Classification Of Finger Movements Using Deep Learning

Abstract

This study aims to classify finger movements using electromyography (EMG) signals and deep learning models. Leveraging the potential of EMG signals for prosthetic control, diagnosis, and biomedical applications, a dataset of 350 trials was created by recording five repetitions of seven different finger movements from ten healthy volunteers. During the data processing stage, signal separation was performed using the FastICA method to resolve conflicts in finger-channel mapping. Subsequently, features were extract-ed using Time-Domain Descriptors (TDD), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) modules. The Mutual Information (MI) method was applied for selecting the most discriminative features, and a Fully Connected Neural Network (FCNN) model was chosen for the classification task. The results demonstrated that the model could identify finger movements with high accuracy rates. These findings indicate that deep learning is an effective tool for movement classification in EMG-based systems and point to its applicability in fields such as prosthetic control and rehabilitation technologies.

Keywords

• Electromyography
• finger movement classification
• deep learning
• convolutional neural networks
• long short-term memory
• independent component analysis
• Fully Connected Neural Network

PARMAK HAREKETLERİNİN DERİN ÖĞRENME İLE SINIFLANMASI

Abstract

Bu çalışma, elektromiyografi (EMG) sinyalleri ve derin öğrenme modelleri kullanılarak parmak hareketlerinin sınıflandırılmasını amaçlamaktadır. EMG sinyallerinin protez kontrolü, teşhis ve biyomedikal uygulamalardaki potansiyelinden yola çıkılarak, 10 sağlıklı gönüllüden yedi farklı parmak hareketinin beşer tekrarla kaydedilmesiyle toplam 350 deneylik veri seti oluşturulmuştur. Veri işleme aşamasında, parmak-kanal eşleşmesindeki çakışmaları çözmek için FastICA yöntemiyle sinyal ayrıştırma gerçekleştirilmiş; ardından zaman alanı tanımlayıcıları (TDD), evrişimli sinir ağı (CNN) ve uzun kısa süreli bellek (LSTM) modülleri kullanılarak öznitelikler çıkarılmıştır. En ayırt edici özniteliklerin seçiminde karşılıklı bilgi (MI) yöntemi uygulanmış, sınıflandırma işlemi içinse tam bağlantılı sinir ağı (FCNN) modeli tercih edilmiştir. Sonuçlar, modelin yüksek doğruluk oranlarıyla parmak hareketlerini tanımlayabildiğini ortaya koymuştur. Bu bulgular, derin öğrenmenin EMG tabanlı sistemlerde hareket sınıflandırma için etkili bir araç olduğunu göstermekte olup, protez kontrolü ve rehabilitasyon teknolojileri gibi alanlarda uygulanabilirliğine işaret etmektedir.

Keywords

• parmak hareketlerini sınıflama
• Elektromiyogram
• derin öğrenme
• evrişimsel sinir ağları
• uzun-kısa süreli hafıza
• bağımsız bileşen analizi
• tam bağlantılı sinir ağı

Thanks

In this study, the software provided by the project numbered 2024-047, supported by the Manisa Celal Bayar University Scientific Research Projects Coordination Unit, was used.

References

1. [1] M. M. Taye, “Understanding of Machine Learning with Deep Learning: Architectures, Workflow, Applications and Future Directions,” 2023. doi: 10.3390/computers1205009.
2. [2] A. Ibrahim, V. R. Gannapathy, L. W. Chong, and I. S. M. Isa, “Analysis of electromyography (EMG) signal for human arm muscle: A review,” in Lecture Notes in Electrical Engineering, 2016. doi: 10.1007/978-3-319-24584-3_49.
3. [3] M. Benzyane, M. Azrour, I. Zeroual, and S. Agoujil, “Exploring the Impact of Convolutions on LSTM Networks for Video Classification,” in Lecture Notes in Networks and Systems, 2024. doi: 10.1007/978-3-031-48573-2_4.
4. [4] M. Atzori, M. Cognolato, and H. Müller, “Deep learning with convolutional neural networks: a resource for the control of robotic prosthetic hands via electromyography,” Front Neurorobot, vol. 10, 2016.
5. [5] T. Triwiyanto, V. Abdullayev, and A. A. Ahmed, “Deep Convolution Neural Network to Improve Hand Motion Classification Performance Against Varying Orientation Using Electromyography Signal,” International Journal of Precision Engineering and Manufacturing, vol. 25, no. 6, 2024, doi: 10.1007/s12541-024-00985-x.
6. [6] B. Azhiri, M. Esmaeili, and M. Nourani, “Real-Time EMG Signal Classification via Recurrent Neural Networks,” in Proceedings- 2021 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2021, 2021. doi: 10.1109/BIBM52615.2021.9669872.
7. [7] K. H. Lee, J. Y. Min, and S. Byun, “Electromyogram-based classification of hand and finger gestures using artificial neural networks,” Sensors, vol. 22, no. 1, 2022, doi: 10.3390/s22010225.
8. [8] Ari, A. (2023). EMG signal classification using deep learning and time domain descriptorsbased feature extraction for hand grip movement recognition. Traitement du Signal, Vol. 40, No. 3, pp. 949-960. https://doi.org/10.18280/ts.400311.

9. [9] V. Medved, S. Medved, and I. Kovač, “Critical Appraisal of Surface Electromyography (sEMG) as a Taught Subject and Clinical Tool in Medicine and Kinesiology,” 2020. doi: 10.3389/fneur.2020.560363.
10. [10] S. K. Dwivedi, J. Ngeo, and T. Shibata, “Extraction of Nonlinear Synergies for Proportional and Simultaneous Estimation of Finger Kinematics,” IEEE Trans Biomed Eng, vol. 67, no. 9, 2020, doi: 10.1109/TBME.2020.2967154.
11. [11] A. Tharwat, “Independent component analysis: An introduction,” 2018. doi: 10.1016/j.aci.2018.08.006.
12. [12] M. Chen and P. Zhou, “A novel framework based on FastICA for high density surface EMG decomposition,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 24, no. 1, 2016, doi: 10.1109/TNSRE.2015.2412038.
13. [13] R. Yamashita, M. Nishio, R. K. G. Do, and K. Togashi, “Convolutional neural networks: an overview and application in radiology,” 2018. doi: 10.1007/s13244-018-0639-9.
14. [14] P. Mahajan, "Fully connected vs Convolutional Neural Networks," Medium. [Online].Available:https://medium.com/swlh/fully-connected-vs-convolutional-neural-networks-813ca7bc6ee5.