In our paper we describe a novel technique for harmonic sound scene decomposition. It is based on the "frequency-place" mapping of a sound event. The defined set of harmonic neural networks follows the same organizational principles as can be seen in the auditory pathways and in the cerebral cortex of the human brain (tonotopic and vertical organizations of the neurons). The defined set of harmonic networks consists of 1066 first-class neurons, 23344 second-class neurons and 23344 interneurons. The experimental results have shown that the defined set of networks is capable to perform similar "frequency-place" mapping of the sound events, as is sought to be carried out in the auditory pathways. The results also indicate that the defined set might be useful in corrupted speech processing and speaker-independent speech processing and recognition

Keywords: harmonic sound event, isofrequency neural column, tonotopic organization, frequency-place mapping, harmonic neural network.