[2025] SAR ADC-Based Time-Domain Feature Extraction for Implantable Neural Interfaces

Closed-loop neuromodulation facilitates precise and personalized treatment for neurological disorders that currently have no cure, including epilepsy, treatment-resistant depression, and Parkinson’s disease (PD). A typical closed-loop system comprises an analog frontend (AFE) for signal acquisition and digitization, followed by a feature extraction unit (FEU) that derives informative features from the raw neural data. Subsequently, a classifier uses these extracted features for brain-state classidication, while a stimulator performs neuromodulation based on the classifier's decisions. To enable implantable closed-loop neuromodulation systems, a low-power and compact feature extraction unit is essential. Time domain features such as Maximum, Minimum, Average, and Line Length have proven to be both hardware-friendly and effective for seizure detection. Traditionally, the process of computing these features involves analog-to-digital conversion followed by feature extraction in the digital domain.

Assignment

This project aims to adapt SAR (Successive Approximation Register) algorithms to allow for direct time domain feature extraction during the analog-to-digital conversion stage.
The expected assignments within this project include, but are not limited to, the following tasks:
1. Modelling the proposed approach and validating the performance in system level.
2. Comparing the conventional method with the proposed approach in terms of power
consumption, area efficiency, and accuracy performance.
3. Designing the proposed approach at both the schematic and layout levels.

Requirements

MSc student in Microelectronics with an interest in Analog/mixed signal IC design and signal processing.

Required courses:

• Analog Circuit Design Fundamentals (EE4C10)

• Analog CMOS Design I (ET4295)]

• Nyquist-Rate Data Converters (ET4369)

• Analog IC Design (ET4252)

Contact

Last modified: 2025-03-10