Hi all

I know this is not exactly forum about sound but i need some help. Do you know any good algorithm, which should change e.g. speech rate? I have recorded speech and I need something to slow ti down or fasten it up. I know that audacity can do this but i have my application and I'd like to know how to work with sound. Do you know any idea or some algorithms to work with it?

thx.

You mean, you want to implement such features for your app?

I'd suggest you look at the sources of audacity plug-ins as well LADSPA plug-ins that do these kinds of things.

Probably audacity dev list, alsa lists or such could give you more information.

to do it the quick and dirty way:
to speed up the sample rate::duplicate samples
to slow it own::skip samples

Do you want mathematical background or do you want implementation ?

If the latter, for the whole sound system or, say, batch processing for files ?

I might be mistaken, but when you do this by adding or deleting samples, you are changing the pitch of the sound.

Imagine that you have a sine for which one period is described with 20 samples. If you play it by omitting every second sample, you play this sine in the same time, but only using 10 samples per period. 10 samples in the same time as previously held 20 samples with a constant sample rate really increases the frequency.

jlinkels

but you change the tag to match saying to play it at half speed

Cool, then you have exactly the same speed of speech, just with half of the samples. What have you achieved?

jlinkels

as for what the op wanted
the orignal method will work
nomatter how you change the speed of the sound
it will always change pitch too

No, it does not. Modern software programs are able to chnage the speed without changing the pitch. I believe that is done by changing the sampling rate, and shifting the frequency by an equal amount in the frequency domain.

jlinkels

Isn't the OP simply wanting to speed up or slow down the playback rate? Kind of like spinning the LP disk faster or slower on the turntable? Digitized sound is basically a time ordered series of discreet samples. Accurate playback fidelity implies using the same conversion rate for the DAC as was used for the ADC in the recording phase. Not doing so will alter the pitch. Most digital sound files contain sample rate information, which can potentially be altered, such that the playback rate will differ from the recording rate. The exact format of the file will dictate what would need to be changed, and also what changes would be seen as valid. There may be tools which make this kind of editing simple, however with knowledge of the file format, one should be able to use a generic binary-style editor.

On the other hand, the original poster may be asking how to re-sample a recorded sound file. If the sound is re-sampled with fewer samples per unit of time, it will decrease both the file size and fidelity. If the sound is re-sampled with more samples per unit of time, it will increase the file size, but will not increase fidelity. A quick Google search for 'audio resampling algorithm' turned up numerous useful hits for me.

--- rod.