Quality of musical instrument sound identification for various levels of accompanying sounds

Research on automatic identification of musical instrument sounds has already been performed through last years, but mainly for monophonic singular sounds. In this paper we work on identification of musical instrument in polyphonic environment, with added accompanying orchestral sounds for the training purposes, and using mixes of two instrument sounds for testing. Four instruments of definite pitch has been used. For training purposes, these sounds were mixed with orchestral recordings of various levels, diminished with respect to the original recording level. The level of sounds added for testing purposes was also diminished with respect to the original recording level, in order to assure that the investigated instrument actually produced the sound dominating in the recording. The experiments have been performed using WEKA classification software.