Abstract:In recent years, deep learning algorithms have become increasingly more prominent for their unparalleled ability to automatically learn discriminant features from large amounts of data. However, within the field of electromyography-based gesture recognition, deep learning algorithms are seldom employed as they require an unreasonable amount of effort from a single person, to generate tens of thousands of examples. This work's hypothesis is that general, informative features can be learned from the large amounts of data generated by aggregating the signals of multiple users, thus reducing the recording burden while enhancing gesture recognition. Consequently, this paper proposes applying transfer learning on aggregated data from multiple users, while leveraging the capacity of deep learning algorithms to learn discriminant features from large datasets. Two datasets comprised of 19 and 17 able-bodied participants respectively (the first one is employed for pre-training) were recorded for this work, using the Myo Armband. A third Myo Armband dataset was taken from the NinaPro database and is comprised of 10 able-bodied participants. Three different deep learning networks employing three different modalities as input (raw EMG, Spectrograms and Continuous Wavelet Transform (CWT)) are tested on the second and third dataset. The proposed transfer learning scheme is shown to systematically and significantly enhance the performance for all three networks on the two datasets, achieving an offline accuracy of 98.31% for 7 gestures over 17 participants for the CWT-based ConvNet and 68.98% for 18 gestures over 10 participants for the raw EMG-based ConvNet. Finally, a use-case study employing eight able-bodied participants suggests that real-time feedback allows users to adapt their muscle activation strategy which reduces the degradation in accuracy normally experienced over time.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in gesture recognition based on surface electromyography (sEMG) signals, the application of deep - learning algorithms faces the problem of excessive data volume requirements. Specifically: 1. **Large data volume requirements**: Traditional deep - learning methods require a large amount of training data to automatically learn discriminative features, and it is difficult for a single user to generate tens of thousands of samples in one experiment. 2. **Differences between users**: There are significant differences in sEMG signals among different users, even when the electrode placement positions are exactly the same. Therefore, directly applying a classifier of one user to another user usually has poor performance. 3. **Difficulty in data acquisition**: Due to the non - stationarity of sEMG signals and the complexity of high - quality data collection, it is very difficult to obtain a large amount of high - quality sEMG data. To solve these problems, the paper proposes the following hypotheses and solutions: - **Hypothesis**: By aggregating signal data from multiple users, general and information - rich features can be learned from a large amount of data, thereby reducing the recording burden on each user and improving the performance of gesture recognition. - **Solution**: Apply transfer learning (TL) technology, use the aggregated data of multiple users for pre - training, and then fine - tune the data of new users. This can achieve higher recognition accuracy with a smaller amount of data. ### Specific contributions 1. **New TL scheme**: A new transfer learning scheme is proposed, using a convolutional neural network (ConvNet) to utilize multi - user data, thereby improving the effect of sEMG gesture recognition. 2. **Three input modalities**: Three different deep - learning networks are tested, using the original sEMG signal, spectrograms, and continuous wavelet transform (CWT) as input modalities respectively. 3. **Public data set**: A new sEMG gesture recognition data set containing 36 healthy participants is released, aiming to provide a useful tool for the research community. 4. **Real - time feedback experiment**: Experiments prove that real - time feedback can help users adapt to muscle activation strategies and reduce the decrease in accuracy over time. ### Experimental results - On the CWT - based ConvNet, the offline accuracy rate for 7 gestures reaches 98.31%. - On the raw EMG - based ConvNet, the offline accuracy rate for 18 gestures reaches 68.98%. - The real - time feedback experiment shows that users can reduce the impact of signal drift by adjusting muscle activation strategies. Through these methods, the paper shows how to effectively use transfer learning and multi - user data to improve the performance of sEMG - based gesture recognition systems.

Deep Learning for Electromyographic Hand Gesture Signal Classification Using Transfer Learning

A deep learning approach using attention mechanism and transfer learning for electromyographic hand gesture estimation

Online cross session electromyographic hand gesture recognition using deep learning and transfer learning

Deep end-to-end transfer learning for robust inter-subject and inter-day hand gesture recognition using surface EMG

A Deep Learning Sequential Decoder for Transient High-Density Electromyography in Hand Gesture Recognition Using Subject-Embedded Transfer Learning

Hand Gesture Recognition based on Surface Electromyography using Convolutional Neural Network with Transfer Learning Method

Machine Learning-based sEMG Signal Classification for Hand Gesture Recognition

Surface EMG-Based Inter-Session Gesture Recognition Enhanced by Deep Domain Adaptation.

Gesture Classification in Electromyography Signals for Real-Time Prosthetic Hand Control Using a Convolutional Neural Network-Enhanced Channel Attention Model

Transfer learning in hand movement intention detection based on surface electromyography signals

Towards Robust and Interpretable EMG-based Hand Gesture Recognition using Deep Metric Meta Learning

Interpreting Deep Learning Features for Myoelectric Control: A Comparison with Handcrafted Features

Dual Stream Long Short-Term Memory Feature Fusion Classifier for Surface Electromyography Gesture Recognition

EMG-based Multi-User Hand Gesture Classification via Unsupervised Transfer Learning Using Unknown Calibration Gestures

A Transferable Adaptive Domain Adversarial Neural Network for Virtual Reality Augmented EMG-Based Gesture Recognition

sEMG-based Hand Gesture Recognition with Deep Learning

Deep Learning with Convolutional Neural Networks Applied to Electromyography Data: A Resource for the Classification of Movements for Prosthetic Hands

From Forearm to Wrist: Deep Learning for Surface Electromyography-Based Gesture Recognition.

Gesture Recognition Through sEMG with Wearable Device Based on Deep Learning

Compact Deep Neural Networks for Computationally Efficient Gesture Classification From Electromyography Signals