Analysis of Cetacean Data and Algorithm Development


  1. Bougher, B.
  2. Hood, J.
  3. MacKillop, E.
Corporate Authors
Defence Research and Development Canada, Atlantic Research Centre, Halifax NS (CAN);Akoostix Inc, Dartmouth NS (CAN)
One of the many challenges faced by cetacean data analysts is the large volume of data they must review in order produce concise results. The objective of this project was to improve on algorithms that support automation of this process. Algorithms intended to produce a list of vocalization events from click-detection data, and to produce an estimate of the number of vocalizing animals based on different time difference of arrival for a vocalization sequence were investigated. During the course of this work, methods for reducing the number of false detections due to multipath and general methods for reducing the number of false detections were also investigated. The algorithm that produces a list of vocalization events examines detection density, attempting to group associated clicks, while discarding random detection events. This is useful because it provides the analyst with a reduced number of events to validate during further analysis, which is beneficial for data where detection is infrequent. The most promising algorithm for counting the number of vocalizing animals results in an image, where animals are represented by lines in the image (traces) and the shape of the line is representative of motion. The method requires frequent vocalization (i.e. a click train) and at least two sensors spaced sufficiently for signal time difference (STD) estimation. The analyst can quickly examine the image and isolate individuals, though multipath can result in overestimation of the a
sonar;marine mammal
Report Number
DRDC-RDDC-2017-C115 — Contract Report
Date of publication
01 May 2017
Number of Pages
Electronic Document(PDF)

Permanent link

Document 1 of 1

Date modified: