A Wake-Up Circuit for Event-Driven Duty-Cycling of Wearable IoT Sensor Nodes
Open access
Author
Date
2018Type
- Doctoral Thesis
ETH Bibliography
yes
Altmetrics
Abstract
The design of IoT sensing nodes mandate unprecedented tight constraints on power consumption, physical volume, and production costs which limits the on board processing and data-transfer capabilities. Efficient real-time processing can be achieved by exploiting system-level duty-cycled architectures and novel mixed-signal processing circuits that detect and transmit only events of interest reducing power-hungry raw data transfer.
In this dissertation we report an always-on wake-up circuit and a high-performance biomedical SoC designs and implementations that can be employed in IoT sensing nodes.
The wake-up circuit is based on a level-crossing analog-to-digital converter (LC-ADC) employed as feature-extraction block with automatic activity-sampling rate scaling behaviour. A novel asynchronous digital logic classifier for sequential pattern recognition is presented and it is supervised-learning trained to maximize events detection accuracy. It is driven by the LC-ADC activity and trained to minimize classification errors due to falsely detected events. A prototype has been first validated by interfacing it with a commercial accelerometer to classify hand gestures in real-time, reaching 81% of accuracy with only 2.2 μW at 1 V supply. To highlight the flexibility of the design, a second application, detecting pathologic ECG beats is also discussed.
Furthermore, an application based on the novel biomedical System-On-Chip (SoC) for signal acquisition and processing combining a homogeneous multi-core cluster with a versatile bio-potential front-end is described. The presented implementation acquires raw EMG signals from 3 passive gel-electrodes and classifies 3 hand gestures using a Support Vector Machine (SVM) pattern recognition algorithm. Performance matches state-of-the-art high-end systems both in terms of recognition accuracy (> 85%) and of real-time execution (gesture recognition time 300 ms), requiring only 10 mW. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000281405Publication status
publishedExternal links
Search print copy at ETH Library
Publisher
ETH ZurichSubject
Wake-up circuit; classifier; Ultra Low Power; event driven; IoT; sensing node; Wearable; CMOSOrganisational unit
02636 - Institut für Integrierte Systeme / Integrated Systems Laboratory03996 - Benini, Luca / Benini, Luca
Funding
157048 - Transient Computing Systems (SNF)
162524 - MicroLearn: Micropower Deep Learning (SNF)
More
Show all metadata
ETH Bibliography
yes
Altmetrics