EEG-TCNet: An Accurate Temporal Convolutional Network for Embedded Motor-Imagery Brain–Machine Interfaces

Ingolfsson, Thorir Mar; Hersche, Michael; Wang, Xiaying; Kobayashi, Nobuaki; Cavigelli, Lukas; Benini, Luca

doi:10.1109/smc42975.2020.9283028

Cited by 141 publications

(88 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 3 presents the classification accuracy and kappa scores of each subject for several state-of-the-art MI-EEG algorithms employing a subject-specific approach on the BCI Competition IV-2a dataset. The suggested MBEEGNET and MBShallowConvNet, EEG-TCNet [ 30 ], both fixed and variable EEGNet [ 16 , 30 ], ShallowConvNet [ 16 ], and Incep-EEGNet [ 29 ] are the approaches compared. MBEEGNET and MBShallowConvNet, the proposed models, have an accuracy of 82.01% and 81.15%, respectively.…”

Section: Resultsmentioning

confidence: 99%

“…In [ 30 ], the authors used temporal convolutional networks (TCNs) with EEGnet to boost the performance accuracy. A standard causal convolution can only expand the size of its receptive field linearly as the network depth increases, which can be a significant drawback because a big receptive field size requires either an exceptionally deep network or one with a very large kernel size.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Multibranch of Convolutional Neural Network Models for Electroencephalogram-Based Motor Imagery Classification

Altuwaijri

Muhammad

2022

Biosensors

View full text Add to dashboard Cite

Automatic high-level feature extraction has become a possibility with the advancement of deep learning, and it has been used to optimize efficiency. Recently, classification methods for convolutional neural network (CNN)-based electroencephalography (EEG) motor imagery have been proposed, and have achieved reasonably high classification accuracy. These approaches, however, use the CNN single convolution scale, whereas the best convolution scale varies from subject to subject. This limits the precision of classification. This paper proposes multibranch CNN models to address this issue by effectively extracting the spatial and temporal features from raw EEG data, where the branches correspond to different filter kernel sizes. The proposed method’s promising performance is demonstrated by experimental results on two public datasets, the BCI Competition IV 2a dataset and the High Gamma Dataset (HGD). The results of the technique show a 9.61% improvement in the classification accuracy of multibranch EEGNet (MBEEGNet) from the fixed one-branch EEGNet model, and 2.95% from the variable EEGNet model. In addition, the multibranch ShallowConvNet (MBShallowConvNet) improved the accuracy of a single-scale network by 6.84%. The proposed models outperformed other state-of-the-art EEG motor imagery classification methods.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Multibranch of Convolutional Neural Network Models for Electroencephalogram-Based Motor Imagery Classification

Altuwaijri

Muhammad

2022

Biosensors

View full text Add to dashboard Cite

show abstract

“…Hence, researching compact yet accurate algorithms [7], [17] and designing low-power processors with high capabilities [18] has become an emerging trend. Most of the MI-BMI models, particularly CNNs, are too demanding for low-power MCUs [19]. TPCT [20] reached the state-of-the-art (SoA) accuracy of 88.87% on the BCI Competition IV-2a dataset [21].…”

Section: Introductionmentioning

confidence: 99%

Mixed-Precision Quantization and Parallel Implementation of Multispectral Riemannian Classification for Brain-Machine Interfaces

Wang

Schneider

Hersche

et al. 2021

2021 IEEE International Symposium on Circuits and Systems (ISCAS)

Self Cite

View full text Add to dashboard Cite

With Motor-Imagery (MI) Brain-Machine Interfaces (BMIs) we may control machines by merely thinking of performing a motor action. Practical use cases require a wearable solution where the classification of the brain signals is done locally near the sensor using machine learning models embedded on energy-efficient microcontroller units (MCUs), for assured privacy, user comfort, and long-term usage. In this work, we provide practical insights on the accuracy-cost tradeoff for embedded BMI solutions. Our proposed Multispectral Riemannian Classifier reaches 75.1% accuracy on 4-class MI task. We further scale down the model by quantizing it to mixed-precision representations with a minimal accuracy loss of 1%, which is still 3.2% more accurate than the state-of-theart embedded convolutional neural network. We implement the model on a low-power MCU with parallel processing units taking only 33.39 ms and consuming 1.304 mJ per classification.Index Terms-brain-machine interface, edge computing, parallel computing, machine learning, deep learning, motor imagery.

show abstract

“…Deep neural networks (DNNs) have demonstrated impressive results in many fields, from outperforming humans in computer vision [17], vastly improving solutions for image processing [18] and speech recognition [19], natural language processing [20], large-scale recommender systems [21], and data analysis for sensors such as radar [22] that are not directly understandable by humans. Especially convolutional neural networks (CNNs) are largely applied in the image domain, but also for MI-BMI classification achieving state-of-the-art (SoA) accuracy [23]- [26]. However, they tend to grow in numbers of parameters making them often unfit for deployment on low-power low-cost MCUs, and require a large amount of training data to prevent overfitting.…”

Section: Introductionmentioning

confidence: 99%

“…1 https://github.com/pulp-platform/multispectral-riemannian Q-EEGNet [13] EEGNet [24] S. ConvNet [23] MSFBCNN [25] EEG-TCNet [26] FBCSP [40] MRC-Mr. Wolf [39] MRC-Vega…”

Section: Introductionmentioning

confidence: 99%

Sub-100uW Multispectral Riemannian Classification for EEG-based Brain--Machine Interfaces

Wang¹,

Cavigelli²,

Schneider³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

motor imagery (MI) brain-machine interfaces (BMIs) enable us to control machines by merely thinking of performing a motor action. Practical use cases require a wearable solution where the classification of the brain signals is done locally near the sensor using machine learning models embedded on energy-efficient microcontroller units (MCUs), for assured privacy, user comfort, and long-term usage. In this work, we provide practical insights on the accuracy-cost trade-off for embedded BMI solutions. Our multispectral Riemannian classifier reaches 75.1% accuracy on a 4-class MI task. The accuracy is further improved by tuning different types of classifiers to each subject, achieving 76.4%. We further scale down the model by quantizing it to mixed-precision representations with a minimal accuracy loss of 1% and 1.4%, respectively, which is still up to 4.1% more accurate than the state-of-the-art embedded convolutional neural network. We implement the model on a low-power MCU within an energy budget of merely 198µJ and taking only 16.9 ms per classification. Classifying samples continuously, overlapping the 3.5 s samples by 50% to avoid missing user inputs allows for operation at just 85µW. Compared to related works in embedded MI-BMIs, our solution sets the new state-of-the-art in terms of accuracy-energy trade-off for near-sensor classification.

show abstract

EEG-TCNet: An Accurate Temporal Convolutional Network for Embedded Motor-Imagery Brain–Machine Interfaces

Cited by 141 publications

References 25 publications

A Multibranch of Convolutional Neural Network Models for Electroencephalogram-Based Motor Imagery Classification

A Multibranch of Convolutional Neural Network Models for Electroencephalogram-Based Motor Imagery Classification

Mixed-Precision Quantization and Parallel Implementation of Multispectral Riemannian Classification for Brain-Machine Interfaces

Sub-100uW Multispectral Riemannian Classification for EEG-based Brain--Machine Interfaces

Contact Info

Product

Resources

About