of the cochlea can pass through, continuing onto areas of the cochlea that still properly transduce lower frequencies. It's quite possible that by sending in *amplified* high-frequencies, you could be accelerating damage to lower-frequency areas of your inner ear.
(Also, when this is the case, you won't hear any better, in fact you'll arguably hear worse as the extra off-frequency energy further down the cochela can cause distortions as the cochlea is highly non-linear)
I'm not sure that's entirely accurate. Granted I'm neither an audiologist nor a hearing aid designer, but I am a hearing researcher and I don't know of frequency shifting devices until you start talking about cochlear implants, which "shift" frequencies by the nature of not being able to stimulate the low-freq. region of your cochlea because which is harder to access, so the signals from low freqs are used to drive relatively higher-frequency locations on the cochlea.
For one, if the hearing impairment is conductive (i.e. something wrong between the outer and inner ear), amplification fit to the shape of hearing loss is exactly what you want to do as the only thing you need to overcome is a mechanical attenuation of acoustic energy.
If the hearing aid is sensorineural (like the damaged hair cells you mention), then, as you mention, at some point amplification won't do squat, since you don't have the cells to transduce those frequencies anymore. So here frequency scaling could be a strategy I suppose, though it adds a lot of computing (and thus battery) power needed, and adds a steep learning curve to the device (which according to some cochlear implant research may not ever be fully learnable)
But as to the overall message, absolutely agreed. You certainly can further damage your hearing with poorly made or fitted hearing aids. For starters, make sure you know not only the shape but mechanism of your hearing loss.
Another thing to keep in mind with GV transcriptions, in addition to norminator's point, is that it's working with crappy audio as received in the voicemail (even I can't understand half the words in a voicemail sometimes). Speech-to-text running natively on the phone, or through a web service as the case may be, has a lot more flexibility to use a representation of the audio optimized for machine transcription. Of course, I don't really know specifics here.
The Verizon-Bing thing applies to Blackberries, not the Droid. A quick Google search on my Droid confirmed those headlines.
-Runs on something other than Verizon (unlike the Droid), namely ATT & TMo.
Again, hardly unique to Google, but this is probably as close as it gets. Also,
It is unique when comparing to the Droid though.
The question was droid/G1/etc. How 'bout the Hero (Sprint)? The MyTouch (T-Mobile)? Or others? Granted, the two I mentioned aren't (?) running Android 2.0 (yet).But that goes into your Andoird 2.1 point; "cheaper plans" is kind of false.
But hey, while we're comparing to the droid,
-More storage space
is the opposite of true. Droid comes with 16 GB, Nexus One with only 4. Both are expandable to 32. Nexus one has twice the RAM of the Droid though (512 MB to 256 MB).
if there was an entry in the white pages saying "Houses that are empty this weekend, and location of keys"
So tell me, where's the "illegal filesharing links" Google search filter?
Norsefire's analogy ain't perfect. Yours is worse. If we really wanted to try to make this analogy fit, it might somewhere in between Norsefire's and the following: the Yellow Pages listing a business that sells you catalogs of houses that are empty this weekend.
I see you got (back) here first. To be honest I'm not sure; most of the mp3s I have I acquired in their current state. The ones I have ripped have been done with some fraction in Winamp using whatever LAME encoder was available at the time (IIRC, which I might not be), and more recently, simply with iTunes. I can't recall though if I've ripped any continuous albums or I just happened to acquire all the ones that matter in their current state.
If you have two identical pieces of music [...] then the results will necessarily be perfectly random. [...].
Conversely, if the results are not random, then the people could necessarily tell differences in them.
This is not the converse, this is the contrapositive (which is necessarily true if the original statement was true). The converse illustrates the flaw in this test with regards to the question "is there a perceivable difference" -- If the results are random, this does not imply that the two pieces of music are perceptually identical. There could be a perceivable difference but no population-consistent preference.
Thus if you asked people "do you hear a difference," you are likely to get many false positives since you are predisposing people to seek a difference.
A typical psychophysical experiment involves a "same/different" task, but you have half of your trials with an actual difference (MP3 / FLAC), and half of your trials with no difference (FLAC/FLAC or MP3/MP3). This allows you to measure performance accounting for bias.
The test in the article is less appropriate to answer "is there a perceptible difference" as it cannot distinguish between the being no difference and there being a difference but no overall preference, or even a difference but no strong individual preferences. You could even imagine a result where the difference was easy to tell, but half the population prefers one, and half the other. I hope from this you can see why this test is not well-suited to answer "is there a perceivable difference."
That is not to say that the test is completely incapable of answering the question; certainly if it results in a response distribution that significantly differs from chance you could from that infer that there is a perceivable difference. However it only answers the question if there is, in addition to a perceivable difference, also an overall population preference. It is like asking the question "is my die roll larger than 2?" and designing a test that only gives a positive result if you roll a 5 or 6.
The MP3 format introduces a small amount of silence at the beginning of the track track as it encodes - and decodes - the file. Because the standard has no way of accounting for this padding, it can not be removed during playback, resulting in annoying gaps
Though I know and have experienced what you're referring to, I do not have gaps in my more recently mp3-encoded albums. Are you sure the gap is / was specifically mp3-related, or simply a silly feature of the encoding program? I do recall most programs having a on-by-default "insert silence" preference, I never understood why. Regardless, I listen to many continuous and semi-continuous (i.e., a few tracks run into each other, but not all) albums, all mp3 encoded, and do not have the issue of gaps between tracks.
"A worrying preponderance" is not "a fraction statistically significantly above chance." As with any such tests you have to keep in mind the possible influence of randomness before assuming that a difference in two conditions means anything. Here I interpret "a worrying preponderance" to mean "not significantly above chance in favor of FLAC", which seems to be what they were hoping for.
Additionally, we need to distinguish between there being a difference and one being preferable over another. Just having a perceivable difference does not imply that FLAC sounds "better". Of course this is more subjective and thus harder to rigorously answer.
If you're really trying to save battery when you're on the go all day, faster boot > sleep. For some reason, my laptop takes longer to resume-from-hibernate than to boot up, so I rarely use that feature anymore. Maybe I'm doing something wrong. If it resume-from-hibernate were faster, then I'd agree with you.
Slashdot Mirror
[...] alcohol doesn't benefit society [...]
That's an interesting claim
Where is your outrage over [negligent] deaths? [...] should we bring back prohibition for the safety of the children?
The argument was for banning things with a primary purpose to kill, so you're not really providing an effective rebuttal by talking about deaths due to negligence or related to alcohol
How about instead of banning things, we focus our resources on figuring out why people go nuts and try to kill children? Why don't we try to help the nutters before they kill our children?
There I agree with you. But I still ask the question if it is worth taking other measures to reduce homicide rates in the meantime, because surely you're not suggesting that we can solve everyone's psychiatric issues (just as I would not suggest that we could eliminate homicides or even gun-related homicides with gun control)
If someone wants to kill people, they don't need guns.
No, but in many cases access to guns does dramatically increase the damage they can do. (Yes, some people can make bombs...how many?)
That's a good point; in addition to tracking the hum from the signal to be modified, you could use other recordings and power plant data to get information about the hum, line it up to your recording, and create an inverse filter.
(Inverting the hum signal in the time domain as I think you were getting at (were you?) wouldn't work well because the each recording will have different spectral and phase responses, so the signal well never cancel out completely and in fact in some cases may amplify. Like how if were to record a clean signal with the exact same model mic at two different spots in a room, those two signals will not cancel each other just by subtraction. However, once you know the time-varying frequency content of the hum signal, you could knock out those spectro-temporal regions. Of course the resulting notches in the spectrogram is its own signature that screams "this recording was tampered with" but there are probably reasonable ways of then filling in the gaps to obscure this. Then you could superimpose a different hum signal.)
I wouldn't worry too much about this being used as evidence, or i guess I should say I hope i shouldn't have to worry; as others have said elsewhere, this method is good for disproving authenticity, not proving it. But, I suppose, never underestimate what bullshit lawyers can work with.
1. Pet peeve: DFT-based, or freq. domain, filter. FFT, being the algorithm to compute the DFT, has little to do with the actual filtering being done. Anyway, not important.
2. You still can't eliminate the hum that specifically quite like you say, digital or no. Yes it's a lot easier to implement sharper filters in the digital domain, but that has little to do with whether the filtering is done in the freq. or time domain and a lot to do with the ability to run long for loops. Keep in mind that the DFT is only a discrete sampling of the DTFT. Zeroing a point tells you little about the energy immediately bordering that point (between DFT samples). Zeroing out several contiguous points does not remove all energy between those points. The hum varies in frequency, and the analysis window imposes further spectral spread, so just trying to zero out some points of the DFT (if that's what you were implying) will probably not be the best approach.
Even if you, using a well-designed filter, knock out the range of frequencies that you expect the hum to be in, you still have the harmonics of the hum, and eliminating those is going to start eliminating the signal. Maybe if you could track the hum as TFA suggests and implement a time-varying notch filter at all the harmonics...well someone must have tried this, right? I wonder how well it works.
First, if you can find significantly better deals online, then the store is overpriced, period. You have a choice when you run a store: sell cheap and make it up in volume, or sell expensive and lose the sale.
I agree with some of your points, but I object to this simplification as it does accommodate my shopping considerations (and it's fine, of course, that mine are different than yours). I like to support small business that seem to be doing a decent job even if they don't have the volume that allows lower prices (not necessarily "underperforming" in my book). I feel that if nobody did that, markets would more quickly become dominated by a few huge companies, and I believe that would stifle competition and ultimately result in the downfall of the quality of services that those companies would be motivated to provide (e.g. cable providers in many US locations). Perhaps this argument is better suited toward producers and not retailers, but I haven't thought out that distinction too thoroughly yet and I'm still willing to spend a little extra money on that belief.
I'd be curious to hear other opinions regarding the value or folly of supporting small business competitors.
Falk and her colleagues also noticed an unusual feature in the right somatosensory cortex, which receives sensory information from the body. In this part of Einstein’s brain, the region corresponding to the left hand is expanded, and the researchers suggest that this may have contributed to his accomplished violin playing.
It is already quite known that experience can cause expanded representation in various cortical areas, so failing to address that this "unusual" feature might have been caused by practicing the violin, rather than being the cause of violin skill, does little to boost the credibility of this article.
I'm not an expert on the subject, but I know there are at least several landmarks in the "nooks and crannies" (search for: gryi and sulci) that are shared across different people, and different areas of the brain are very typically found by looking relative to these landmarks (though often some amount of individual mapping is required in brain imaging studies).
If using a significance level of 0.05, then if you have 20 independent parameters and the null hypothesis is true for all of them, then the probability of all 20 statistical tests showing no difference is 0.95^20 = 0.36. Therefore the probability of getting at least one false positive is 64%. (I think I'm doing that right, anyway. Feel free to correct me)
Of course not all the measurement are independent, etc, and perhaps the authors already corrected for multiple comparisons. I don't really know, I'm not that interested to find and read through the full manuscript. I also can't tell if the unfounded jumps to causal relationships came from the study authors or the summary author.
If these are all true, they should have had some sense of the beating in the dissonance, and been able to at least detect with accuracy greater than chance dissonant notes.
Unless the dissonant notes were played using pure tones and intervals wider than auditory filters, in which case dissonance can exist without beating, no?. I'm sure other methods might have be used as well to tease them apart, this is just what I'm thinking of without having immediate access to the article.
Or maybe the idea that beating and dissonance are related is incorrect.[...] if there was no preference for harmonic tones with amusia, the study cannot exclude beating while including harmonicity as a foundation of musical preference [...] If anything, I would have concluded that beating is not the foundation of dissonance.
I'm a little confused by your objections here. The authors' conclusion is exactly that the idea that beating and dissonance are (perceptually) related is incorrect, as you just stated. The study shows no difference in preference for beating, but differences in dissonance aversion, and presents this as evidence that it is therefore unlikely that beating aversion underlies dissonance aversion.
What I'd question is how trustworthy the measure of "preference" is; such things can be very sketchy in a lab experiment setting.
I believe you missed my point. I admitted that there's likely some reason for an upper limit (and no, IANAP), that's not what at issue with violation of Shannon's limit as per OP, which is what I was addressing.
I'm glad you can sort-of derive that upper limit. Good for you. You know far more physics than I.
Maybe you and I just have different interpretations of what "in principle" means in this context. I'm only referring to the point at hand, which was violation of a the mathematical principle. How we realize the application of that mathematical principle in the physical world is subject to whatever limitations due to our choice of medium. That the medium of EMR has the limitations you point out is wonderful, but is irrelevant to the principle of shannon's limit.
I simply wanted to point out how that mathematical principle was not violated. Is that not clear from my statements? If you think my statements imply in any way that I think there's a potential practical realization of infinite channel capacity, I have clearly failed to convey my meaning.
I believe you missed my point. I admitted that there's likely some reason for an upper limit (and no, IANAP), that's not what at issue with violation of Shannon's limit as per OP, which is what I was addressing. Even still, doesn't your argument about infinite energy still leave us unbounded? It does not provide a strict upper limit to what frequencies we can manipulate, and therefore, still infinite in principle. In practice, obviously, we can't expect to keep upping the energy without bound, I don't and didn't argue against that obvious point.
An aside; seriously, why is everyone so fucking combative in internet posts?
*In principle* is a key modifier, so let's think this through. Doesn't the EMR spectrum *in principle* have infinite capacity if we could use arbitrarily high frequencies to transmit information? I suppose there may be some modern-physics-predicted upper limit to what frequencies of radiation can exist, just as I suppose there may be a limit also to the orbital angular momentum states, but that's not what's at issue with violation of Shannon's limit. Shannon's limit applies to one-dimensional signal. I'm sure for a given bandwidth and a given resolvable orbital angular momentum state, Shannon's limit still applies. But if we have a theoretically unbound number of accessible, resolvable, states to use in each frequency bandwidth, just as there are a theoretically unbound number of finite-frequency bands in the entire EMR spectrum, would it not be appropriate to state that *in principle* the technology would allow for infinite capacity?
If the input waveform would have been unknown, the reconstruction would not aid at all in knowing what word the patient listened to.
Why do you say "not aid at all"? Audio can be reconstructed from spectrograms, and computer matching algorithms can match spectrograms to words (that's basically how they usually do speech recognition). So while accuracy might not have been quite 90% on an open set of unknown words, the procedure would still aid at least a little.
The electrical signals transmitted from the eardrum to the brain would naturally have a pretty tight correspondence with the sound waves received
Near the ear (i.e., auditory nerve), that would be mostly correct. For low frequencies (<2 kHz), the summed impulses on the auditory nerve look very much like the acoustic signal. For high frequencies, they look only like the overall envelope of the acoustic signal, but that's still pretty good and good enough to figure out most speech.
and I would naturally expect that electrical activity in the brain corresponding to regions associated with hearing would be similarly correspondent.
Mostly incorrect. Neurons in cortex fire very slowly (on the order of Hz), often eve nonce per "signal" (loosely defined, discrete chunk of audio). The pathway goes auditory nerve > cochlear nucleus > superior olive > lateral lemniscus > inferior colliculus > medial geniculate > auditory cortex. Very cool, complex computations are being performed at every stage that transform a time waveform received at the eardrum to a perception of sound and the world around you. To understand what the signals at the cortex mean is a huge insight toward reverse-engineering the brain, so don't knock it before you understand it.
Always love a good DIY project, but in this case at least consult with an audiologist first--even if you may never get their approval, you may save your remaining hearing. Depending on the nature of your hearing loss, you can further damage your hearing this way. For example, if you have a problem affecting the mechanics of your inner ear at high frequencies, high-frequency energy that normally goes to driving that portion of the cochlea can pass through, continuing onto areas of the cochlea that still properly transduce lower frequencies. It's quite possible that by sending in *amplified* high-frequencies, you could be accelerating damage to lower-frequency areas of your inner ear.
(Also, when this is the case, you won't hear any better, in fact you'll arguably hear worse as the extra off-frequency energy further down the cochela can cause distortions as the cochlea is highly non-linear)
I'm not sure that's entirely accurate. Granted I'm neither an audiologist nor a hearing aid designer, but I am a hearing researcher and I don't know of frequency shifting devices until you start talking about cochlear implants, which "shift" frequencies by the nature of not being able to stimulate the low-freq. region of your cochlea because which is harder to access, so the signals from low freqs are used to drive relatively higher-frequency locations on the cochlea.
For one, if the hearing impairment is conductive (i.e. something wrong between the outer and inner ear), amplification fit to the shape of hearing loss is exactly what you want to do as the only thing you need to overcome is a mechanical attenuation of acoustic energy.
If the hearing aid is sensorineural (like the damaged hair cells you mention), then, as you mention, at some point amplification won't do squat, since you don't have the cells to transduce those frequencies anymore. So here frequency scaling could be a strategy I suppose, though it adds a lot of computing (and thus battery) power needed, and adds a steep learning curve to the device (which according to some cochlear implant research may not ever be fully learnable)
But as to the overall message, absolutely agreed. You certainly can further damage your hearing with poorly made or fitted hearing aids. For starters, make sure you know not only the shape but mechanism of your hearing loss.
Another thing to keep in mind with GV transcriptions, in addition to norminator's point, is that it's working with crappy audio as received in the voicemail (even I can't understand half the words in a voicemail sometimes). Speech-to-text running natively on the phone, or through a web service as the case may be, has a lot more flexibility to use a representation of the audio optimized for machine transcription. Of course, I don't really know specifics here.
-Runs on something other than Verizon (unlike the Droid), namely ATT & TMo.
Again, hardly unique to Google, but this is probably as close as it gets. Also,
It is unique when comparing to the Droid though.
The question was droid/G1/etc. How 'bout the Hero (Sprint)? The MyTouch (T-Mobile)? Or others? Granted, the two I mentioned aren't (?) running Android 2.0 (yet).But that goes into your Andoird 2.1 point; "cheaper plans" is kind of false.
But hey, while we're comparing to the droid,
-More storage space
is the opposite of true. Droid comes with 16 GB, Nexus One with only 4. Both are expandable to 32. Nexus one has twice the RAM of the Droid though (512 MB to 256 MB).
if there was an entry in the white pages saying "Houses that are empty this weekend, and location of keys"
So tell me, where's the "illegal filesharing links" Google search filter?
Norsefire's analogy ain't perfect. Yours is worse. If we really wanted to try to make this analogy fit, it might somewhere in between Norsefire's and the following: the Yellow Pages listing a business that sells you catalogs of houses that are empty this weekend.
I see you got (back) here first. To be honest I'm not sure; most of the mp3s I have I acquired in their current state. The ones I have ripped have been done with some fraction in Winamp using whatever LAME encoder was available at the time (IIRC, which I might not be), and more recently, simply with iTunes. I can't recall though if I've ripped any continuous albums or I just happened to acquire all the ones that matter in their current state.
If you have two identical pieces of music [...] then the results will necessarily be perfectly random. [...].
Conversely, if the results are not random, then the people could necessarily tell differences in them.
This is not the converse, this is the contrapositive (which is necessarily true if the original statement was true). The converse illustrates the flaw in this test with regards to the question "is there a perceivable difference" -- If the results are random, this does not imply that the two pieces of music are perceptually identical. There could be a perceivable difference but no population-consistent preference.
Thus if you asked people "do you hear a difference," you are likely to get many false positives since you are predisposing people to seek a difference.
A typical psychophysical experiment involves a "same/different" task, but you have half of your trials with an actual difference (MP3 / FLAC), and half of your trials with no difference (FLAC/FLAC or MP3/MP3). This allows you to measure performance accounting for bias.
The test in the article is less appropriate to answer "is there a perceptible difference" as it cannot distinguish between the being no difference and there being a difference but no overall preference, or even a difference but no strong individual preferences. You could even imagine a result where the difference was easy to tell, but half the population prefers one, and half the other. I hope from this you can see why this test is not well-suited to answer "is there a perceivable difference."
That is not to say that the test is completely incapable of answering the question; certainly if it results in a response distribution that significantly differs from chance you could from that infer that there is a perceivable difference. However it only answers the question if there is, in addition to a perceivable difference, also an overall population preference. It is like asking the question "is my die roll larger than 2?" and designing a test that only gives a positive result if you roll a 5 or 6.
The MP3 format introduces a small amount of silence at the beginning of the track track as it encodes - and decodes - the file. Because the standard has no way of accounting for this padding, it can not be removed during playback, resulting in annoying gaps
Though I know and have experienced what you're referring to, I do not have gaps in my more recently mp3-encoded albums. Are you sure the gap is / was specifically mp3-related, or simply a silly feature of the encoding program? I do recall most programs having a on-by-default "insert silence" preference, I never understood why. Regardless, I listen to many continuous and semi-continuous (i.e., a few tracks run into each other, but not all) albums, all mp3 encoded, and do not have the issue of gaps between tracks.
"A worrying preponderance" is not "a fraction statistically significantly above chance." As with any such tests you have to keep in mind the possible influence of randomness before assuming that a difference in two conditions means anything. Here I interpret "a worrying preponderance" to mean "not significantly above chance in favor of FLAC", which seems to be what they were hoping for.
Additionally, we need to distinguish between there being a difference and one being preferable over another. Just having a perceivable difference does not imply that FLAC sounds "better". Of course this is more subjective and thus harder to rigorously answer.
If you're really trying to save battery when you're on the go all day, faster boot > sleep. For some reason, my laptop takes longer to resume-from-hibernate than to boot up, so I rarely use that feature anymore. Maybe I'm doing something wrong. If it resume-from-hibernate were faster, then I'd agree with you.
"I keep my [computer] asleep most of the time, so that I have a completely usable machine in two seconds."
And you don't see the potential benefit of a quick boot time?