Dynamic clustering of transient signals

A computer-implemented method for signal analysis of noisy transient signals observed in time or space, comprising the steps of:

• capturing the noisy transient signals with a measurement system;
• configuring a processor with a memory connected thereto to:
• dynamically build up clusters of similar transient signals in a hyperspace based on comparison and clustering rules stored in the memory so that each new one of the noisy transient signals captured ends up in a cluster stored in the memory with similar noisy transient signals formerly captured at the time of the acquirement, by comparing each new one of the noisy transient signals captured with cluster signatures defined by noisy transient signals formerly captured, the comparison and clustering rules using metrics modelling each cluster close to a hypersphere having a radius characteristic of noise in the noisy transient signals gathered in the cluster;
• and
• analyze the clusters stored in the memory so that respective signatures defined by the noisy transient signals gathered in the clusters stored in the memory are determined;
• and
• process the signatures so that a phenomenon connectable to an intrinsic attribute of the noisy transient signals is detected, wherein the building up of the clusters comprises merging similar noisy transient signals and similar clusters when a predetermined maximum number of clusters is reached, and computing new parameters of the comparison and clustering rules applicable to the clustering of new noisy transient signals captured, and wherein, to reduce a computational time of the processor to detect the phenomenon, the processor refrains from determining a signature for each of the transient signals, and determines one respective signature for each dynamically built-up cluster.