Speaker

Prof. Antonia Papandreou-Suppappola

Title

"Analysis of Cetacean Mammal Sounds Using Time-Frequency Signal Processing Techniques"

Abstract

Over the ages, humans have always been fascinated by cetacean mammals such as whales and dolphins: from show performing dolphins in aquariums, to popular dolphin TV shows, to unfounded myths that dolphins are the smartest of animals, with an intelligence approaching that of humans. In order to gain some insight into these amazing creatures, we study their sounds using time-varying signal processing techniques. We consider two main sounds: (i) short broadband clicks which are used by dolphins for echolocation, and (ii) whistles which are continuous, frequency-modulated, narrowband signals used by dolphins primarily for communication. We use quadratic time-frequency representations (QTFRs) such as the Wigner distribution, the spectrogram, and the Altes Q-distribution to analyze the time-frequency structure of these sounds. We determine the nature and properties of the signals, and we investigate which types of QTFRs are best suited for analyzing them. We investigate the use of synthetic signals like hyperbolic frequency-modulated (FM) signals as approximations to cetacean mammal whistles. We estimate the parameters of the signal models using maximum likelihood estimation techniques, and demonstrate with examples that these signals have time-frequency structures that match those of dolphin whistles. We further study the applicability of mammal sounds towards the problem of echo ranging of acoustic targets. Since cetacean mammals have evolved a biological active sonar system as an effective method of hunting and hence survival, it is highly likely that there are some intrinsic properties of the cetacean mammal sounds which make them effective for echo ranging. As our analysis tool, we use the ambiguity function to show that hyperbolic FM signals are effective in determining the range of a target.