An Empirical Analysis of Training Algorithms of Neural Networks: A Case Study of EEG Signal Classification Using Java Framework

Mental state detection is the need of today's age due to increase in cases of mental disorders. Emotion describes the current mental state of the human being. The development of Emotion based Non Invasive Electroencephalogram brain-computer interface will be useful to analyze brain activity and to read hidden brains of people in need that most of us take for granted. The behavior of Electroencephalogram EEG signal is categorized in linear, nonlinear, stationary and non stationary. Behavioral analysis of the EEG signal is necessary to understand complex brain activity. The focus of this paper is the Initial analysis of brain EEG signal for mental state detection of human being. This paper presents initial analysis of EEG signal, databases and emotion classification system for the development of Intelligent Emotion Recognition System.

Electroencephalogram (EEG) signals are widely used in neuroscience and clinical medicine, which are non-invasive, real-time and low-cost. However, their complexity and noise interference limit their application in brain stripe recognition, emotion recognition, motor imagination, and epilepsy detection. This paper aims to explore the application of EEG signals in different fields, analyzing the role of deep learning in EEG signal classification and recognition tasks, so as to propose the challenges faced by current technologies and future development. Through the literature review, the definition and analysis of EEG signals are summarized, and its applications in brain striate recognition, emotion recognition, motor imagination and epilepsy detection are discussed in detail. Meanwhile, this paper analyzes the contributions to improving classification accuracy, feature extraction and model generalization ability. Finally, the research status of four sub-fields is discussed, and their research prospects are prospected. With the technology advancement and in-depth interdisciplinary cooperation, EEG signal processing technology will make greater contributions to human health and scientific and technological development.

In order to investigate the nonlinear relations of the electroencephalogram (EEG) signals under different brain functional states, higher-order statistics is used to study the nonlinear interrelation of the EEG components for the purpose of further understanding of the EEG generation and its construction. A parametric bispectral estimation for the analysis of EEG signals has been presented as an useful tool for detecting the nonlinearity of EEG signals. The bicoherence pattern is proposed in the paper to extract more. information beyond first and second-order statistics or spectral structure. Several EEG signals with normal subjects in different brain functional states are investigated by employing the non-Gaussian parametric model. The experimental results demonstrate that practical EEG signals provide obvious quadratic nonlinear coupling phenomena. The bicoherence structures of EEG signals is also different from that corresponding to the brain functional states. It is suggest that the bispectral analysis can be used as an effective way for nonlinear analysis and automatic classification of EEG signals and other biomedical measurements.

Depression is a common phenomenon in the present scenario. Due to the fast pace at which our lives move and immense pressure that we face adolescents, office goers and even the elders face depression. Diagnosing depression in the early curable stages is very important and may even save the life of a patient. EEG signal analysis has been used for medical research like epilepsy, sleep disorder, insomnia etc. Similarly, video signal analysis has been used for facial features detection, eye movement, emotion recognition etc. Collaborating both the methods accuracy of depression detection can be improved upon. This paper describes a novel method for combining both EEG signal analysis and facial emotion recognition through video analysis to successfully categorize depression into various levels. For this aim, power spectrum of three frequency bands (alpha, beta, and theta) and the whole bands of EEG are used as features along with standard deviation, mean and entropy.

In this study, a new combination scheme has been proposed for detecting Parkinson's disease (PD) from electroencephalogram (EEG) signal recorded from normal subjects and PD patients. The scheme is based on discrete wavelet transform (DWT), sample entropy (SampEn), and the three-way decision model in analysis of EEG signal. The EEG signal is noisy and nonstationary, and, as a consequence, it becomes difficult to distinguish it visually. However, the scheme is a well-established methodology in analysis of EEG signal in three stages. In the first stage, the DWT was applied to acquire the split frequency information; here, we use three-level DWT to decompose EEG signal into approximation and detail coefficients; in this stage, we aim to remove the useless and noise information and acquire the effective information. In the second stage, as the SampEn has advantage in analyzing the EEG signal, we use the approximation coefficient to compute the SampEn values. Finally, we detect the PD patients using three-way decision based on optimal center constructive covering algorithm (O_CCA) with the accuracy about 92.86%. Without DWT as preprocessing step, the detection rate reduces to 88.10%. Overall, the combination scheme we proposed is suitable and efficient in analyzing the EEG signal with higher accuracy.

Epilepsy is one of the most common and diverse set of chronic neurological disorders characterized by an abnormal excessive or synchronous neuronal activity in the brain that is termed "seizure", affecting about 50 million individuals worldwide. Electroencephalogram (EEG) signal processing technique plays a significant role in detection and prediction of epileptic seizure. Recently, many research works have been devoted to detect/predict of epileptic seizure based on analysis of EEG signals. Even though remarkable works have been conducted on seizure detection/prediction, experimental results are not mature enough in terms of sensitivity, specificity, and accuracy. In this paper we present a new approach for seizure detection to analysis preictal (before seizure onset) and interictal (period between seizures) EEG signals by extracting different features from gamma frequency band by decomposing the signals using discrete wavelet transformation. Note that the detection of preictal and interictal EEG signals leads to predict the epileptic seizure. Experimental results demonstrate that the propose method outperforms the state-of-the-art method in terms of sensitivity, specificity and accuracy to classify seizure by analyzing EEG signals to the benchmark dataset in different brain locations.

<p>In this paper, a wavelet-based neural network (WNN) classifier for recognizing EEG signals is implemented and tested under three sets EEG signals (healthy subjects, patients with epilepsy and patients with epileptic syndrome during the seizure). First, the Discrete Wavelet Transform (DWT) with the Multi-Resolution Analysis (MRA) is applied to decompose EEG signal at resolution levels of the components of the EEG signal (δ, θ, α, β and γ) and the Parseval-s theorem are employed to extract the percentage distribution of energy features of the EEG signal at different resolution levels. Second, the neural network (NN) classifies these extracted features to identify the EEGs type according to the percentage distribution of energy features. The performance of the proposed algorithm has been evaluated using in total 300 EEG signals. The results showed that the proposed classifier has the ability of recognizing and classifying EEG signals efficiently.</p>

A novel feature extraction called discretization-based entropy is proposed for use in the classication of EEG signals. To this end, EEG signals are decomposed into frequency subbands using the discrete wavelet transform (DWT), the coecients of these subbands are discretized into the desired number of intervals using the discretization method, the entropy values of the discretized subbands are calculated using the Shannon entropy method, and these are then ?used as the inputs of the adaptive neuro-fuzzy inference system (ANFIS). The equal width discretization (EWD) and equal frequency discretization (EFD) methods are used for the discretization. In order to evaluate their performances in terms of classication accuracy, three dierent experiments are implemented using dierent combinations of healthy segments, epileptic seizure-free segments, and epileptic seizure segments. The experiments show that the EWD-based entropy approach achieves higher classication accuracy rates than the EFD-based entropy approach. articial neural networks (ANNs) are used in the analysis of EEG signals. The adaptive neuro-fuzzy inference system (ANFIS) can be used as a classication tool for the rule-based analysis of EEG signals, since it is an

This thesis explores latent-variable probabilistic models for the analysis and classification of electroenchephalographic (EEG) signals used in Brain Computer Interface (BCI) systems. The first part of the thesis focuses on the use of probabilistic methods for classification. We begin with comparing performance between 'black-box' generative and discriminative approaches. In order to take potential advantage of the temporal nature of the EEG, we use two temporal models: the standard generative hidden Markov model, and the discriminative input-output hidden Markov model. For this latter model, we introduce a novel 'apposite' training algorithm which is of particular benefit for the type of training sequences that we use. We also asses the advantage of using these temporal probabilistic models compared with their static alternatives. We then investigate the incorporation of more specific prior information about the physical nature of EEG signals into the model structure. In particular, a common successful assumption in EEG research is that signals are generated by a linear mixing of independent sources in the brain and other external components. Such domain knowledge is conveniently introduced by using a generative model, and leads to a generative form of Independent Components Analysis (gICA). We analyze whether or not this approach is advantageous in terms of performance compared to a more standard discriminative approach, which uses domain knowledge by extracting relevant features which are subsequently fed into classifiers. The user of a BCI system may have more than one way to perform a particular mental task. Furthermore, the physiological and psychological conditions may change from one recording session and/or day to another. As a consequence, the corresponding EEG signals may change significantly. As a first attempt to deal with this effect, we use a mixture of gICA in which the EEG signal is split into different regimes, each regime corresponding to a potentially different realization of the same mental task. An arguable limitation of the gICA model is the fact that the temporal nature of the EEG signal is not taken into account. Therefore, we analyze an extension in which each hidden component is modeled with an autoregressive process. The second part of the thesis focuses on analyzing the EEG signal and, in particular, on extracting independent dynamical processes from multiple channels. In BCI research, such a decomposition technique can be applied, for example, to denoise EEG signals from artifacts and to analyze the source generators in the brain, thereby aiding the visualization and interpretation of the mental state. In order to do this, we introduce a specially constrained form of the linear Gaussian state-space model which satisfies several properties, such as flexibility in the specification of the number of recovered independent processes and the possibility to obtain processes in particular frequency ranges. We then discuss an extension of this model to the case in which we don't know a priori the correct number of hidden processes which have generated the observed time-series and the prior knowledge about their frequency content is not precise. This is achieved using an approximate variational Bayesian analysis. The resulting model can automatically determine the number and appropriate complexity of the underlying dynamics, with a preference for the simplest solution, and estimates processes with preferential spectral properties. An important contribution from our work is a novel 'sequential' algorithm for performing smoothed inference, which is numerically stable and simpler than others previously published.

Analysis of EEG signals using deep learning to highlight effects of vibration-based therapy on brain

Background: Drivers drowsiness is one of the prime reasons for road accidents. Electroencephalogram (EEG) signals provide crucial information regarding drowsy state due to neurological changes in the brain. But the complex nature of EEG signals makes it difficult to study these changes. A detailed analysis of the EEG signal can be done if it is decomposed into multi-modes. Method: In this paper, adaptive variational mode decomposition (AVMD) is used for accurate analysis and synthesis of EEG signals. The number of modes (J) and quadratic penalty factor (α) is selected adaptively to find out representative information from EEG signals. Selection of J and α is done by minimizing the reconstruction error using the Jaya optimization algorithm. Features are extracted from the adaptively decomposed modes. Entropy-based features selected by statistical analysis are classified with different classification algorithms. Eight performance parameters are evaluated to test the system's effectiveness. Results: The reconstruction error of 4.035 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-09</sup> and 1.564 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-09</sup> for the alert and drowsy state shows that the proposed method gives a better synthesis of signals. An accuracy, sensitivity, specificity, F-1 score, Kappa, false-positive rate, error, and precision of 97.19%, 97.01%, 97.46%, 0.976, 94.23%, 2.54%, 2.81%, and 98.18% shows that the proposed method provides representative modes for analysis. Conclusion: The comparison shows that AVMD is superior over conventional and existing methods by about 7% and 1%, respectively. The solution provided in this paper takes a step ahead for efficient synthesis and analysis of EEG signals to detect the drowsy state of drivers.

This paper covers the initial research and analysis of the EEG signal for the purpose of designing a neural interface for identification of the mental state. Such neural interface can be beneficial in various fields of automation and industry and can also potentially serve as a safety feature for safety critical processes. In the first section of this paper we discuss the performed experiment and also the technical means for the EEG data acquisition. In following chapter, we are describing the data itself and we are also performing the basic data analysis as well as the correlation identification. Final part of this paper we are evaluating our hypothesis to finding correlations in the dataset.

The activity of neurons inside the human brain produces electrical signals that contain frequencies. An electroencephalogram (EEG) system with a noninvasive device can record brain signals directly from the scalp, these signals are called EEG signals. In motor imaging (MI) task the human brain imagines moving a part of the body without any physical movement. Speech imagery (SI) is also a type of MI task in which the subject imagines speaking without moving the vocal organ or any other articulations. In the last two decades, Brain Computer Interface (BCI) system has been developed to analyze SI and MI tasks of human brain aiding in overcoming critical motor non-functionalities. A BCI system involves the collection, pre-processing, selection, extraction of features, and classification of EEG signals. This systematic literature review (SLR) aims to assist researchers in knowing EEG signals, non-invasive EEG devices and analyzing EEG signals by making use of ML models. This survey is divided into four subsections which explain analysis of SI task for imaging of digits, alphabets or word, MI task for visualization of a picture or a video and left-hand right-hand movement. Based on utilizations of number of channels of EEG device, accuracy of classification models is compared.

This paper mainly studies the EEG signals, eye movement signals, ECG signal analysis under natural human-computer interaction and the fusion of multiple information sources to generate decision signals to judge whether the driver is fatigued or not. It is mainly used to automatically monitor the driver's mental state in real time while the vehicle is running. Alerts when the driver just shows signs of fatigue. Forced deceleration or even forced parking in the event of a dangerous situation. The analysis of EEG signals mainly includes EEG signal acquisition, feature extraction and classification. The analysis of eye movement signals mainly includes face detection, human eye positioning, pupil and eye angle detection, etc. Integration and implementation mainly include the integration of multiple information sources, decision signal transmission, and fatigue driving monitoring system design.

This work describes the development of a computer-aided diagnostic model for the analysis and classification of EEG signals. The main objective of this study is to achieve an accurate as well as timely classification model which would help in the detection of epileptic EEG signals. This is very important as the patient suffering from epilepsy should receive proper medical attention hours before seizures occur. Thus the importance of fast and accurate analysis of different biomedical signals is growing at an ever increasing rate. In this study we have developed a feature extractor which when integrated with a classifier based on the Learning Vector Quantization (LVQ) algorithm classifies EEG signals into two categories viz. healthy and epileptic. The feature extractor uses the Hilbert Transform to convert real-time series EEG signals into an analytic signal which makes it easier to perform the requisite analysis. 5 sets of EEG signals from a publicly available EEG time series database were used to develop the proposed model on MATLAB. The average accuracy of classification of our proposed methodology is obtained to be as high as 89.31%.