Downloads & Free Reading Options - Results

A three-step approach has been developed for detecting and classifying the foraging calls of the blue whale, Balaenoptera musculus, and fin whale, Balaenoptera physalus, in passive acoustic recordings. This approach includes a pattern recognition algorithm to reduce the effects of ambient noise and to detect the foraging calls. The detected calls are then classified as blue whale D-calls or fin whale 40-Hz calls using a machine learning technique, a logistic regression classifier. These algorithms have been trained and evaluated using the Detection, Classification, Localization, and Density Estimation (DCLDE) annotated passive acoustic data, which were recorded off the Central and Southern California coast from 2009 to 2013. By using the cross-validation method and DCLDE scoring tool, this research shows high out-of sample performance for these algorithms, namely 96% recall with 92% precision for pattern recognition and 96% accuracy for the logistic regression classifier. The result was published by the Institute of Electrical and Electronics Engineers (2016). The advantages of this automated approach over traditional manual methods are reproducibility, known performance, cost-efficiency, and automation. This approach has the potential to conquer the challenges of detecting and classifying the foraging calls, including the analysis of large acoustic data sets and real-time acoustic data processing.

A three-step approach has been developed for detecting and classifying the foraging calls of the blue whale, Balaenoptera musculus, and fin whale, Balaenoptera physalus, in passive acoustic recordings. This approach includes a pattern recognition algorithm to reduce the effects of ambient noise and to detect the foraging calls. The detected calls are then classified as blue whale D-calls or fin whale 40-Hz calls using a machine learning technique, a logistic regression classifier. These algorithms have been trained and evaluated using the Detection, Classification, Localization, and Density Estimation (DCLDE) annotated passive acoustic data, which were recorded off the Central and Southern California coast from 2009 to 2013. By using the cross-validation method and DCLDE scoring tool, this research shows high out-of-sample performance for these algorithms, namely 96% recall with 92% precision for pattern recognition and 96% accuracy for the logistic regression classifier. The result was published by the Institute of Electrical and Electronics Engineers (2016). The advantages of this automated approach over traditional manual methods are reproducibility, known performance, cost-efficiency, and automation. This approach has the potential to conquer the challenges of detecting and classifying the foraging calls, including the analysis of large acoustic data sets and real-time acoustic data processing.