Recognition and separation of sounds played by various instruments is very useful in labeling audio files with semantic information. This is a non-trivial task requiring sound analysis, but the results can aid automatic indexing and browsing music data when searching for melodies played by user specified instruments. In this chapter, we describe all stages of this process, including sound parameterization, instrument identification, and also separation of layered sounds. Parameterization in our case represents power amplitude spectrum, but we also perform comparative experiments with parameterization based mainly on spectrum related sound attributes, including MFCC, parameters describing the shape of the power spectrum of the sound waveform, and also time domain related parameters. Various classification algorithms have been applied, including k-nearest neighbor (KNN) yielding good results. The experiments on polyphonic (polytimbral) recordings and results discussed in this chapter allow us to draw conclusions regarding the directions of further experiments on this subject, which can be of interest for any user of music audio data sets.