Enhancing the decoding accuracy of EEG signals by the introduction of anchored-STFT and adversarial data augmentation method

Ali, Omair; Saif-ur-Rehman, Muhammad; Dyck, Susanne; Glasmachers, Tobias; Iossifidis, Ioannis; Klaes, Christian

doi:10.1038/s41598-022-07992-w

Cited by 18 publications

(14 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Henceforth, the development of a reliable and robust muscle-to-machine interfaces decoder remains one of the primary goals of the brain-machine interface (BMI) community. Real life applications, such as control of prosthesis, rehabilitation-aiding exoskeletons, and BMI for gaming are active research areas [1]. Recently published studies on movement classification based on sEMG showed promising results with high accuracy and real-time processing capabilities [2][3][4].…”

Section: Introductionmentioning

confidence: 99%

Deep transfer learning compared to subject-specific models for sEMG decoders

Johann¹,

Saif-ur-Rehman²,

Glasmachers³

et al. 2022

J. Neural Eng.

Self Cite

View full text Add to dashboard Cite

Objective Accurate decoding of surface electromyography (sEMG) is pivotal for muscle-to-machine-interfaces (MMI) and their application e.g. rehabilitation therapy. sEMG signals have high inter-subject variability, due to various factors, including skin thickness, body fat percentage, and electrode placement. Deep learning algorithms require long training time and tend to overfit if only few samples are available. In this study, we aim to investigate methods to calibrate deep learning models to a new user when only a limited amount of training data is available. Approach Two methods are commonly used in the literature, subject-specific modeling and transfer learning. In this study, we investigate the effectiveness of transfer learning using weight initialization for recalibration of two different pretrained deep learning models on new subjects data and compare their performance to subject- specific models. We evaluate two models on three publicly available databases (NinaPro Database 2,3 and 4) and compare the performance of both calibration schemes in terms of accuracy, required training data, and calibration time. Main results On average over all settings, our transfer learning approach improves 5 %-points on the pretrained models without fine-tuning, and 12 %-points on the subject-specific models, while being trained for 22% fewer epochs on average. Our results indicate that transfer learning enables faster learning on fewer training samples than user-specific models. Significance To the best of our knowledge, this is the first comparison of subject- specific modeling and transfer learning. These approaches are ubiquitously used in the field of sEMG decoding. But the lack of comparative studies until now made it difficult for scientists to assess appropriate calibration schemes. Our results guide engineers evaluating similar use cases.

show abstract

Section: Introductionmentioning

confidence: 99%

Deep transfer learning compared to subject-specific models for sEMG decoders

Johann¹,

Saif-ur-Rehman²,

Glasmachers³

et al. 2022

J. Neural Eng.

Self Cite

View full text Add to dashboard Cite

show abstract

“…But the proposed model has outperformed the baseline methods by generating a TFR of the MI data from 2 channels, thereby having lesser complexity than the model used in [38]. The method proposed in [28] has classified the left and right hand MI tasks with an accuracy of 81.2% by using the MI EEG data from 3 channels, an anchored STFT for feature extraction and a skip-net CNN model with data augmentation techniques. The method proposed in this study has attained a slightly higher accuracy as compared to [28] by using the data from lesser number of channels and a CNN model with a simple architecture.…”

Section: Comparison Of Results For Different N Valuesmentioning

confidence: 99%

“…The method proposed in [28] has classified the left and right hand MI tasks with an accuracy of 81.2% by using the MI EEG data from 3 channels, an anchored STFT for feature extraction and a skip-net CNN model with data augmentation techniques. The method proposed in this study has attained a slightly higher accuracy as compared to [28] by using the data from lesser number of channels and a CNN model with a simple architecture. Similarly, it gives a higher accuracy than [39] which uses Sliding Window (SW) to extract nine timeframes of length 1 sec from the MI data of multiple channels and uses CSP and LDA for feature extraction and classification of data in each frame respectively.…”

Section: Comparison Of Results For Different N Valuesmentioning

confidence: 99%

“…Deep learning methods are recently being used in BCI studies as they effectively deal with non-stationary data and can automatically learn the underlying features from the data [27][28][29]. In [30] different architectures of the convolutional neural networks (CNN) named ConvNets were built using the recent advances in deep learning, for decoding the MI tasks.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Classification of Motor Imagery EEG signals using high resolution time-frequency representations and convolutional neural network

Srimadumathi,

Ramasubba Reddy

2024

Biomed. Phys. Eng. Express

View full text Add to dashboard Cite

A Motor Imagery (MI) based Brain Computer Interface (BCI) system aims to provide neuro-rehabilitation for the motor disabled people and patients with brain injuries (e.g., stroke patients) etc. The aim of this work is to classify the left and right hand MI tasks by utilizing the occurrence of event related desynchronization and synchronization (ERD\ERS) in the Electroencephalogram (EEG) during these tasks. This study proposes to use a set of Complex Morlet Wavelets (CMW) having frequency dependent widths to generate high-resolution time-frequency representations (TFR) of the MI EEG signals present in the channels C3 and C4. A novel method for the selection of the value of number of cycles relative to the center frequency of the CMW is studied here for extracting the MI task features. The generated TFRs are given as input to a Convolutional neural network (CNN) for classifying them into left or right hand MI tasks. The proposed framework attains a classification accuracy of 82.2% on the BCI Competition IV dataset 2a, showing that the TFRs generated in this work give a higher classification accuracy than the baseline methods and other existing algorithms.

show abstract

“…These techniques involve a variety of methodologies aimed at increasing the adaptability and accuracy of MI-EEG classification models. Broadly categorized, these techniques can be divided into two principal categories: those that operate on the manipulation of raw EEG waveforms and those that leverage domain transformation techniques [18][19][20], such as short-time Fourier transform (STFT) [21], to derive augmented data in the alternative domains (e.g. frequency domain, time-frequency domain, spatial domain).…”

Section: Related Workmentioning

confidence: 99%

Improving cross-subject classification performance of motor imagery signals: a data augmentation-focused deep learning framework

Ozelbas,

Tülay,

Ozekes

2024

Mach. Learn.: Sci. Technol.

View full text Add to dashboard Cite

Motor Imagery Brain-Computer Interfaces (MI-BCIs) have gained a lot of attention in recent years thanks to their potential to enhance rehabilitation and control of prosthetic devices for individuals with motor disabilities. However, accurate classification of motor imagery signals remains a challenging task due to the high inter-subject variability and non-stationarity in the electroencephalogram (EEG) data. In the context of MI-BCIs, with limited data availability, the acquisition of EEG data can be difficult. In this study, several data augmentation techniques have been compared with the proposed data augmentation technique Adaptive Cross-Subject Segment Replacement (ACSSR). This technique, in conjunction with the proposed deep learning framework, allows for a combination of similar subject pairs to take advantage of one another and boost the classification performance of MI-BCIs. The proposed framework features a multi-domain feature extractor based on Common Spatial Patterns (CSP) with a sliding window and a parallel two-branch Convolutional Neural Network (CNN). The performance of the proposed methodology has been evaluated on the multi-class BCI Competition IV Dataset 2a through repeated 10- fold cross-validation. Experimental results indicated that the implementation of the ACSSR method (80.46%) in the proposed framework has led to a considerable improvement in the classification performance compared to the classification without data augmentation (77.63%), and other fundamental data augmentation techniques used in the literature. The study contributes to the advancements for the development of effective MI-BCIs by showcasing the ability of the ACSSR method to address the challenges in motor imagery signal classification tasks.

show abstract

Enhancing the decoding accuracy of EEG signals by the introduction of anchored-STFT and adversarial data augmentation method

Cited by 18 publications

References 48 publications

Deep transfer learning compared to subject-specific models for sEMG decoders

Deep transfer learning compared to subject-specific models for sEMG decoders

Classification of Motor Imagery EEG signals using high resolution time-frequency representations and convolutional neural network

Improving cross-subject classification performance of motor imagery signals: a data augmentation-focused deep learning framework

Contact Info

Product

Resources

About