The frequent activity taking place at Mount Etna may pose risks to tourists and specialists visiting the summit craters as well as to the nearby population and thus requires constant monitoring. Infrasound recordings play an important role in volcanic observation because shallow explosive activity, shallow tremor processes or other volcanic phenomena coupled with the atmosphere are easier to identify with sound waves travelling through the air than with seismic waves, which are more effective for the characterization of buried sources but are strongly affected by scattering within the volcano edifice. Similar to seismic waves, infrasound signals often interfere with noise sources, in case of Mount Etna mostly wind induced noise. The manual distinction of noisy data from real volcanic signals is often not easy, and becomes unrealistic when a large amount of data has to be processed. Currently five summit craters at Mt Etna are active, showing intermittent and fluctuating levels of activity. This leads to a wide variety of infrasound signal patterns coming along with changing noise levels. In order to distinguish between noise induced signals and different kinds of volcano induced signals – such as infrasound events and tremor – in the waveform data we apply unsupervised pattern recognition techniques. We show that by extracting features from the amplitude spectrum, a simple clustering analysis with K-Means yields a set of different activity regimes. A more indepth analysis of the patterns is possible by means of Self-Organizing maps (SOMs) which also allow for the identification of transitional activity regimes and provide the option to color-code the results for an intuitive interpretation. We create a reference data set from multiple months of infrasound waveforms to include as many activity regimes as possible. We apply the results to a test set of data showing how new data can be treated fast and efficiently.