Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Ultra-Low Power Wearable BMI System with Continual Learning Capabilities

Lan Mei, Thorir Mar Ingolfsson, Cristian Cioflan, Victor Kartsch, Andrea Cossettini, Xiaying Wang, Luca Benini

TL;DR

This work proposes a comprehensive BMI workflow based on a CNN-based Continual Learning (CL) framework, allowing the system to adapt to inter-session changes and meets users’ needs for improved privacy, reduced latency, and enhanced inter-session performance.

Abstract

Driven by the progress in efficient embedded processing, there is an accelerating trend toward running machine learning models directly on wearable Brain-Machine Interfaces (BMIs) to improve portability and privacy and maximize battery life. However, achieving low latency and high classification performance remains challenging due to the inherent variability of electroencephalographic (EEG) signals across sessions and the limited onboard resources. This work proposes a comprehensive BMI workflow based on a CNN-based Continual Learning (CL) framework, allowing the system to adapt to inter-session changes. The workflow is deployed on a wearable, parallel ultra-low power BMI platform (BioGAP). Our results based on two in-house datasets, Dataset A and Dataset B, show that the CL workflow improves average accuracy by up to 30.36% and 10.17%, respectively. Furthermore, when implementing the continual learning on a Parallel Ultra-Low Power (PULP) microcontroller (GAP9), it achieves an energy consumption as low as 0.45mJ per inference and an adaptation time of only 21.5ms, yielding around 25h of battery life with a small 100mAh, 3.7V battery on BioGAP. Our setup, coupled with the compact CNN model and on-device CL capabilities, meets users' needs for improved privacy, reduced latency, and enhanced inter-session performance, offering good promise for smart embedded real-world BMIs.

An Ultra-Low Power Wearable BMI System with Continual Learning Capabilities

TL;DR

This work proposes a comprehensive BMI workflow based on a CNN-based Continual Learning (CL) framework, allowing the system to adapt to inter-session changes and meets users’ needs for improved privacy, reduced latency, and enhanced inter-session performance.

Abstract

Driven by the progress in efficient embedded processing, there is an accelerating trend toward running machine learning models directly on wearable Brain-Machine Interfaces (BMIs) to improve portability and privacy and maximize battery life. However, achieving low latency and high classification performance remains challenging due to the inherent variability of electroencephalographic (EEG) signals across sessions and the limited onboard resources. This work proposes a comprehensive BMI workflow based on a CNN-based Continual Learning (CL) framework, allowing the system to adapt to inter-session changes. The workflow is deployed on a wearable, parallel ultra-low power BMI platform (BioGAP). Our results based on two in-house datasets, Dataset A and Dataset B, show that the CL workflow improves average accuracy by up to 30.36% and 10.17%, respectively. Furthermore, when implementing the continual learning on a Parallel Ultra-Low Power (PULP) microcontroller (GAP9), it achieves an energy consumption as low as 0.45mJ per inference and an adaptation time of only 21.5ms, yielding around 25h of battery life with a small 100mAh, 3.7V battery on BioGAP. Our setup, coupled with the compact CNN model and on-device CL capabilities, meets users' needs for improved privacy, reduced latency, and enhanced inter-session performance, offering good promise for smart embedded real-world BMIs.
Paper Structure (30 sections, 8 figures, 5 tables)

This paper contains 30 sections, 8 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Acquisition Platforms
  4. Continual and Transfer Learning for Data
  5. Online On-Device Learning for Biosignals
  6. Materials and Methods
  7. Acquisition Device
  8. BioGAP
  9. Headband
  10. Electrodes
  11. Dataset Description
  12. Preprocessing and Classification
  13. Preprocessing
  14. Classification
  15. Inter-session Continual Learning
  16. ...and 15 more sections

Figures (8)

  • Figure 1: Top: data acquisition setup, comprising the headband, BioGAPfrey2023biogap, and the dry active electrodes. Bottom: timing scheme of data trials.
  • Figure 2: The network architecture MI-BMINet wang2022mi used in this work, selective layer update is shown for the $\theta_1$ part of the network, this part of the network is fine-tuned in floating precision, while others are kept in int8 format.
  • Figure 3: Inter-session learning workflows for systems.
  • Figure 4: Performance across all sessions when progressively adding more sessions, comparing TL and CL in 2-class MM classification on Dataset A.
  • Figure 5: Performance across all sessions when progressively adding more sessions, comparing TL and CL in 2-class MM classification on Dataset B.
  • ...and 3 more figures