Skip to main content

Cristiano MERIO

Using artificial intelligence for near field communication and low-energy signal demodulation


Keywords: NFC, SNN, AI, Low Power, Demodulation, HLS

Abstract: Standard digital solutions for demodulating signals uses Nyquist’s analog-to-digital converters (based on the Shannon theorem). In order to reduce the power consumption at least by a factor of 10, new approaches need to be explored. This thesis will use event-based techniques and a sparse analog-to-digital conversion in order to reduce the computational load. Such an approach is particularly efficient when coupled to event-driven circuits, also known as asynchronous circuits.
In order to smartly demodulate RF and especially NFC signals, the PhD thesis will target a low-power system based on a microcontroller unit and a dedicated hardware accelerator. This latter will be based on a neural network. This solution will be devised thanks to a High Level Synthesis (HLS) approach to evaluate several neural network configurations. Moreover, the neural network inputs will benefit from a sparse sampling and an asynchronous implementation. Such a challenging approach helps for drastically reducing the processed samples and, thus, the system power consumption.
The PhD will contribute to setting up a design framework for event-based circuitry and advanced technology. Case studies will be implemented as a demonstration of the technology.
The main areas of focus will be the following:
- Asynchronous circuit design
- Non-uniform sampling and processing
- Formal models and methods for CAD automation (HLS)
- Signal demodulation based on AI and NN accelerator
- Application to NFC test cases and low-power circuits
Thanks to this, the PhD student will propose some innovative solutions based on event-based strategies applied to NFC communications. Then, the relevant solutions and models will be implemented in a CAD framework and IPs will be investigated more thoroughly at silicon level. The PhD work will target a demodulation digital block based on artificial intelligence and event-based techniques in order to improve energy efficiency.


Thesis director: Laurent FESQUET (TIMA - CDSI)
Thesis supervisor: Sylvain ENGELS (TIMA - CDSI / STMicroelectronics)
Thesis started on: Oct. 2022
Doctoral school: EEATS

Submitted on November 29, 2022

Updated on December 12, 2023