Postgraduate research opportunities Neuromorphic sensor signal processing for spatio-temporal tracking in low SNR

Neuromorphic (NM) Engineering provides an attractive alternative to conventional sensing and processing, with the key characteristic being that NM systems generate and propagate spikes as means of processing data. The information is encoded in the timing and rate of spikes, generated by each neuron in the network once a set of incoming stimuli crosses an inner threshold within the neuron. This type of neural network is known as a spiking neural network (SNN). On the sensing side, NM (or event-based) sensors have matured over recent years, with vision sensors becoming particularly popular. Event-based sensing is done typically through change detection, where a large enough change in the signal causes the sensor to output spikes. Another very desirable feature of NM sensors is their high dynamic range. This allows event-based cameras to see in a wide variety of lighting conditions, from quickly changing brightness conditions, to low light ones, where traditional cameras would not be able to detect anything.

The goal of this research is to create an end-to-end neuromorphic processing pipeline that uses SNN-based techniques to detect, identify and track low contrast targets in high noise settings. The high noise (low SNR) environments coupled with a low contrast present a highly challenging problem space. This proposal looks at exploiting jointly neuromorphic sensing and processing techniques to further extend current capabilities for any application environments with inherent low SNR.

An initial non-exhaustive list of possible tasks for this research is:

• Identify suitable sensor modalities to investigate, to incorporate in the processing pipeline; for example radar, lidar, frame-based visual-band and IR cameras, event-based cameras.
• Identify suitable AI applications and tasks that incorporate detection and identification of targets in low SNR.
• Define appropriate and relevant spiking conversion techniques for all non-event-based sensor modalities selected.
• Devise a testing regime for each conversion technique defined.
• Devise the best methods to combine the separate spiking streams coming from each sensor modality.
• Design suitable SNNs for detection and tracking of targets in low SNR, to address the AI application/tasks selected.
• Devise a testing regime for the overall processing pipelines.

This research will be carried out through two parallel PhD programmes of 36 months each, starting in January 2024. The two PhD students will be part of the Neuromorphic Sensor Signal Processing (NSSP) Lab led by Dr Gaetano Di Caterina, in the Centre for Signal and Image Processing (CeSIP) Group, in the Electronic and Electrical Engineering Department at the University of Strathclyde, and also supported by PostDoc researchers in the Lab and other academics in the wider CeSIP Group.

As PhD researchers at Strathclyde, the students will be automatically enrolled in the Researcher Development Programme which supports PGR students in their continued personal and professional development, with the additional opportunity of gaining a PG Certificate in Researcher Professional Development.