Forums

Downshifted pitches and sample-rates to decrease bandwidth-usage of audio files.

Started by Green Xenon [Radium] September 11, 2007
Hi:

Decreasing the pitch of the audio in an audio linear-PCM [wave] file 
means the sample-rate of the wave file can be decreased without causing 
aliasing. If the bit-resolution and # of channels [1 in mono, 2 in 
stereo] of the file are kept constant, then decreasing the sample-rate 
will decrease the file size. Adobe Audition allows the alteration of 
pitch without changing speed.

http://www.adobe.com/products/audition/overview2.html#kmhead3

"Time and pitch processing: shift pitch without changing tempo - and 
never introduce audio artifacts."

If the pitch of the audio can be decreased this way, then sample-rate 
can also be decreased until it is just 2.5x the maximum frequency of the 
audio signal in the file. Mathematically, only 2x the max frequency is 
necessary, however, due to physical factors, its best to use at least 
2.5x the max frequency.

Suppose it is desired to decrease the file size, keep the good audio 
quality, and not use any compression. In this case, the pitch of the 
audio can be decreased sufficiently so the sample-rate can also be 
decreased sufficiently to store this file in devices with not much space 
or transfer on the internet on low-bandwidths. Instructions regarding 
how the file was processed and what the audio signals and sample-rate 
were like before the processing can be stored with the file. Upon 
reading the file, these instructions tell the software to increase the 
sample-rate and pitch of the audio file back to what is was prior to the 
processing. Then the audio can be played back w/out any artifacts.

For best results, the pitch would be decreased all the way until the 
lowest-frequency signal can finish its cycle in the appropriate amount 
of time � e.g. if a song is 5 minutes long, the lowest frequency signal 
is downshifted until it has a frequency of 1 cycle every 5 minutes. This 
ensures that the information is not lost or cut-off. After this, the 
software would assign an appropriate sample rate for the file. If the 
highest frequency signal is 1,000 Hz, then the sample rate would be 
2,500 Hz.

Has this technique ever been tried in the past? If so, was it as 
efficient as I think it would be?


Thanks,

Radium
Green Xenon [Radium] wrote:
> Hi: > > Decreasing the pitch of the audio in an audio linear-PCM [wave] file > means the sample-rate of the wave file can be decreased without causing > aliasing. If the bit-resolution and # of channels [1 in mono, 2 in > stereo] of the file are kept constant, then decreasing the sample-rate > will decrease the file size. Adobe Audition allows the alteration of > pitch without changing speed. > > http://www.adobe.com/products/audition/overview2.html#kmhead3 > > "Time and pitch processing: shift pitch without changing tempo - and > never introduce audio artifacts." > > If the pitch of the audio can be decreased this way, then sample-rate > can also be decreased until it is just 2.5x the maximum frequency of the > audio signal in the file. Mathematically, only 2x the max frequency is > necessary, however, due to physical factors, its best to use at least > 2.5x the max frequency. > > Suppose it is desired to decrease the file size, keep the good audio > quality, and not use any compression. In this case, the pitch of the > audio can be decreased sufficiently so the sample-rate can also be > decreased sufficiently to store this file in devices with not much space > or transfer on the internet on low-bandwidths. Instructions regarding > how the file was processed and what the audio signals and sample-rate > were like before the processing can be stored with the file. Upon > reading the file, these instructions tell the software to increase the > sample-rate and pitch of the audio file back to what is was prior to the > processing. Then the audio can be played back w/out any artifacts. > > For best results, the pitch would be decreased all the way until the > lowest-frequency signal can finish its cycle in the appropriate amount > of time � e.g. if a song is 5 minutes long, the lowest frequency signal > is downshifted until it has a frequency of 1 cycle every 5 minutes. This > ensures that the information is not lost or cut-off. After this, the > software would assign an appropriate sample rate for the file. If the > highest frequency signal is 1,000 Hz, then the sample rate would be > 2,500 Hz. > > Has this technique ever been tried in the past? If so, was it as > efficient as I think it would be? > > > Thanks, > > Radium
-- Engineering is the art of making what you want from things you can get. ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Green Xenon [Radium] wrote:
> Hi: > > Decreasing the pitch of the audio in an audio linear-PCM [wave] file > means the sample-rate of the wave file can be decreased without causing > aliasing. If the bit-resolution and # of channels [1 in mono, 2 in > stereo] of the file are kept constant, then decreasing the sample-rate > will decrease the file size. Adobe Audition allows the alteration of > pitch without changing speed. > > http://www.adobe.com/products/audition/overview2.html#kmhead3 > > "Time and pitch processing: shift pitch without changing tempo - and > never introduce audio artifacts." > > If the pitch of the audio can be decreased this way, then sample-rate > can also be decreased until it is just 2.5x the maximum frequency of the > audio signal in the file. Mathematically, only 2x the max frequency is > necessary, however, due to physical factors, its best to use at least > 2.5x the max frequency. > > Suppose it is desired to decrease the file size, keep the good audio > quality, and not use any compression. In this case, the pitch of the > audio can be decreased sufficiently so the sample-rate can also be > decreased sufficiently to store this file in devices with not much space > or transfer on the internet on low-bandwidths. Instructions regarding > how the file was processed and what the audio signals and sample-rate > were like before the processing can be stored with the file. Upon > reading the file, these instructions tell the software to increase the > sample-rate and pitch of the audio file back to what is was prior to the > processing. Then the audio can be played back w/out any artifacts. > > For best results, the pitch would be decreased all the way until the > lowest-frequency signal can finish its cycle in the appropriate amount > of time � e.g. if a song is 5 minutes long, the lowest frequency signal > is downshifted until it has a frequency of 1 cycle every 5 minutes. This > ensures that the information is not lost or cut-off. After this, the > software would assign an appropriate sample rate for the file. If the > highest frequency signal is 1,000 Hz, then the sample rate would be > 2,500 Hz. > > Has this technique ever been tried in the past? If so, was it as > efficient as I think it would be?
The scheme won't work. I leave it to you to think for once. Silly questions merely advertise your ignorance. Jerry -- Engineering is the art of making what you want from things you can get. ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
-snip-

> Has this technique ever been tried in the past? If so, was it as > efficient as I think it would be? > > Thanks, > > Radium
As always, what's important is the total bandwidth of the signal. Shifting things around in frequency doesn't buy you anything; your minimum sampling frequency is still determined by the total bandwidth of the signal, which you can't change without distorting it in some way. Basically, all you're describing is shifting some bandpass signal down to (almost) baseband, which doesn't do anything for you with respect to compressibility. Jason
cincydsp@gmail.com wrote:
> -snip- > >> Has this technique ever been tried in the past? If so, was it as >> efficient as I think it would be? >> >> Thanks, >> >> Radium > > As always, what's important is the total bandwidth of the signal. > Shifting things around in frequency doesn't buy you anything; your > minimum sampling frequency is still determined by the total bandwidth > of the signal, which you can't change without distorting it in some > way. Basically, all you're describing is shifting some bandpass signal > down to (almost) baseband, which doesn't do anything for you with > respect to compressibility.
No, no, the trap he laid for himself is more seductive than that. He's contemplating pitch shifting, not frequency shifting. As conservation of energy rules out perpetual-motion devices, just so does conservation of information rule out his compression scheme. I won't write more about why until he's had time to come up with a coherent explanation. Jerry -- Engineering is the art of making what you want from things you can get. ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
On Sep 11, 5:06 pm, "Green Xenon [Radium]" <gluceg...@excite.com>
wrote:
> "Time and pitch processing: shift pitch without changing tempo - and > never introduce audio artifacts." > > If the pitch of the audio can be decreased this way, then
The problem is that in the general case it can't. Pitch shifting downwards either throws away time domain snippets or frequency bins, or increases the total duration of the audio. Sometimes there is no audible information in the snippets or bins which the processing throws away, and you can compress the bits down closer to their information content (like an mp3 coder tries to). Sometimes there is audible information, and too much compression will sound bad. IMHO. YMMV.
Ron N. wrote:

 > Pitch shifting
 > downwards either throws away time domain snippets or
 > frequency bins, or increases the total duration of the audio.

The frequencies are downshifted but the shapes of the audio signals 
remain similar.