Joint localization and separation of sperm whale clicks

Paul M. Baggenstoss


In this paper we consider the joint problems of separating and localizing sperm whale click trains. Click train separation is the single-sensor problem of grouping the clicks from each animal together when the clicks of more than one animal are present at a given sensor. Localization is the problem of localizing the animals based on the measurement of time delays of the same click events at multiple sensors. The two problems are inherently connected. We first consider the two problems independently using novel applications of statistical signal processing methods. For separation, we employ an algorithm inspired by the Viterbi algorithm from dynamic programming. For localization, we employ an algorithm inspired by the expectation-maximization (EM) algorithm. Finally, we use the two algorithms to "assist" each other in a joint localization/separation solution. We demonstrate the algorithm on real data.


Blind source separation; Boolean functions; Isomers; Locomotives; Mathematical programming; Railroad cars; Sensors; Separation; Signal processing; Statistical methods; Systems engineering; Multiple sensors; Novel applications; Sperm whales; Viterbi

Full Text:



  • There are currently no refbacks.