Browsing by Author "Guren, Onan"

Now showing 1 - 4 of 4

Citation - Scopus: 4
Classification of Hand Gestures Using Semg Signals and Hilbert-Huang Transform
(IEEE, 2022) Kisa, Deniz Hande; Ozdemir, Mehmet Akif; Guren, Onan; Akan, Aydin
Artificial intelligence is effectively utilized for hand gesture classification in myoelectric systems. In this study, hand movement classification is performed with ML algorithms using electromyography (EMG) signals of 7 hand gestures. The Hilbert-Huang Transform (HHT) was applied to the preprocessed EMG signals to obtain the Hilbert-Huang spectrum (HHS). Six different Gray Level Co-occurrence Matrix (GLCM)-based features were extracted from HHS images. In order to validate the proposed method, the same features were extracted from the snapshots of EMG signals and intrinsic mode functions (IMF) extracted by empirical mode decomposition (EMD), separately. These features are classified with 29 different Machine learning (ML) approaches in the MATLAB (R) environment. Among these three approaches, the HHS-based novel method yielded the best performance, with an accuracy of 90.87% from the Cubic Support Vector Machine (SVM). The novel HHS and GLCM-based approach may be used in EMG-based biomedical systems as a promising alternative.
Citation - WoS: 36
Citation - Scopus: 52
Dataset for Multi-Channel Surface Electromyography (semg) Signals of Hand Gestures
(Elsevier, 2022) Ozdemir, Mehmet Akif; Kisa, Deniz Hande; Guren, Onan; Akan, Aydin
This paper presents an electromyography (EMG) signal dataset for use in human-computer interaction studies. The dataset includes 4-channel surface EMG data from 40 participants with an equal gender distribution. The gestures in the data are rest or neutral state, extension of the wrist, flexion of the wrist, ulnar deviation of the wrist, radial deviation of the wrist, grip, abduction of all fingers, adduction of all fingers, supination, and pronation. Data were collected from 4 forearm muscles when simulating 10 unique hand gestures and recorded with the BIOPAC MP36 device using Ag/AgCl surface bipolar electrodes. Each participant's data contains five repetitive cycles of ten hand gestures. A demographic survey was applied to the participants before the signal recording process. This data can be utilized for recognition, classification, and prediction studies in order to develop EMG-based hand movement controller systems. The dataset can also be useful as a reference to create an artificial intelligence model (especially a deep learning model) to detect gesture-related EMG signals. Additionally, it is encouraged to use the proposed dataset for benchmarking current datasets in the literature or for validation of machine learning and deep learning models created with different datasets in accordance with the participant-independent validation strategy. (c) 2022 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
Citation - WoS: 69
Citation - Scopus: 85
Hand Gesture Classification Using Time-Frequency Images and Transfer Learning Based on Cnn
(Elsevier Sci Ltd, 2022) Ozdemir, Mehmet Akif; Kisa, Deniz Hande; Guren, Onan; Akan, Aydin
Hand gesture-based systems are one of the most effective technological advances and continue to develop with improvements in the field of human-computer interaction. Surface electromyography (sEMG) is utilized as a popular input data for gesture classification with elevated accuracy and advanced control capability. This paper presents a comparative hand gesture classification approach using time-frequency (TF) images of the spontaneous sEMG signals and the transfer learning method. 4-channel sEMG signals are collected from 30 subjects performing 7 specific hand gestures. After the required pre-processing, segmentation, and windowing steps, three TF analysis methods, namely Short-Time Fourier Transform (STFT), Continuous Wavelet Transform (CWT), and Hilbert-Huang Transform (HHT), are applied to EMG signals to obtain TF images. Spectrograms from STFT, scalograms from CWT, and Hilbert-Huang spectra (HHS) from HHT obtained from multi-channel sEMG data are separately fused. TF images are then utilized to extract distinct features using seven state-of-the-art, pre-trained Convolutional Neural Network (CNN) architectures and classify seven hand gestures. Two different robust crossvalidation strategies are conducted to evaluate the proposed method; stratified k-fold cross-validation (SKCV) and leave-one-subject-out cross-validation (LOOCV). We also investigate the effect of window size and the combination of Intrinsic Mode Functions (IMFs) on classification performance. The results demonstrated that the HHT utilizing IMFs obtained by Empirical Mode Decomposition (EMD) provided improved TF resolution and better results than STFT and CWT in the classification of sEMG signals. Finally, the best average accuracies (93.75% for SKCV) and (94.41% for LOOCV) are obtained by the HHT method with the pre-trained ResNet-50 model.
Citation - Scopus: 3
Investigating the Effect of Signal Channels and Features in Various Domains on the Emg-Based Hand Gesture Classification
(IEEE, 2022) Kisa, Deniz Hande; Yildirim, Muhiddin Ceyhun; Ozdil, Belkis; Ozdemir, Mehmet Akif; Guren, Onan; Akan, Aydin
A variety of artificial intelligence (AI) approaches are applied for the classification of hand movements in systems that use electromyography (EMG), which measures the electrical activity of muscles. In AI approaches, machine learning (ML) is frequently preferred and researched for this classification issue. In this study, hand gesture classification was performed with ML algorithms using EMGs of 10 hand movements. Features were extracted from the time domain (TD), frequency domain (FD), time-frequency domain (TFD) (via Wavelet-based Synchrosqueezing Transform), and Fractional Fourier Transform (FrFT) domain. After training 31 ML models with all features, Subspace k-Nearest Neighbor (kNN), which is ensemble-based learning, was determined as the best model. This model was trained with different feature and channel combinations, and the classification performances were examined as channel-based and domain-based, separately. In all cases, an accuracy of 97.10% was obtained as the highest via the TD-FD-FrFT domain feature combination, including all channels. When all the results are examined, an alternative classification approach is presented to the literature by proving that the computational load decreases while the accuracy value increases by determining and utilizing the channels and features that contain more related information about hand movement.