Time-series changing frequency

Synthetic time-series: 20 time-series of three types are analysed. The signals of the second type increase their frequency at time step t1=150 and get back to their original frequency at time step t2=300. The other signals remain unchanged. (Top) Current signals. (Bottom left) IKSC model: cluster prototypes in the eigenspace. (Bottom right) Actual kernel matrix. In the initialization phase the IKSC algorithm recognizes 2 clusters. After some time, we notice that IKSC successfully detects the first change in frequency of the signals of the second type by creating a new cluster at time step t=223. Moreover, the second change point is detected at time step t=382, when a merging of two clusters is performed.