Reference : UMR8520-FRELEF-136
Workplace : VILLENEUVE D ASCQ
Date of publication : Tuesday, September 20, 2022
Scientific Responsible name : Dr Christophe LOYEZ et Pr François DANNEVILLE
Type of Contract : PhD Student contract / Thesis offer
Contract Period : 36 months
Start date of the thesis : 2 November 2022
Proportion of work : Full time
Remuneration : 2 135,00 € gross monthly
Description of the thesis topic
The preservation of biodiversity is fundamental for the survival of species. In the context of underwater fauna, monitoring the activity of large cetaceans (sperm whales, bottlenose dolphins, ....) is essential for multiple reasons: (i) firstly, the preservation of species, (ii) secondly, as large cetaceans are at the "end" of the food chain, their depletion results in a depletion of fisheries resources, and consequently, a direct economic impact on the fishing industry. Access to and processing of natural data (sounds emitted) from large cetaceans is also relevant for a better understanding of their individual and/or group behaviour (cognitive capacities, learning, communication between individuals) as well as for locating them in deep waters.
Large cetaceans - e.g. sperm whales - emit clicks (acoustic signatures) which they use to communicate with each other when hunting giant squid in cohorts (they also use clicks to locate their prey, thanks to a "sonar" technique). These clicks can be picked up by hydrophones and then analysed (for classification, counting of individuals, etc.). For this purpose and for energy efficiency, we propose to use spiking neural networks (SNN), designed from artificial plastic neurons and synapses (patented ultra low power neuromorphic technology (ULP), developed by IEMN / IRCICA). The input layer of these SNNs will have to receive natural data in the form of electrical pulses (spikes), which encode the information carried by the acoustic signal in frequency and time form. The study will consist of :
- A bibliographical study of artificial cochleas
- The definition of disruptive hardware architectures, compatible with the ULP neuromorphic technology mentioned above, adapted to the "clicks" of large cetaceans (particularly their bandwidth)
- The design (simulation) of monaural and/or binaural artificial cochleas
- Their coupling with an SNN for the recognition (learning) of acoustic signatures emitted by large cetaceans and/or an ad-hoc circuit allowing their geolocation.
This PhD subject is part of a larger ANR project "ULP SMART 3D COCHLEA" (partners are CRISTAL - Univ. Lille - and INPS - Univ. Toulon -) started on January 1st 2022, for a duration of 4 years. The PhD student will be hosted at IRCICA
Based in Villeneuve d'Ascq on the Haute Borne Science Park, the Research Institute on Software and Hardware Components for Advanced Information and Communication (IRCICA) is a Service and Research Unit (USR-3380) associating the CNRS and the University of Lille. For the past ten years, IRCICA has been developing interdisciplinary research to imagine and create responsible information and communication technologies. This allows researchers and teacher-researchers with complementary skills to be brought together in the same place to tackle major scientific and societal challenges with interdisciplinary and/or disruptive approaches from the embryonic stage of the projects.
IRCICA houses around 120 teacher-researchers, researchers, students, engineers and technicians with complementary skills from the four partner laboratories CRIStAL - Lille Research Centre for Computer Science, Signals and Automation, IEMN - Institute of Electronics, Microelectronics and Nanotechnologies, PhLAM - Laboratory of Laser, Atom and Molecule Physics and L2EP - Lille Laboratory of Electrotechnics and Electronics and Power
We talk about it on Twitter!