The use of computers to analyze sounds spectrally has great musical potential, because music is all about repetition and variation. The creative potential of spectral analysis resides in an understanding of how sound behaves through time.
This project begins with finding some objects that have interesting sonic properties. These could be natural objects, or pieces of junk, or indeed musical instruments.
The first exercise is to try to describe each sound. Is it periodic or aperiodic? What are its timbral characteristics? How does the sound change over time? Try to make a diagram of the envelope of the sound. How does it begin and end?
Now listen closely and try to deduce spectral information. Are the harmonics true, or partial? Or a combination of these? Is there variation over time in the harmonics, or do they seem stable? Can individual harmonics be identified and do any take precedence during the sound?
Finally, make a recording of the sound. Try cutting off the attack - is it still recognizable? Now consider the complete sound: look at the waveform and then at a spectrogram. To what extent does this confirm the impression gained by your ears? Are there any obvious differences?
Apply this process to a number of 'sonic objects' and compare the differences and similarities between them. Can you classify the results or divide the sounds into meaningful groupings? Finally, ask whether thinking of these as 'sonic objects' is a useful thing to do.
This can be a useful substitute for stage 1 of the . . . from scratch project, if that is being undertaken.