Application of artificial neural networks to the recognition of musical sounds

摘要: The aim of the presented work was to train a neural network in order recognize class chosen musical instrument. As problems related analysis sounds are human subjective perception abilities, then it seems that such tools analyses as nets should be used for recognition processes. On other hand, an artificial not trained directly with subsequent samples sound, thus feature extraction procedure is needed at first. As, there no consensus regarding selection parameters vector extraction, experiment aimed check whether calculated sufficient creating set sound patterns training. Some were investigated experiment, they so-called ELEVEN and FOURTEEN types. After learning executed, examples previously created database (but seen by network) nets. Results show NNs (neural networks) able generalize information included vectors. Therefore, when presenting data NN inputs, problem variation within data, consequently clustering, because has ability during phase. In paper, experimental results will carried on, conclusions derived from performed tests presented.