Representation of temporal coherence: CHAINS algorithm and FPGA implementation

CHAINS is a conceptual framework and a computational theory for separating temporal events that are mixed together and is proposed as a model of auditory scene analysis. It relies on computational structures that involve competition and co-operation between low-level event predictors, which are represented by graphs. CHAINS attempts to capture streaming phenomena, i.e. dynamic percepts of a group of successive and/or simultaneous sounds as a coherent whole appearing from a single source. The CHAINS algorithm as originally casted is a bottom up algorithm for self organization. It captures sequential processes at multiple scales from tens of milli-seconds to 100s of milli-seconds and up to a few seconds long.