Unsupervised Learning for Exploring Hidden Structures in Self-Talk DOI

Kellen Tyrrell,

Masoumeh Heidari Kapourchali

Published: Oct. 9, 2023

Innerspeech decoding from EEG data holds significant importance due to its potential revolutionize human-machine interaction and communication systems. Leveraging the power of temporal shift-invariant sparse coding, this study explores unsupervised learning inner-speech patterns using EEG, a prominent modality in body sensor networks. By analyzing data, we investigate characteristics code activities distinguish between different classes conditions. The results showcase effectiveness model, emphasizing for accurate inner speech without need explicit class labels. Furthermore, assess significance an ANOVA test, providing statistical evidence their discriminative across To discriminatory dictionaries, compare Multilayer Perceptron (MLP) Convolutional Neural Networks (CNN) classifiers on both raw dictionary outputs. findings demonstrate that accuracy does not decrease when employing approach, showcasing decoding. This research significantly contributes field signal processing networks, paving way advancements innerspeech applications diverse range domains.

Language: Английский

Time–frequency domain machine learning for detection of epilepsy using wearable EEG sensor signals recorded during physical activities DOI
Shaswati Dash, Dinesh Kumar Dash, Rajesh Kumar Tripathy

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 100, P. 107041 - 107041

Published: Oct. 16, 2024

Language: Английский

Citations

1
Detecting Microstate Transition in Human Brain via Eigenspace of Spatiotemporal Graph DOI
Raghav Dev, Sandeep Kumar, Tapan Kumar Gandhi

et al.

IEEE Sensors Letters, Journal Year: 2023, Volume and Issue: 7(5), P. 1 - 4

Published: April 26, 2023

In this letter, a novel approach for detecting the transition of electroencephalography (EEG) microstates human brain has been proposed. We have considered EEG electrodes as nodes graph and correlation between electrodes' signals edge weights. Then, using spectral analysis graph, method proposed microstates. The is comprised two steps. First, spatiotemporal constructed Laplacian spatial at consecutive time instants. principal angles eigenspace microstate detected. Experimental results on publicly available datasets show that performs more accurately than state-of-the-art. On first dataset 15 out 20 subjects, improved accuracy time. second dataset, it missed only transitions opposed to which failed detect overall ten across all subjects.

Language: Английский

Citations

2
Analysis of Machine Learning Models Using Proposed EEG Vowel Dataset DOI Creative Commons
Asif Iqbal, Arpit Bhardwaj,

Ashok Kumar Suhag

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: March 27, 2024

Abstract Electroencephalogram (EEG) signals are produced by neurons of human brain and contain frequencies electrical properties. It is easy for a Brain to Computer Interface (BCI) system record EEG using non-invasive methods. Speech imagery (SI) can be used convert speech imaging into text, researches done so far on SI has made use multichannel devices. In this work, we propose signal dataset imagined a/e/i/o/u vowels collected from 5 participants NeuroSky Mindwave Mobile2 single channel device. Decision Tree (DT), Random Forest (RF), Genetic Algorithm (GA) Machine Learning (ML) classifiers trained with proposed dataset. For the dataset, average classification accuracy DT found lower in comparison RF GA. GA shows better performance vowel e/o/u resulting 80.8%, 82.36%, 81.8% 70 − 30 data partition, 80.2%, 81.9%, 80.6% 60 40 partition 79.8%, 81.12%, 78.36% 50–50 partition. Whereas improved a/i which 83.44%, 81.6% 82.2%, 81.2% 81.4%, 80.2% Some other parameters like min. value, max. value accuracy, standard deviation, sensitivity, specificity, precision, F1 score, false positive rate receiver operating characteristics also evaluated anal- ysed. Research proven that functions remains normal patients vocal disorders. Completely disabled equipped such technol- ogy as may one best way them have access over essential day basic requirement.

Language: Английский

Citations

0
Resting state EEG assisted imagined vowel phonemes recognition by native and non-native speakers using brain connectivity measures DOI
Ruchi Juyal, M. Hariharan, Niraj Kumar

et al.

Physical and Engineering Sciences in Medicine, Journal Year: 2024, Volume and Issue: 47(3), P. 939 - 954

Published: April 22, 2024

Language: Английский

Citations

0
An Analysis of EEG Signal Classification for Digit Dataset DOI Creative Commons
Asif Iqbal, Arpit Bhardwaj,

Ashok Kumar Suhag

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: May 2, 2024

Abstract Artificial Intelligence (AI) and Machine Learning has brought significant atten- tion to the human brain, making it a prominent research area in engineering technology other non-medical sciences. Electroencephalogram (EEG) are one of many biological signals that produced by brain. EEG contain electrical properties frequency ranging between 0-100Hz. Fea- tures various attributes recorded which associated with state The data comprises values correspond frequencies signals, specifically delta, theta, alpha, beta, gamma. Additionally, includes information about level attention, medita- tion, eye blinking subject. This given notion how imagined digit is classified from an signal using machine learning algorithms. We have done analysis models like k-Nearest Neighbor (kNN), Convolutional Neural Network (CNN) Genetic Program- ming (GP). An original set created for 0–9 non invasive single electrode (channel) device. obtained accuracy kNN 66.8%, 73.1% GP calculated equal 82%. If lower channel device improved further achieved more then day they may replace higher chan- nel bulky devices. As or portable easy use therefore implementation this work future meet variety applications biomedical engineering, smart health care, personal assistance automation.

Language: Английский

Citations

0
Does Distance Between Electrodes Affect the Accuracy of Decoding the Motor Imagery Using EEG? DOI
Raghav Dev, Sandeep Kumar, Tapan Kumar Gandhi

et al.

IEEE Sensors Letters, Journal Year: 2024, Volume and Issue: 8(8), P. 1 - 4

Published: July 12, 2024

Language: Английский

Citations

0
An Empirical Model-Based Algorithm for Removing Motion-Caused Artifacts in Motor Imagery EEG Data for Classification Using an Optimized CNN Model DOI Creative Commons
Rajesh Kannan Megalingam, Kariparambil Sudheesh Sankardas,

Sakthiprasad Kuttankulangara Manoharan

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(23), P. 7690 - 7690

Published: Nov. 30, 2024

Electroencephalography (EEG) is a non-invasive technique with high temporal resolution and cost-effective, portable, easy-to-use features. Motor imagery EEG (MI-EEG) data classification one of the key applications within brain-computer interface (BCI) systems, utilizing signals from motor tasks. BCI very useful for people severe mobility issues like quadriplegics, spinal cord injury patients, stroke etc., giving them freedom to certain extent perform activities without need caretaker, driving wheelchair. However, motion artifacts can significantly affect quality recordings. The conventional enhancement algorithms are effective in removing ocular muscle stationary subject but not as when motion, e.g., wheelchair user. In this research study, we propose an empirical error model-based artifact removal approach cross-subject (MI) using modified CNN-based deep learning algorithm, designed assist users issues. method applies real tasks measured data, focusing on accurately interpreting practical application. model evolved inertial sensor-based acceleration weight wheelchair, subject, surface friction terrain under Three different wheelchairs five terrains, including road, brick, concrete, carpet, marble, used recording. After evaluating benchmarking proposed CNN model, accuracy achieved 94.04% distinguishing between four specific classes: left, right, front, back. This demonstrates model's effectiveness compared other state-of-the-art techniques. comparative results show that potentially way raise decoding efficiency BCI.

Language: Английский

Citations

0
Systematic Review of EEG-Based Imagined Speech Classification Methods DOI Creative Commons
Safiah Ahmed Alzahrani, Haneen Banjar, Rsha Mirza

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(24), P. 8168 - 8168

Published: Dec. 21, 2024

This systematic review examines EEG-based imagined speech classification, emphasizing directional words essential for development in the brain–computer interface (BCI). study employed a structured methodology to analyze approaches using public datasets, ensuring evaluation and validation of results. highlights feature extraction techniques that are pivotal classification performance. These include deep learning, adaptive optimization, frequency-specific decomposition, which enhance accuracy robustness. Classification methods were explored by comparing traditional machine learning with role brain lateralization effective recognition classification. discusses challenges generalizability scalability recognition, focusing on subject-independent multiclass scalability. Performance benchmarking across various datasets methodologies revealed varied accuracies, reflecting complexity variability EEG signals. concludes remain despite progress, particularly classifying words. Future research directions improved signal processing techniques, advanced neural network architectures, more personalized, BCI systems. is critical future efforts develop practical communication tools individuals motor impairments BCIs.

Language: Английский

Citations

0
A reference free non-negative adaptive learning system for health care monitoring and adaptive physiological artifact elimination in brain waves DOI Creative Commons

Chintalpudi S.L. Prasanna,

Md. Zıa Ur Rahman

Healthcare Analytics, Journal Year: 2023, Volume and Issue: 4, P. 100225 - 100225

Published: July 8, 2023

Electroencephalogram (EEG), also referred to as brain wave (BW), is a physiological phenomenon that depicts how the human functions. Brain analysis fundamental in applications like brain-computer interference (BCI), beamforming, sleep analysis, epilepsy detection, and emotion recognition. In real-time applications, encounters many non-physiological artifacts during acquisition. Due this phenomenon, method complicated obscures wave's tiny features. This study proposes an intelligent signal enhancement unit (SEU) for processing EEG signals enable decision-making under certainty. The proposed SEU enables healthcare professionals analyze high-resolution components various applications. A new singular spectrum decomposition (SSD) based on score reconstruction (score RC) used first phase of generate artifact nature, which then reference adaptive cancellation (AAC) method. SSD performs embedding, decomposition, grouping, procedures provide signal. modified Logarithmic Non-Negative Adaptive Learning Algorithm (MLNNAL) employed second stage improve With help learning, system with lower computing complexity stable has non-negative weights can be realized. learning algorithm's weight recursion continually reweights vector each iteration eliminate from contaminated waves. Excess mean square error (EMSE), noise ratio improvement (SNRI) computational cost algorithm are evaluate performs.

Language: Английский

Citations

0
Sparsity-Enhanced Model-Based Method for Intelligent Fault Detection of Mechanical Transmission Chain in Electrical Vehicle DOI Open Access
Wangpeng He, Yue Zhou, Xiaoya Guo

et al.

Computer Modeling in Engineering & Sciences, Journal Year: 2023, Volume and Issue: 137(3), P. 2495 - 2511

Published: Jan. 1, 2023

In today's world, smart electric vehicles are deeply integrated with energy, transportation and cities.In (EVs), owing to the harsh working conditions, mechanical parts prone fatigue damages, which endanger driving safety of EVs.The practice has proved that identification periodic impact characteristics (PICs) can effectively indicate faults.This paper proposes a novel model-based approach for intelligent fault diagnosis transmission train in essential idea this lies fusion statistical information model from dynamic process.In algorithm, fractal wavelet decomposition (FWD) is used investigate time-frequency representation input signal.Based on sparsity PIC Hilbert envelope spectrum, method evaluating energy ratio (PICER) defined based an over-complete Fourier dictionary.A compound indicator considering kurtosis PICER signal designed.Using index, evaluations impulsiveness cycle-stationary process be enabled, thus avoiding serious interference sporadic during measurements.The robustness proposed noise demonstrated via numerical simulations, engineering application employed validate its effectiveness.

Language: Английский

Citations

0