In collaboration with:
Université de Bretagne Occidentale
The project SABRE aims to build an open database based on the acoustic profiles of various auditoriums, likely to be characteristic of any place where audio programs identifiable are involved, through a simple test protocol and a comparative procedure.
The acoustic measurement instrument is an analyzer with Spectral Density Integration which stands out from traditional spectrum analyzers by its frequency analysis and the kind of signals processed, from which results a specific time-related functioning. While a standard analyzer divides the audio spectrum in regular parts, the SDI adapts to the spectral sensitivity of human ear thanks to an irregular dividing which bounds are empirical. While a standard analyzer uses stationary or almost stationary signals, the SDI uses real stimulus perceptible by human ears in a specific situation. While a standard analyzer does not have a "memory" because of the "stationary" hypothesis, the SDI represents the distant-like sensation, that is to say the fact that hearing is influenced by sounds that preceded those perceived at every moment.
By mimicking, even approximately, some human hearing features, the SDI gets an essential characteristic: returning a "verdict" only until a certain level of learning is reached. For human hearing, there is no reference, no standards, but only "memorized experiences" to which any current perception can be compared and likely to enter and enrich this memory, when certain criteria are met. Composed of SDI profiles obtained in typical situations (as far as places, sounds and number are concerned), the database reproduces this learning process through variation, selection and memorization. At this stage of the project, the test protocol as well as its content must be precisely defined, and most importantly, normalization procedures must be set up to guarantee the device’s "robustness", especially in order to authorize improvement without distorting the "base" on which comparisons can be made.
However the database, just as any memory, would only be abstract if it was not connected to the world of actions and representations, thanks to various processes of association. For the human being connected to his environment, these associations result from his biological organization and from what Gilbert Simondon calls "individuation". As far as the database is concerned, these associations must be explicitly programmed so that a categorization can be established and efficient representations can be inferred.