A deep CNN approach to decode motor preparation of upper limbs from time–frequency maps of EEG signals at source level

Mammone, Nadia; Ieracitano, Cosimo; Morabito, Francesco Carlo

doi:10.1016/j.neunet.2020.01.027

Cited by 128 publications

(86 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The model takes in input the raw EEG data, which has been spatially re-organised in a topology-preserving representation that is able to keep the information about the spatiotemporal dependencies of the channels. We tested our approach using the same publicly available dataset used by Mammone et al [24], consisting in a high density EEG recording acquired simultaneously with motion data (described in detail in Ofner et al [14]). The major contributions of our work can are the following:…”

Section: Contributions Of This Workmentioning

confidence: 99%

“…Indeed, several recent works are focusing on the detection of motor anticipation rather than the detection of motor-imagery/execution, with envisaged applications ranging from driving [21] to the control of robotic devices [22,23]. The application of DL techniques for classifying motor planning of different movements has been recently investigated by the important work of Mammone et al [24], providing also an updated overview of relevant works about EEG-based motor anticipation. In their work, they proposed to classify 6 different movement classes, plus the rest class, by using only a sub-part of the EEG signal corresponding to the second before the movements onset.…”

Section: Introductionmentioning

confidence: 99%

“…Section 2.4 introduces the detailed architecture of the model, specifically focusing on the Encoder component in Section 2.4.1 and on the Classifier component in Section 2.4.2. The training scheme of the model is reported in Section 2.5, while in Section 2.6 we describe how the results presented by [24] have been adapted in order to allow a direct comparison with our results. Section 3 presents the empirical analysis, first by reporting the classification accuracy on the selected epoch for the different subjects (Section 3.1), then by showing the accuracy of the classifier over time around the movement onset (Section 3.2).…”

mentioning

confidence: 99%

See 2 more Smart Citations

A Deep Classifier for Upper-Limbs Motor Anticipation Tasks in an Online BCI Setting

Valenti

Barsotti²,

Bacciu

et al. 2021

Bioengineering

View full text Add to dashboard Cite

Decoding motor intentions from non-invasive brain activity monitoring is one of the most challenging aspects in the Brain Computer Interface (BCI) field. This is especially true in online settings, where classification must be performed in real-time, contextually with the user’s movements. In this work, we use a topology-preserving input representation, which is fed to a novel combination of 3D-convolutional and recurrent deep neural networks, capable of performing multi-class continual classification of subjects’ movement intentions. Our model is able to achieve a higher accuracy than a related state-of-the-art model from literature, despite being trained in a much more restrictive setting and using only a simple form of input signal preprocessing. The results suggest that deep learning models are well suited for deployment in challenging real-time BCI applications such as movement intention recognition.

show abstract

Section: Contributions Of This Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

A Deep Classifier for Upper-Limbs Motor Anticipation Tasks in an Online BCI Setting

Valenti

Barsotti²,

Bacciu

et al. 2021

Bioengineering

View full text Add to dashboard Cite

show abstract

“…A value of 1 is recommended because the input and output spatial resolution must be the same [ 61 ]. There are diverse nonlinear activation functions such as sigmoid, hyperbolic tangent, rectified linear unit (ReLu), among others, being the ReLu, f ( Y j ) = max(0, Y j ), the fastest and most effective to learn the nonlinear properties of each feature map, Y j , in a CNN [ 62 ].…”

Section: Theoretical Backgroundmentioning

confidence: 99%

“…It moves a filter of size K 1 × K 2 with a stride s 2 across the feature maps by taking the average (average pooling) or maximum (max pooling) of the neighbor values chosen by the filter. Hence, a sub-sampled representation of Y j , with a size of Z 1 × Z 2 , is obtained as follows [ 62 ]:

…”

Section: Theoretical Backgroundmentioning

confidence: 99%

Convolutional Neural Network and Motor Current Signature Analysis during the Transient State for Detection of Broken Rotor Bars in Induction Motors

Valtierra-Rodríguez

Rivera-Guillen

Basurto-Hurtado

et al. 2020

Sensors

View full text Add to dashboard Cite

Although induction motors (IMs) are robust and reliable electrical machines, they can suffer different faults due to usual operating conditions such as abrupt changes in the mechanical load, voltage, and current power quality problems, as well as due to extended operating conditions. In the literature, different faults have been investigated; however, the broken rotor bar has become one of the most studied faults since the IM can operate with apparent normality but the consequences can be catastrophic if the fault is not detected in low-severity stages. In this work, a methodology based on convolutional neural networks (CNNs) for automatic detection of broken rotor bars by considering different severity levels is proposed. To exploit the capabilities of CNNs to carry out automatic image classification, the short-time Fourier transform-based time–frequency plane and the motor current signature analysis (MCSA) approach for current signals in the transient state are first used. In the experimentation, four IM conditions were considered: half-broken rotor bar, one broken rotor bar, two broken rotor bars, and a healthy rotor. The results demonstrate the effectiveness of the proposal, achieving 100% of accuracy in the diagnosis task for all the study cases.

show abstract

Classification of EMG signals taken from arm with hybrid CNN‐SVM architecture

Tuncer

Alkan

2021

Concurrency and Computation

View full text Add to dashboard Cite

Analysis based on the classification of electromyography (EMG) signals, the bioelectrical signs that appear during the contraction of the muscles, can be used in many prosthetic control applications. For this purpose, the classification of the EMG signal is considered a pattern‐recognition problem that can be used to improve the functionality and ease of control of reinforced upper‐limb prostheses. Four different EMG‐signal patterns taken from the biceps and triceps muscles were analyzed via hybrid deep‐learning methods after spectrogram‐based preprocessing. Four hundred EMG spectrograms obtained by preprocessing were classified with hybrid deep‐learning techniques based on AlexNet, GoogLeNet, and ResNet18. The classification was conducted using a support vector machine instead of the classification layers after the pooling layer of deep‐learning architectures used in the hybrid system. In general, acceptable classification results were achieved with all techniques used, and the highest performance was obtained with the hybrid system created with AlexNet architecture. The hybrid‐classification achievements with AlexNet, GoogLeNet, and Resnet18 were 99.17%, 95.83%, and 93.33%, respectively. These results show that the proposed architectures can be used in prosthetic controls created using EMG signals.

show abstract

A deep CNN approach to decode motor preparation of upper limbs from time–frequency maps of EEG signals at source level

Cited by 128 publications

References 45 publications

A Deep Classifier for Upper-Limbs Motor Anticipation Tasks in an Online BCI Setting

A Deep Classifier for Upper-Limbs Motor Anticipation Tasks in an Online BCI Setting

Convolutional Neural Network and Motor Current Signature Analysis during the Transient State for Detection of Broken Rotor Bars in Induction Motors

Classification of EMG signals taken from arm with hybrid CNN‐SVM architecture

Contact Info

Product

Resources

About