💾 Archived View for oberdada.pollux.casa › gemlog › 2024-06-08_solutions.gmi captured on 2024-06-16 at 12:14:54. Gemini links have been rewritten to link to archived content
-=-=-=-=-=-=-
Music Information Retrieval (MIR) uses feature extraction for purposes such as classification into categories, source recognition, and distance measures between audio excerpts. In order to capture some of the human auditory capabilities, feature extraction in MIR usually relies on frequency domain techniques and various transforms. It occured to me that much simpler time domain signal descriptors might be useful in some field I haven't thought of, not necessarily audio applications which is what I've been concerned with.
My collection of signal descriptors is now available for download along with detailed explanations of what they do.
https://ristoid.net/prog/timedomain.html
There are interesting applications such as adaptive effects where some of the descriptors might be used. Dynamic processors such as compression and expansion are adaptive effects, they multiply the input with a gain factor that depends on the current value of a signal descriptor, which in this case is the signal's amplitude. Adaptive synthesis is a similar case, where signal descriptors from an input sound controls synthesis parameters.
I'm sure there are lots of other applications. So far I haven't had much need for the new signal descriptors, except once as a pitch dependent gain control while mixing one of my new pieces.
My generative poetry has traveled the world (small parts of it, at least) as mail art and there are a few animated readings. Now I'm considering publishing a selection as an artist book. Although I write the programs that generate the texts I hardly understand a word of it. Even if a seemingly familiar word such as "then" might show up, nothing in the context indicates that it is the English word, and the same goes for any word that might be recognisable in any other language. It's a fluke, an accident, the words only belong to the poem. Still, it is puzzling how the words take on appearances of being noun-like, verb-like, adjectives or adverbs, or belonging to the fuzzy class of small words including nouns, articles, and prepositions. Or maybe not, because I have intentionally constructed the program so that it places the short words before the longer ones and uses a limited word pool for each poem, with repeated words gaining some significance.
It is as if there were significance by signifiers without a signified, if that makes sense. Conlangs, as far as I know, begin in the other end, by setting up a grammar and introducing words with specified meanings. My procedure is more like Tristan Tzara's hat in which you put letters and fragments of words and grab them in random order; there you have your poem, unique and reflecting your personality, if beyond the layman's comprehension, as Tzara himself put it. Whatever semblance of grammar there is may be deduced in retrospect by analysing the texts. Their meaning, if any, floats untethered in the air, free to be captured by anyone's associations.
Un coup de dés is being mastered these days and will be released on CD in a sufficiently small edition that I can have some hope of selling it out before the EOC (end of civilization). I'll just have to design a cover and organise a few other things. It will be available for digital download as well, but probably not on any streaming platform. So far I have relied on bandcamp, but its near monopoly status is reason enough to look for a backup plan.