Wireless biomedical sensors should dramatically reduce the costs and risks associated with personal health care while being more and more exploited by telemedicine and efficient e-health systems.

However, because of the large power consumption of continuous wireless transmission, the battery life of the sensors is reduced for long-term use. Sub-Nyquist continuous-time discrete-amplitude (CTDA) sampling approaches using level-crossing analogto-digital converters (ADCs) have been developed to reduce the sampling rate and energy consumption of the sensors. However, traditional machine learning techniques and architectures are not compatible with the non-uniform sampled data obtained from levelcrossing ADCs.

This project aims to develop analog algorithms, circuits, and systems for the implementation of machine learning techniques in CTDA sampled data in wireless biomedical sensors. This “near-sensor computing” approach, will help reduce the wireless transmission rate and therefore the power consumption of the sensor. The output rate of the CTDA is directly proportional to the activity of the analog signal at the input of the sensor. Therefore, artificial intelligence hardware that processes CTDA data should consume significantly less energy.

For demonstration purposes, a prototype biomedical sensor for the detection and classification of sleep apnea will be developed using integrated circuit prototypes and a commercially available analog front-end interface. The sensor will acquire electrocardiogram and bioimpedance signals from the subject and will use data fusion techniques and machine learning techniques to achieve high accuracy.

Project partners

• Institut d'électronique, de microélectronique et de nanotechnologie - France
• Tallinn University of Technology - Estonia
• University College Dublin -  Ireland

Main results

The significant recent results of the project are:

• The proposition of a non-uniformly sampled process and evaluation of level-crossing ADC, the creation of a non-uniformly sampled ECG database, and the definition of an architecture based on LC-ADC and artificial neural network suitable for the classification of cardiac arrhythmia [IEEE ISCAS 2021 and IEEE TBioCAS 2022].
• The demonstration of event-driven slope-based feature extraction [IEEE BioCAS 2021] and antidictionary-based classification [IEEE ISCAS 2022] architectures.
• The creation of an ICG database and study of the electrode locations with a focus on wrist-based systems [IEEE EMBC 2021]
• The evaluation of ICG feature point extraction and possible architectures to extract hemodynamic parameters [IEEE SOCC 2021 and IEEE I2MTC 2022]
• The proposition of architectures for energy-efficient embedded sleep apnea detection
• The development of an embedded evaluation and test platform for joint ICG ECG (PPG) measurements