Convolutional Neural Networks with Reusable Full-Dimension-Long Layers for Feature Selection and Classification of Motor Imagery in EEG Signals

Tokovarov, Mikhail

doi:10.1007/978-3-030-61609-0_7

Cited by 5 publications

(7 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That is, fixing the sliding short-time window length parameter τ =2 s with an overlapping step of 1 s, resulting in N τ =5 EEG segments. For implementing the filter bank strategy, the following bandwidths of interest: ∆ f ∈{µ∈ [8][9][10][11][12], β∈ [12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30]} Hz. These bandwidths belong to µ, and β rhythms, commonly associated with electrical brain activities provoked by MI tasks [53].…”

Section: Preprocessing and Feature Extraction Of Image-based Represen...mentioning

confidence: 99%

“…As a solution for ineffective discriminability in decoding MI tasks using EEGs, Deep Learning (DL) algorithms have increasingly been applied to boost the classification accuracy of subject-independent classifiers. Among DL models, convolutional neural networks (CNN) with kernels that share weights for multidimensional planes have achieved outstanding success in extracting, directly from raw EEG data, unlocked local/general patterns over different domain combinations like time, space, and frequency [13][14][15][16][17][18]. The earlier layers learn low-level features, while the deeper layers learn high-level representations.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Image-Based Learning Using Gradient Class Activation Maps for Enhanced Physiological Interpretability of Motor Imagery Skills

Collazos-Huertas

Álvarez-Meza

Castellanos-Domínguez

2022

Applied Sciences

View full text Add to dashboard Cite

Brain activity stimulated by the motor imagery paradigm (MI) is measured by Electroencephalography (EEG), which has several advantages to be implemented with the widely used Brain–Computer Interfaces (BCIs) technology. However, the substantial inter/intra variability of recorded data significantly influences individual skills on the achieved performance. This study explores the ability to distinguish between MI tasks and the interpretability of the brain’s ability to produce elicited mental responses with improved accuracy. We develop a Deep and Wide Convolutional Neuronal Network fed by a set of topoplots extracted from the multichannel EEG data. Further, we perform a visualization technique based on gradient-based class activation maps (namely, GradCam++) at different intervals along the MI paradigm timeline to account for intra-subject variability in neural responses over time. We also cluster the dynamic spatial representation of the extracted maps across the subject set to come to a deeper understanding of MI-BCI coordination skills. According to the results obtained from the evaluated GigaScience Database of motor-evoked potentials, the developed approach enhances the physiological explanation of motor imagery in aspects such as neural synchronization between rhythms, brain lateralization, and the ability to predict the MI onset responses and their evolution during training sessions.

show abstract

Section: Preprocessing and Feature Extraction Of Image-based Represen...mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Image-Based Learning Using Gradient Class Activation Maps for Enhanced Physiological Interpretability of Motor Imagery Skills

Collazos-Huertas

Álvarez-Meza

Castellanos-Domínguez

2022

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…The authors in [34] combine CSP with Riemannian distances obtaining a classification accuracy of 77.82% and a kappa value of 0.7086 with an average of 15.2 selected channels over the nine subjects of the IV-2a dataset for the 4-class MI task. CNN-based channel reduction approaches are also found in the literature [37], [28]. Dose et al [37] manually select subsets of EEG channels to compare with related works and demonstrate that their proposed architecture based on shallow ConvNet outperforms most of the other models.…”

Section: Related Workmentioning

confidence: 99%

“…Dose et al [37] manually select subsets of EEG channels to compare with related works and demonstrate that their proposed architecture based on shallow ConvNet outperforms most of the other models. More recent work by Tokovarov et al [28] applies an automatic channel selection method based on CNN feature maps obtaining 82.34% accuracy with only 14 instead of 64 channels on the 2-class MI task of the MM/MI dataset outperforming the manual selection of [37]. Given the many advantages of channel selection, we propose in this work an automated method based on the spatial filters of the proposed CNN.…”

Section: Related Workmentioning

confidence: 99%

“…However, a reduced number of channels makes the task even more difficult, often causing accuracy degradation [24]. Hence, sophisticated methods for channel selection are necessary to extract relevant features for the task while maintaining accurate performance [26], [27], [28]. Another commonly used technique to reduce the resource requirements is quantization, e.g., using 8 bits to represent numbers instead of 32 bits [29], [30].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

MI-BMInet: An Efficient Convolutional Neural Network for Motor Imagery Brain--Machine Interfaces with EEG Channel Selection

Wang¹,

Hersche²,

Magno³

et al. 2022

Preprint

View full text Add to dashboard Cite

A brain-machine interface (BMI) based on motor imagery (MI) enables the control of devices using brain signals while the subject imagines performing a movement. It plays an important role in prosthesis control and motor rehabilitation and is a crucial element towards the future Internet of Minds (IoM). To improve user comfort, preserve data privacy, and reduce the system's latency, a new trend in wearable BMIs is to embed algorithms on low-power microcontroller units (MCUs) to process the electroencephalographic (EEG) data in real-time close to the sensors into the wearable device. However, most of the classification models present in the literature are too resourcedemanding, making them unfit for low-power MCUs. This paper proposes an efficient convolutional neural network (CNN) for EEG-based MI classification that achieves comparable accuracy while being orders of magnitude less resource-demanding and significantly more energy-efficient than state-of-the-art (SoA) models for a long-lifetime battery operation. We propose an automatic channel selection method based on spatial filters and quantize both weights and activations to 8-bit precision to further reduce the model complexity with negligible accuracy loss. Finally, we efficiently implement and evaluate the proposed models on a parallel ultra-low power (PULP) MCU. The most energy-efficient solution consumes only 50.10 µJ with an inference runtime of 5.53 ms and an accuracy of 82.51% while using 6.4f ewer EEG channels, becoming the new SoA for embedded MI-BMI and defining a new Pareto frontier in the three-way trade-off among accuracy, resource cost, and power usage.

show abstract

Non-linear Feature Selection Based on Convolution Neural Networks with Sparse Regularization

Wu,

Chen,

Ding

et al. 2023

Cogn Comput

View full text Add to dashboard Cite

Convolutional Neural Networks with Reusable Full-Dimension-Long Layers for Feature Selection and Classification of Motor Imagery in EEG Signals

Cited by 5 publications

References 19 publications

Image-Based Learning Using Gradient Class Activation Maps for Enhanced Physiological Interpretability of Motor Imagery Skills

Image-Based Learning Using Gradient Class Activation Maps for Enhanced Physiological Interpretability of Motor Imagery Skills

MI-BMInet: An Efficient Convolutional Neural Network for Motor Imagery Brain--Machine Interfaces with EEG Channel Selection

Non-linear Feature Selection Based on Convolution Neural Networks with Sparse Regularization

Contact Info

Product

Resources

About