Highly Efficient Brain-Machine Edge Devices Using Deep Learning for IoT

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K12789
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 90150:Medical assistive technology-related
• Research Institution: Nihon University
• Principal Investigator: KOBAYASHI Nobuaki 日本大学, 理工学部, 准教授 (50611422)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: EEG / Brain-machine interface / Brain-computer interface / Edge computing / IoT

Outline of Final Research Achievements

Brain-machine interface (BMI) technology that enables carers to actively control machines using electroencephalography (EEG) is expected to improve the comfort of people requiring care in their living environment and reduce the burden on care workers. The main focus of this study is on EEG during motor imagery (MI) and on improving the processing efficiency of MI classification using deep learning, which requires a large amount of data for training, increasing the burden on carers. Therefore, the weights of previously trained models were re-trained and fine-tuned to achieve high classification accuracy and reduce the amount of data required for training. As a result, a BMI that can be used for long periods of time and with minimal response delay has been achieved.

Free Research Field

brain-machine interface, digital circuit

Academic Significance and Societal Importance of the Research Achievements

要介護者のQoLの向上、また、介護職従事者の仕事量を減少することを目的に、実用化に向けたBMI型IoTデバイスの構築が目的である。このためには、「脳活動の高い識別精度」、「居住環境ノイズ下での使用に耐え得る」、「長時間使用によるユーザの使用負担を軽減する」、「遅延量を最小化する」、「消費電力を低減し使用時間を長くする」、といった要件を同時に満たす必要がある。これまでにない低遅延性・低消費電力性を兼ね備える高効率なBMI型IoTデバイスを開発・実証することで、BMIの介護分野における普及、IoT分野の進展に大きく貢献することが期待される。