Selection of the Best Set of Features for sEMG-Based Hand Gesture Recognition Applying a CNN Architecture

Currently, surface EMG signals have a wide range of applications in human–computer interaction systems. However, selecting features for gesture recognition models based on traditional machine learning can be challenging and may not yield satisfactory results. Considering the strong nonlinear generalization ability of neural networks, this paper proposes a two-stream residual network model with an attention mechanism for gesture recognition. One branch processes surface EMG signals, while the other processes hand acceleration signals. Segmented networks are utilized to fully extract the physiological and kinematic features of the hand. To enhance the model’s capacity to learn crucial information, we introduce an attention mechanism after global average pooling. This mechanism strengthens relevant features and weakens irrelevant ones. Finally, the deep features obtained from the two branches of learning are fused to further improve the accuracy of multi-gesture recognition. The experiments conducted on the NinaPro DB2 public dataset resulted in a recognition accuracy of 88.25% for 49 gestures. This demonstrates that our network model can effectively capture gesture features, enhancing accuracy and robustness across various gestures. This approach to multi-source information fusion is expected to provide more accurate and real-time commands for exoskeleton robots and myoelectric prosthetic control systems, thereby enhancing the user experience and the naturalness of robot operation.

Section: Discussionmentioning

confidence: 99%

“…J.A. Sandoval-Espino et al [46] conducted an experiment using CNNs to analyze four sets of the time-domain features of EMG signals.…”

Section: Comparative Analysis With Other Methodsmentioning

confidence: 99%

Study on Gesture Recognition Method with Two-Stream Residual Network Fusing sEMG Signals and Acceleration Signals

Hu,

Wang,

et al. 2024

“…Consequently, they can be employed in real-time applications. Furthermore, they have been demonstrated to achieve high gesture recognition accuracy (exceeding 90%) when combined with machine learning classifiers [ 56 , 57 ]. Nevertheless, time-frequency-domain features, such as Short-Time Fourier Transform (STFT) coefficients, Spectral Moment (SM), and Stockwell Transform coefficients, have the potential to capture non-stationarities in sEMG signals.…”

Section: Methodsmentioning

confidence: 99%

“…Nevertheless, time-frequency-domain features, such as Short-Time Fourier Transform (STFT) coefficients, Spectral Moment (SM), and Stockwell Transform coefficients, have the potential to capture non-stationarities in sEMG signals. However, studies have not conclusively demonstrated superior performance over time-domain features, which would justify the added computational complexity for many applications [ 56 , 58 , 59 ]. After extracting 37 features from each of the eight sEMG channels, we tested the performance of the kNN classification algorithm using the database containing 296 features (37 × 8 = 296).…”

Section: Methodsmentioning

confidence: 99%

Comparative Study of sEMG Feature Evaluation Methods Based on the Hand Gesture Classification Performance

Hellara,

Barioul,

Sahnoun

et al. 2024

Effective feature extraction and selection are crucial for the accurate classification and prediction of hand gestures based on electromyographic signals. In this paper, we systematically compare six filter and wrapper feature evaluation methods and investigate their respective impacts on the accuracy of gesture recognition. The investigation is based on several benchmark datasets and one real hand gesture dataset, including 15 hand force exercises collected from 14 healthy subjects using eight commercial sEMG sensors. A total of 37 time- and frequency-domain features were extracted from each sEMG channel. The benchmark dataset revealed that the minimum Redundancy Maximum Relevance (mRMR) feature evaluation method had the poorest performance, resulting in a decrease in classification accuracy. However, the RFE method demonstrated the potential to enhance classification accuracy across most of the datasets. It selected a feature subset comprising 65 features, which led to an accuracy of 97.14%. The Mutual Information (MI) method selected 200 features to reach an accuracy of 97.38%. The Feature Importance (FI) method reached a higher accuracy of 97.62% but selected 140 features. Further investigations have shown that selecting 65 and 75 features with the RFE methods led to an identical accuracy of 97.14%. A thorough examination of the selected features revealed the potential for three additional features from three specific sensors to enhance the classification accuracy to 97.38%. These results highlight the significance of employing an appropriate feature selection method to significantly reduce the number of necessary features while maintaining classification accuracy. They also underscore the necessity for further analysis and refinement to achieve optimal solutions.

“…For real-time wearable sEMG devices, the total response time of myoelectric control should be limited to 300 ms [ 25 , 26 , 27 ]. A wide window function bandwidth results in a low time-domain resolution, and a narrow window function bandwidth results in a low frequency-domain resolution [ 28 ].…”

Section: Experimental Methodsmentioning

confidence: 99%

High-Performance Surface Electromyography Armband Design for Gesture Recognition

Zhang

Hong

Zhang

et al. 2023

Wearable surface electromyography (sEMG) signal-acquisition devices have considerable potential for medical applications. Signals obtained from sEMG armbands can be used to identify a person’s intentions using machine learning. However, the performance and recognition capabilities of commercially available sEMG armbands are generally limited. This paper presents the design of a wireless high-performance sEMG armband (hereinafter referred to as the α Armband), which has 16 channels and a 16-bit analog-to-digital converter and can reach 2000 samples per second per channel (adjustable) with a bandwidth of 0.1–20 kHz (adjustable). The α Armband can configure parameters and interact with sEMG data through low-power Bluetooth. We collected sEMG data from the forearms of 30 subjects using the α Armband and extracted three different image samples from the time–frequency domain for training and testing convolutional neural networks. The average recognition accuracy for 10 hand gestures was as high as 98.6%, indicating that the α Armband is highly practical and robust, with excellent development potential.