lu.se

Matematikcentrum

Lunds universitet

Denna sida på svenska This page in English

Kalendarium

Bio-modelling Seminar: Time-frequency localization of marine biosonar signals

Seminarium

Tid: 2018-04-11 09:15 till: 10:00
Plats: Konferensrummet at floor 1 in the Ecology building
Kontakt: sara [at] maths [dot] lth [dot] se
Spara händelsen till din kalender


Maria Sandsten will talk at the next biomodelling seminar April 11 at 9:15-10:00 (see abstract below). The venue is Konferensrummet at the first floor in the Ecology building. As usual there will be fika!

Transient signals are by nature difficult to characterize, compared to longer signals. The short duration of the pulses implies that the uncertainty in frequency is high. Still, these signals are essential in fields such as ultrasonic analysis where pulse reflections are close in the time-frequency (TF) domain, and for the ultrasonic biosonar analysis of several toothed whale species, where the measured signals are only a few periods long. The sonar beam of toothed whales contains signal components from various acoustic pathways inside the animals’ forehead. These pathways affect the emitted signal in different directions. To accurately detect and localize the components in the TF domain are essential to understand to what extent the signal can be controlled by the animal and what functions it serves.

 

 A technique to improve the localization of single TF components and enhance the readability of the TF representation of multi-component signals is TF reassignment. The method reassigns signal energy to the center of gravity, giving higher energy concentration at the actual frequencies of the signal. We have invented the scaled reassigned spectrogram, tailored to very short transients, which goes beyond the lower bound of the Gabor transform and the reassignment of longer lasting transients. This reassignment technique finds the TF centers of individual signal components by utilizing that many transients have a Gaussian like shape in time and is therefore well suited for detection and localization of transient signal components, also when they are closely located in the TF domain.