Classification of imagery motor EEG data with wavelet denoising and features selection

Sun, Lei; Feng, Zu Ren

doi:10.1109/icwapr.2016.7731641

Cited by 19 publications

(6 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, the accuracy comparison between our method and other state-of-the-art and baselines are listed in Table IV. Wavelet transform [2], [15]- [18] and independent component analysis (ICA) [19], [20] are state-of-the-art methods to process EEG signals. The Deep Neural Network [17], [19], [21] and Linear discriminant analysis [20] are applied to classify the EEG data.…”

Section: B Overall Comparisonmentioning

confidence: 99%

Converting Your Thoughts to Texts: Enabling Brain Typing via Deep Feature Learning of EEG Signals

Zhang

Yao

Sheng

et al. 2018

2018 IEEE International Conference on Pervasive Computing and Communications (PerCom)

View full text Add to dashboard Cite

An electroencephalography (EEG) based Brain Computer Interface (BCI) enables people to communicate with the outside world by interpreting the EEG signals of their brains to interact with devices such as wheelchairs and intelligent robots. More specifically, motor imagery EEG (MI-EEG), which reflects a subject's active intent, is attracting increasing attention for a variety of BCI applications. Accurate classification of MI-EEG signals while essential for effective operation of BCI systems, is challenging due to the significant noise inherent in the signals and the lack of informative correlation between the signals and brain activities. In this paper, we propose a novel deep neural network based learning framework that affords perceptive insights into the relationship between the MI-EEG data and brain activities. We design a joint convolutional recurrent neural network that simultaneously learns robust high-level feature presentations through low-dimensional dense embeddings from raw MI-EEG signals. We also employ an Autoencoder layer to eliminate various artifacts such as background activities. The proposed approach has been evaluated extensively on a largescale public MI-EEG dataset and a limited but easy-to-deploy dataset collected in our lab. The results show that our approach outperforms a series of baselines and the competitive state-of-theart methods, yielding a classification accuracy of 95.53%. The applicability of our proposed approach is further demonstrated with a practical BCI system for typing.

show abstract

Section: B Overall Comparisonmentioning

confidence: 99%

Converting Your Thoughts to Texts: Enabling Brain Typing via Deep Feature Learning of EEG Signals

Zhang

Yao

Sheng

et al. 2018

2018 IEEE International Conference on Pervasive Computing and Communications (PerCom)

View full text Add to dashboard Cite

show abstract

“…[17] builds one deep belief net (DBN) classifier for each channel and combines them through Ada-boost algorithm and classifies the left and right hand motor imagery.The work achieves average 83% accuracy. [13] adopts SVM as the classifier and achieves an average accuracy of 65% with the input data being denoised by a wavelet denoising algorithm before power spectral density (PSD) feature selection. [19] yields an accuracy of 80% with the foundational universal background models (UBMs) classifier after the data is processed by I-vectors and Joint Factor Analysis (JFA).…”

Section: Related Workmentioning

confidence: 99%

“…First, the data pre-processing, parameters selection and feature engineering are time-consuming and highly dependent on human expertise. Second, current accuracies mostly center around 60 ∼ 85% [4,13,12], which are too low for real-world deployment. Finally, existing research mainly focus on binary intents recognition while multi-intent scenario dominates the practical applications.…”

Section: Introductionmentioning

confidence: 99%

Intent Recognition in Smart Living Through Deep Recurrent Neural Networks

Zhang

Yao

Huang

et al. 2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Electroencephalography (EEG) signal based intent recognition has recently attracted much attention in both academia and industries, due to helping the elderly or motor-disabled people controlling smart devices to communicate with outer world. However, the utilization of EEG signals is challenged by low accuracy, arduous and timeconsuming feature extraction. This paper proposes a 7-layer deep learning model to classify raw EEG signals with the aim of recognizing subjects' intents, to avoid the time consumed in pre-processing and feature extraction. The hyper-parameters are selected by an Orthogonal Array experiment method for efficiency. Our model is applied to an open EEG dataset provided by PhysioNet and achieves the accuracy of 0.9553 on the intent recognition. The applicability of our proposed model is further demonstrated by two use cases of smart living (assisted living with robotics and home automation).

show abstract

“…time. The short-time Fourier transform and the wavelet transform are the most applicable [13,14,15,16]. Spatial filtering, which extracts signals from multiple sensors to look at the activity localized in a particular brain region, is also applied.…”

Section: Introductionmentioning

confidence: 99%

Design of MEG-Based Brain-Machine Interface Control Methodology Through Time-Varying Cortical Neural Connectivity & Extreme Learning Machine

Uyulan

2022

Preprint

View full text Add to dashboard Cite

Human-machine interfaces contribute to the improvement of the life quality of physically disabled users. In this study, a non-invasive brain-machine interface (BMI) design methodology was proposed to control a robot arm through magnetoencephalography (MEG) based on directionally modulated MEG activity that was acquired during the user's imagined wrist movements in four various directions. The partial directed coherence (PDC) measure derived from functional connectivity between cortical brain regions was utilized in the feature extraction process. The time-varying parameters were estimated based on a time-varying multivariate adaptive autoregressive (AAR) model, that can detect task-dependent features and non-symmetric channel relevance for mental task discrimination. An extreme learning machine (ELM), that utilizes Moore-Penrose (MP) generalized inverse to set its weights and does not necessitate a gradient-based backpropagation algorithm was employed to generate a model with the extracted feature set. The output of the task classification model was embedded into the robotic arm model for realizing control-based tasks. The classification results dictate that the proposed BMI methodology is a feasible solution for rehabilitation or assistance systems that are devised to help motor-impaired people. The proposed methodology provides very satisfactory classification performance at a fast learning speed.

show abstract

Classification of imagery motor EEG data with wavelet denoising and features selection

Cited by 19 publications

References 13 publications

Converting Your Thoughts to Texts: Enabling Brain Typing via Deep Feature Learning of EEG Signals

Converting Your Thoughts to Texts: Enabling Brain Typing via Deep Feature Learning of EEG Signals

Intent Recognition in Smart Living Through Deep Recurrent Neural Networks

Design of MEG-Based Brain-Machine Interface Control Methodology Through Time-Varying Cortical Neural Connectivity & Extreme Learning Machine

Contact Info

Product

Resources

About