Deep Learning for Electromyographic Hand Gesture Signal Classification Using Transfer Learning
read more
Citations
A Comprehensive Survey on Transfer Learning
EMG Pattern Recognition in the Era of Big Data and Deep Learning
Deep Learning in Physiological Signal Data: A Survey.
Deep Learning for EMG-based Human-Machine Interaction: A Review
Real-Time Surface EMG Pattern Recognition for Hand Gestures Based on an Artificial Neural Network.
References
Deep Residual Learning for Image Recognition
Adam: A Method for Stochastic Optimization
ImageNet Classification with Deep Convolutional Neural Networks
Dropout: a simple way to prevent neural networks from overfitting
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Related Papers (5)
Deep Learning with Convolutional Neural Networks Applied to Electromyography Data: A Resource for the Classification of Movements for Prosthetic Hands
Frequently Asked Questions (15)
Q2. What future works have the authors mentioned in the paper "Deep learning for electromyographic hand gesture signal classification using transfer learning" ?
Future works will focus on adapting and testing the proposed TL algorithm on upper-extremity amputees. This will provide additional challenges due to the greater muscle variability across amputees and the decrease in classification accuracy compared to able-bodied participants [ 35 ]. Additionally, tests for the application of the proposed TL algorithm for inter-session classification will be conducted as to be able to leverage labeled information for long-term classification.
Q3. Why is the source task difficult in the present context?
Due to the application of the multi-stream AdaBatch scheme, the source task in the present context is to learn the general mapping between muscle activity and gestures.
Q4. How many rounds are available for the evaluation dataset?
For the evaluation dataset three rounds are available with the first round utilized for training (i.e. 140s per participant) and the last two for testing (i.e. 240s per participant).
Q5. What would be the easiest way to address this?
The most straightforward way of addressing this would be to numerically remove the relevant channels from the dataset used for pre-training.
Q6. How many different combinations of hyperparameters were tested for each classifier?
Hyperparameters for each classifier were selected by employing three fold cross-validation alongside random search, testing 50 different combinations of hyperparameters for each participant’s dataset for each classifier.
Q7. How many repetitions were required to perform the gesture?
1) Data Acquisition and Processing: Each participant was asked to hold a gesture for five seconds followed by three seconds of neutral gesture and to repeat this action five more times (total of six repetitions).
Q8. What is the main purpose of this paper?
As this paper’s main purpose is to present a deep learningbased TL approach to the problem of sEMG hand gesture recognition, contextualizing the performance of the proposed algorithms within the current state-of-the-art is essential.
Q9. What is the way to bridge the dimensionality gap between the two networks?
one could replace the learned scalar layers in the target network by convolutions or fully connected layers to bridge the dimensionality gap between potentially vastly different source and second networks.
Q10. How is the average accuracy of the participants calculated?
As such, in an effort to quantify the impact of muscle fatigue on the classifier’s performance, the average accuracy of the eight participants over the five minute session is computed as a function of time.
Q11. Why are these transitions not part of the training dataset?
These transitions are not part of the training dataset, because they are too time consuming to record as the number of possible transitions equals n2−n where n is the number of gestures.
Q12. How will the proposed TL algorithm be used?
tests for the application of the proposed TL algorithm for inter-session classification will be conducted as to be able to leverage labeled information for long-term classification.
Q13. What is the simplest way to calculate the spectrogram of a signal?
When calculating the spectrogram of a signal, the information is structured in a Time x Frequency fashion (Time x Scale for CWT).
Q14. How can deep learning algorithms be efficiently trained?
Showing that deep learning algorithms can be efficiently trained, within the inherent constraints of sEMGbased hand gesture recognition, offers exciting new research avenues for this field.
Q15. What is the significance of the proposed TL algorithm?
This suggests that the proposed TL algorithm enables the network to learn features that can generalize not only across participants but also for never-seenbefore gestures.