This accessible page is a derivative of https://www.researchcatalogue.net/view/2938321/3620177 which it is meant to support and not replace.
Image description: In white text and a diagram on a black background the SNDArchive is depicted and explained. The SNDArchive is an experimental analysis frameworks for feature-based analysis (SCMIR) of sound files and the reassembling of those through playback agents.
Click on https://www.researchcatalogue.net/view/2938321/3620177#tool-3620371 to see the screenshot.
SND Archive
Experimentation with sound analysis on large audio collections led to the development of a custom compositional system named SNDArchive. Its primary function was to analyse sounds, offering normalised ranges for the attributes utilised in the analysis process. The analysis results would be stored in a database that could be queried once the analysis was done. The dataset then forms an archive that can be further explored and researched. The purpose was to discover aspects of the material that I had not understood before. For example, I could build an archive from folders on a hard drive and then access their loudness or analyse pitch as a number from 0.0 to 1.0. In addition to analysing sounds, the sounds could also be split into segments that would be analysed in the same way. This allowed for many features, such as grouping high-frequency segments of a group of sounds, or playing all sounds in a folder in a sequence based on perceived timbral centre. I experimented with numerous ways of querying the archive. These operations became part of the codebase and can be further used by me or others. Since the database used follows a standard SQL design, it is also easy to create custom queries using the well-known format of SQL queries.
Applying the attitudes of data science meant to compose by querying the database. To ask questions or make queries based on data properties of the set as a whole. The SNDArchive system further allowed me to recompose and combine sound parts based on different dimensions discovered through offline analysis processes. My task then became to make queries and selections on top of sound directories and choose how they would be played back. The music emerges through processes that interpret the sound or sound segments, transforming them or developing synthetic sounds based on their properties. The idea is to create methods for engaging with the sound archive, to review it from a different angle or to reveal previously unknown aspects of material already loaded with meaning.
Video description: A video demonstrating the queries that are made possible using the SNDArchive. Composing takes places through the queries made on the dataset. The system contains many pre-made queries and custom queries can also be made. The query view favours a set-based attitude to sound collections. For example, one can ask for a set of high frequency sounds combined with low amplitude ones. By selecting the attributes and asking for results, a query-based approach evolves.
Click on https://www.researchcatalogue.net/view/2938321/3620177#tool-3620377 to watch the video.
Image description: A Screenshot of the project's GitHub repository. The SNDArchive is open source and built with Supercollider, Javascript, NodeJS, SQLite and Sequalize.
Click on https://www.researchcatalogue.net/view/2938321/3620177#tool-3622371 to see the screenshot.