Dolphins Tracked with Sound Classification

Note the statement of the problem of classification here:

Tracking Dolphins With Algorithms You Might Find on Facebook 
The New York Times    By Steph Yin

Researchers at the Scripps Institution of Oceanography have used a machine-learning algorithm similar to Facebook's friend recommender program to track dolphins. The program analyzed 52 million dolphin clicks and identified seven distinct groups of sound, which the researchers think correspond to different kinds of dolphins. Scripps' Kait Fraiser and colleagues first ran a detection program though years of audio recordings and extracted all segments with dolphin clicks, which their algorithm segmented into five-minute blocks, generating an average click rate and frequency shape for each time window. The algorithm then clustered five-minute chunks with similar average click rates and frequency profiles, and Frasier says it took only about four days to sort through several years of data from five sites. The unsupervised algorithm extrapolated seven discrete click clusters, one of which was consistent with the singular click profile of the Risso's dolphin species, which Frasier notes was a "good sanity check" suggesting their method might work.  .... "