Description of the thesis topic

The advent of intelligent and autonomous audio sensor architectures opens up the possibility of conducting extended statistical observation missions spanning multiple years. These sensors find wide-ranging applications, particularly in fields such as eco-acoustics and urban noise monitoring.

These sensors employ supercapacitors rather than traditional batteries, constituting maintenance-free power systems that rely on ambient energy. This configuration may lead to intermittent power availability. Consequently, the system must be able to partition processes effectively to capitalize on moments of energy availability. Moreover, the energy harvesting itself is constrained by the environment, which has its own cycle, such the alternation between day and night.

These sensors possess the capability to process data locally, thereby reducing its size and facilitating transmission over a low-power wide area network (LPWAN). This approach mitigates the risk of storage saturation, a significant impediment to autonomy.

The goal of the thesis work is to take into account these two aspects in order to propose optimized processing algorithms for long-term statistical audio observation. Specifically, the focus lies within the domain of eco-acoustics, with the aim of developing an algorithm tailored for on-sensor bird song classification.

Work Context

The PhD student must hold an MSc in computer science, embedded systems, statistics, or AI; or hold an engineering degree in one of these areas. Below is the expected profile:
1. Scientific curiosity is paramount.
2. An ability to criticize, advance, and transmit the state of the art is required. Some experience in public outreach is useful but not required.
3. A mastery of scientific English, both written and spoken, is required. Fluency in French is useful but not required.
4. Basic knowledge in digital signal processing, e.g., the Fourier transform, is required. Applied knowledge in bioacoustic signal processing is useful but not required.
5. Experience in data science, ideally in speech or audio processing, is required. Experience in deep neural networks is useful but not required.
6. An ability to program in Python, to use a command line, and to use version control (e.g., git), is required. Experience in high-performance computing (GPU) or parallel computing is useful but not required.
7. An ability to program in C is required. Experience with embedded systems (e.g., microcontrollers, Raspberry Pi) is useful but not required.

The PhD student will be a member of the Laboratoire des Sciences du Numérique de Nantes (LS2N), a mixed research unit whose components are the CNRS, Nantes Université, the École Centrale de Nantes, the IMT Atlantique, and Inria. Visit: https://www.ls2n.fr/

At LS2N, the PhD student will be a member of two steams : "Signal, Image et Son" (SIMS) et "Systèmes Temps Réel" (STR). Voir : https://sims.ls2n.fr/

The PhD will be registered to the doctoral school "Sciences de l'ingénierie et des systèmes". Visit: https://ed-sis.doctorat-paysdelaloire.fr/

The PhD student will be supervised by Pierre-Emmanuel Hladik, Sébastien Faucou, and Vincent Lostanlen. Pierre-Emmanuel Hladik will be the main PhD advisor (HDR).
This PhD program belongs to the ANR project "Operating Within Limits" (OWL), involving partners in Nantes, Rennes, and Lannion.

Here is a list of relevant conferences for presenting the results of the PhD: (in alphabetical order)
- EUSIPCO
- GRETSI
- IEEE ICASSP
- IEEE Internet of Things
- IEEE WASPAA
- IEEE/ACM EMSOFT

Journals:
- EURASIP JASMP
- IEEE Sensors
- IEEE TPAMI
- Journal of System Architectures

The PhD student will have a shared office at École Centrale de Nantes. They will have access to a work station and computer equipment, as well as facilities for high-performance computing and embedded computing.

Constraints and risks

This PhD does not involve isolated work, intense physical effort or manipulation of robotic machines.

This PhD involves a significant amount of work on screen, hence well-understood professional risks: mainly musculoskeletal disorders, lower back pain, visual fatigue, and stress. Against these risks, we propose to act on the layout of the workstation, on the choice of equipment and on the organization of work. Read in particular: https://www.inrs.fr/risks/travail-ecran/ce-qu-il-faut-retenir.html

The doctoral student will not be exposed to toxic products, pathogens, noise, vibrations, dangerous radiation, nor high voltage electricity.

The doctoral student is expected to go on a mission to France or internationally two to three times a year. The dates and destinations remain to be negotiated depending on the personal constraints of the doctoral student.

The risks of injury from falling, fire or electrocution are low, and subject to control by the laboratory's prevention assistants.