Motor imagery EEG classification algorithm based on CNN-LSTM feature fusion network

Li, Hongli; Ding, Ming; Xiu, Chunbo

doi:10.1016/j.bspc.2021.103342

Cited by 91 publications

(38 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An 85% classification accuracy was obtained. Hongli Li et al 32 proposed a network model with CNN and LSTM in parallel and fused the extracted middle layer features of the convolutional layer with the straightened features to improve the classification accuracy by enhancing the discriminative nature of the feature vectors input to the fully connected layer. Their method obtained an accuracy of 87.68% for the classification task of four classes of motor imagery EEG signals.…”

Section: Experimental Results and Analysismentioning

confidence: 99%

A motor imagery EEG classification algorithm based on ResCNN-BiGRU

Hu,

Wu,

Rao

et al. 2023

Optical Design and Testing XIII

View full text Add to dashboard Cite

In recent years, deep learning-based methods for motor imagery EEG classification have become increasingly popular in the field of brain-computer interfaces. However, most of the studies tend to use sequence-structured classification networks to extract spatial features when dealing with motor imagery EEG signal classification tasks, ignoring the fact that EEG signals as time-series signals contain rich temporal information and features between neural network layers, resulting in poor classification performance. Therefore, this paper proposes a feature fusion network called ResCNN-BiGRU, which consists of ResNet-based residual convolutional neural network (ResCNN) and bidirectional gated recurrent unit (BiGRU) connected in parallel. The two branches use different forms of EEG signal feature representation as input, the input to the ResCNN branch is a wavelet transformed time-frequency image, and the input to the BiGRU branch is EEG data in a two-dimensional matrix format. ResCNN extracts spatial features and utilizes interlayer features through residual connections. It also introduces convolutional block attention module (CBAM) to avoid introducing too much useless low-level feature information during interlayer feature fusion. BiGRU extracts temporal features. Finally, experiments are conducted on the authoritative four-category motor imagery dataset BCI Competition IV 2a to verify the performance of the proposed algorithm.

show abstract

Section: Experimental Results and Analysismentioning

confidence: 99%

A motor imagery EEG classification algorithm based on ResCNN-BiGRU

Hu,

Wu,

Rao

et al. 2023

Optical Design and Testing XIII

View full text Add to dashboard Cite

show abstract

“…The CNN-LSTM neural network is a parallel architecture that combines the CNN and LSTM, and the combination of the two is used to optimise the network structure [32][33][34]. The former can reduce the computational cost, filter the correlation between features and be good at processing a large amount of image information.…”

Section: Cnn-lstm Networkmentioning

confidence: 99%

Design of an embedded machine vision system for smart cameras

Zhu¹,

Li²,

Cai³

et al. 2022

Applied Mathematics and Nonlinear Sciences

View full text Add to dashboard Cite

With the rapid increase in computer users’ requirements for image information and image processing, and the rapid development of the intelligent process, the ability of the traditional visual system to process image information and data has been difficult to meet the needs of users. Therefore, in this article, we upgrade the vision system of smart cameras by introducing three network algorithm structures: convolutional neural network (CNN), LSTM and CNN-LSTM. We compare the classification performance of the three algorithms and evaluate them with three metrics: accuracy, precision and recall. The experimental results show that using the CNN algorithm, the accuracy of image information processing is 98.2%, the precision can reach 87.5% and the recall rate is 99.8%; the LSTM accuracy is 97.7%, its precision is 89.6% and its recall rate is 87.3%; its precision can be improved to 90.5% and the recall rate to 99.7%.

show abstract

“…Their proposed method achieved 79.9% for the BCI Competition IV-2a dataset. Li et al [25] proposed a feature fusion algorithm for neural networks that combines CNN and LSTM networks and connects them in parallel. The spatial and temporal features are extracted by the CNN and the LSTM networks, respectively.…”

Section: Introductionmentioning

confidence: 99%

Multi frequency band fusion method for EEG signal classification

Wahdow

Alnaanah

Fadel

et al. 2022

SIViP

View full text Add to dashboard Cite

This paper proposes a novel convolutional neural network (CNN) fusion method for electroencephalography (EEG) motor imagery (MI) signal classification. The method is named MFBF, which stands for multifrequency band fusion. The MFBF method relies on filtering the input signal with different frequency bands and feeding each band signal to a duplicate of a CNN model; then, all duplicates are concatenated to form a fusion model. This paper also introduces the second release of Coleeg software, which is used for evaluation. The MFBF method has the advantage of the flexibility of choosing any model and any number of frequency bands. In the experimental evaluation, the CNN1D model and three frequency bands were used to form the CNN1D_MFBF model, and it was evaluated against the EEGNet_fusion model on three different datasets, which are: Physionet, BCI competition IV-2a, and a dataset from the Hungarian Academy of Sciences Research Centre for Natural Sciences (MTA-TTK). The CNN1D_MFBF model had comparable or better accuracy results with less than one-fifth of the training time, which is a significant advantage for the proposed method.

show abstract

Motor imagery EEG classification algorithm based on CNN-LSTM feature fusion network

Cited by 91 publications

References 23 publications

A motor imagery EEG classification algorithm based on ResCNN-BiGRU

A motor imagery EEG classification algorithm based on ResCNN-BiGRU

Design of an embedded machine vision system for smart cameras

Multi frequency band fusion method for EEG signal classification

Contact Info

Product

Resources

About