Impulsive Source Separation for Sperm Whale Click TrainsDate: 2016-11-16 Add to Google Calendar
Location: Holmes Hall 389
Speaker: Jeremy Young, PhD student, University of Hawaiʻi Electrical Engineering
Passive acoustic monitoring (PAM) is a popular method to study marine mammals. As its name suggests, it involves passively monitor the animals by recording the sounds they make over long periods of time. Assuming the data can be processed properly, population densities can be estimated, animals can be localized, and behavior can be studied amongst other things. However, a common stumbling block is how to process the large amount of data extract the sounds of interest. There are many parts to this processing, here we will focus on the source separation problem: When multiple mammals are within the detection range of a hydrophone, their sounds are mixed together and we cannot tell which sound goes with which animal. This is a well studied area in signal processing as there are many applications where the source signals need to be recovered from the outputs of an unknown mixing system; it is called blind source separation (BSS). While the BSS field is quite mature, the application to marine mammal vocalizations is not. Specifically, we consider the problem of separating an impulsive type of vocalization called a "click." In addition to traditional methods, a novel algorithm based on the regularity inter-click-interval will be presented.