Automated detection and localization of bowhead whale sounds in the presence of seismic airgun surveys.

摘要: An automated procedure has been developed for detecting and localizing frequency-modulated bowhead whale sounds in the presence of seismic airgun surveys. The was applied to four years data, collected from over 30 directional autonomous recording packages deployed a 280 km span continental shelf Alaskan Beaufort Sea. six sequential stages that begin by extracting 25-element feature vectors spectrograms potential call candidates. Two cascaded neural networks then classify some as calls, matches calls between recorders triangulate locations. To train networks, manual analysts flagged 219 471 examples 2008 2009. Manual analyses were also used identify 1.17 million transient signals not calls. network output thresholds adjusted reject 20% training data. Validation runs using 2007 2010 data found missed 30%–...