Towards detection of perceptually similar sounds: investigating self-organizing maps

This p j p e r outl ines a s \ s i e m for the detect ion of p e i c e p t u j i l y s imilar sounds ( ' s o u n d spo t t ing ' ) , reports on a ser ies ot prehminaiN experiments and discusses their results T h e sound spot t ing sys tem pursues a f r ame-based app roach and consists of three mam stages, an auditory model , a s e l t -o igamzing m a p and a pat tern ma tch ing a lgor i thm. T h e expel iments described examine how dif ferent types of se l f -o rgan izmg m a p s c lass i fy a set of test sounds preprocessed by an auditory model and e\ aluate their pe r fo rmance by means ot v isual izat ions and qual i ty measures . With these o u t c o m e s in mind we suggest directions for the furthei deve lopment of the sound spot t ing sys tem.