N.R. Pinto¹ , T.K. Chandrayadula¹

¹Indian Institute of Technology Madras

Abstract:

The HA08 hydrophone triads to the north and south of Diego Garcia constantly record low frequency whale calls. This work develops detectors for different types of blue whales; the Antarctic, Sri Lankan, Madagascan, and an unidentified type. There are significant challenges for constructing detectors. The calls are complicated, and show variations in intensity across frequency bands, and within the three hydrophones of a triad. There is hence a variability to the Signal to Noise Ratio (SNR) available for detection. A major cause for the variability is the scattering due to the island bathymetry. This work proposes a robust framework to overcome these issues in two stages. The first stage uses subspace approaches to detect calls across respective frequency bands. The second stage compensates for the SNR variability to improve the detection rates. In this stage, a distributed approach fuses detections across call frequencies, and the hydrophone triads. The other focus of this work is building a three dimensional propagation model to predict sound scattering, which helps suggest improvements to the detector. Finally, the work uses a subset of calls (recorded over a year) to estimate the probability of detection. The new method has higher detection rates than results previously published in the literature.