Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!usc!chaph.usc.edu!alcor.usc.edu!alves
From: alves@alcor.usc.edu (William Alves)
Newsgroups: comp.music
Subject: Re: Books on massaging waveforms...
Keywords: samples sound
Message-ID: <10494@chaph.usc.edu>
Date: 29 Jun 90 22:23:33 GMT
References: <4090@uqcspe.cs.uq.oz.au> <987@anaxagoras.ils.nwu.edu>
Sender: news@chaph.usc.edu
Organization: University of Southern California, Los Angeles, CA
Lines: 41

>In article <4090@uqcspe.cs.uq.oz.au>, tonyg@batserver.cs.uq.oz.au (Tony
>Gedge) writes:
>> Can someone point me towards a good starting book in manipulating
>> sampled sound?  I want to do things like:
>> 
>>   * Raise and lower the pitch of samples without changing the speed
>>         that samples are played.
>>   * Mix two or more samples together.
>>   * Change the sampling rate.
>> 
Try "Programs for Digital Signal Processing" (IEEE press, 1979). It con-
tains many sample programs (unfortunately all in Fortran) to do all of
what you're asking and much more. To mix 2 samples, simply add them, taking
care that the sum doesn't go out of range. The easy (but not optimal) way
to do this is to divide each one by 2 before adding. The optimal way would
be to have a normalization pass.

In article <987@anaxagoras.ils.nwu.edu> sandell@ils.nwu.edu (Greg Sandell) writes:
>For changing the
>sampling rate, if you are going from a higher sampling rate to a lower
>one, you can keep one sample, delete the next (new_samp_rate/old_samp_rate) 
>samples, and keep doing this for the rest of the soundfile.  I have done this
>in going from 22000k to 8000k and the results sound just fine.  Somehow
>I suspect this is not the most HiFi way to do this task, though.  As for going
>from a smaller sampling rate to a larger one, the strategy of duplicating
>samples in the equivalent way will *definitely* not work!  

Your results probably sounded okay because you didn't have too much frequency
information above the new Nyquist frequency. If you did, you would experience
aliasing. Thus your suspicions are correct. One should first apply a low-pass
filter to make sure this doesn't happen. Likewise, when interpolating samples
(smaller to larger sampling rate), one must take care that the new samples
don't introduce unwanted frequencies.

Because this is a fundamental function of samplers, several solutions have
been developed. One of the fastest is detailed in the following paper:
Andy Duncan and Dave Rossum, "Fundamentals of Pitch Shifting," presented 
at the 85th convention of the Audio Engineering Society, preprint #2714 
(1988). These are the fellows who developed the Emulator at Emu.

Bill Alves