Power Spectral Density Features Research Articles

Surgical Hand Gesture Recognition Utilizing Electroencephalogram as Input to the Machine Learning and Network Neuroscience Algorithms.

Surgical gestures detection can provide targeted, automated surgical skill assessment and feedback during surgical training for robot-assisted surgery (RAS). Several sources including surgical videos, robot tool kinematics, and an electromyogram (EMG) have been proposed to reach this goal. We aimed to extract features from electroencephalogram (EEG) data and use them in machine learning algorithms to classify robot-assisted surgical gestures. EEG was collected from five RAS surgeons with varying experience while performing 34 robot-assisted radical prostatectomies over the course of three years. Eight dominant hand and six non-dominant hand gesture types were extracted and synchronized with associated EEG data. Network neuroscience algorithms were utilized to extract functional brain network and power spectral density features. Sixty extracted features were used as input to machine learning algorithms to classify gesture types. The analysis of variance (ANOVA) F-value statistical method was used for feature selection and 10-fold cross-validation was used to validate the proposed method. The proposed feature set used in the extra trees (ET) algorithm classified eight gesture types performed by the dominant hand of five RAS surgeons with an accuracy of 90%, precision: 90%, sensitivity: 88%, and also classified six gesture types performed by the non-dominant hand with an accuracy of 93%, precision: 94%, sensitivity: 94%.

Big data network security index correlation measure based on the fusion of modified two order cone programming model

The existing technology can not meet the needs of large data network security index measurement in the era of big data. New theories and methods need to be explored to support the application of large data. In this paper, an improved second-order cone programming model (MTOCPM) is used as the data expression vector. The power spectral density feature of large data is extracted from a large number of noisy and fuzzy data. The second-order cone programming model of large data information flow is constructed by rough concept lattice method. Combining with the finite convergence of the second-order cone programming model algorithm, the reliability of each clustering sample is measured, and the accurate search of data clustering centre is realised. Experiments show that the method based on MTOCPM fusion can not only greatly reduce the communication cost between network nodes, but also improve the global measurement accuracy by about 10%.

Research on dynamic response characteristics and control effect of mega-sub controlled structural system under long-period ground motions

New structure system, named as Mega-Sub Controlled Structure System (MSCSS) is proposed after the combination of structural response control theory and the construction principle of megastructures. This newly proposed structure, MSCSS, is based on a new structural response control mechanism and has a significant response control effect. The response control effect of the structure under long-period ground motions is studied in this work. Firstly, Hilbert-Huang transform (HHT) is employed to analyze the time–frequency power spectral density characteristics of three typical ground motions, and their spectrum distribution is discussed. Under long- period ground motions, Single-Degree-of-Freedom system showed different trend in its response spectrum. Subsequently, a reasonable cross-section was designed for MSCSS and a finite element model was established using SAP2000 and ABAQUS software respectively and the accuracy of the MSCSS finite element model was verified. Three structural response control indexes are defined and the response control characteristics of MSCSS under long-period ground motion are explored. The results show that the response control rate of MSCSS under the action of long-period ground motions is above 10%, and the control rate of base shear force is about 20%, which reflects the good energy dissipation effect of MSCSS. In addition, this study provides a basis for the dynamic matching design between the mega-structure and substructure of MSCSS.

Performance of deer hunting optimization based deep learning algorithm for speech emotion recognition

This paper proposes a speech emotion recognition technique based on Optimized Deep Neural Network. The speech signals are denoised by presenting a novel adaptive wavelet transform with a modified galactic swarm optimization algorithm (AWT_MGSO). From the noise removed speech signals, the spectral features like LPC (Linear Prediction Coefficients), MFCC (Mel frequency cepstral coefficients), PSD (power spectral density) and prosodic features like energy, entropy, formant frequencies and pitch are extracted and certain features are selected by ASFO (Adaptive Sunflower Optimization Algorithm). The optimized DNN-DHO (Deep Neural Network with Deer Hunting Optimization Algorithm) is proposed for emotion classification. An enhanced squirrel search algorithm is proposed to update the weight in the optimized DNN_DHO classifier. In this study, all the eight emotions of the speech from RAVDESS (Ryerson Audio-Visual Database of Emotional Speech and Song) and TESS (Toronto Emotional Speech Set) databases for English and IITKGP-SEHSC (Indian Institute of Technology Kharagpur Simulated Emotion Hindi Speech Corpus) database for Hindi are classified. The experimental results are obtained and compared with the classifiers such as DNN_DHO, DNN (Deep Neural Network) and DAE (Deep Auto Encoder). The experimental results show that the proposed algorithm obtains maximum accuracy as 97.85% by the TESS dataset, 97.14% by the RAVDESS dataset and 93.75% by the IITKGP-SEHSC dataset by the DNN-HHO classifier.

Acoustic emission based condition monitoring study of piston rod seals by varying speed and pressure parameters

Hydraulic cylinders are used in a wide range of applications such as oil drilling equipment, construction vehicles and manufacturing machines. Seal failure is one of the primitive causes of failure in hydraulic cylinders, possibly leading to fluid spill, unscheduled maintenance, reduced availability and thus leading to lower productivity. Regular visual inspection of seals without affecting the productivity is difficult as the seals are placed internally in the hydraulic cylinder requiring disassembly of the piston. Therefore, condition monitoring is required to assess the current health of the seals. There have been successful attempts made in literature for the assessment of seal quality using acoustic emission-based condition monitoring. However, there have been very few studies performed to diagnose the seal failure under varying speed and pressure parameters. Therefore, this study aims at increasing the understanding of seal failure under varying speed and pressure conditions through correlation with the acoustic emission signal. Experiments were performed on a hydraulic test rig using unworn, semi-worn and worn piston rod seals. For each seal wear condition, experiments were performed for five strokes at pressure conditions of 10, 20, 30 and 40 bar and speeds of 50 mm/s and 100 mm/s. Continuous acoustic emission data were acquired during all the tests. The acoustic emission signal of each piston rod stroke was analyzed using different acoustic emission features such as power spectral density, root mean square, peak, mean frequency, median frequency and band power. From the acoustic emission analysis, by using power spectral density, mean frequency and median frequency feature it is possible to identify and segregate unworn seal, leakage due to semi-worn seal and leakage due to worn seal in the test rig. The acoustic emission-based condition monitoring methodology developed in this study lays a strong foundation for further research to develop real-time monitoring of the piston rod seal in hydraulic cylinders that are used in the offshore industry.

Affective EEG-Based Person Identification Using the Deep Learning Approach

Electroencephalography (EEG) is another mode for performing Person Identification (PI). Due to the nature of the EEG signals, EEG-based PI is typically done while the person is performing some kind of mental task, such as motor control. However, few works have considered EEG-based PI while the person is in different mental states (affective EEG). The aim of this paper is to improve the performance of affective EEG-based PI using a deep learning approach. \textcolor{red}{We proposed a cascade of deep learning using a combination of Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs)}. CNNs are used to handle the spatial information from the EEG while RNNs extract the temporal information. \textcolor{red}{We evaluated two types of RNNs, namely, Long Short-Term Memory (CNN-LSTM) and Gated Recurrent Unit (CNN-GRU). } The proposed method is evaluated on the state-of-the-art affective dataset DEAP. The results indicate that CNN-GRU and CNN-LSTM can perform PI from different affective states and reach up to 99.90--100\% mean Correct Recognition Rate (CRR), significantly outperforming a support vector machine (SVM) baseline system that uses power spectral density (PSD) features. Notably, the 100\% mean \emph{CRR} comes from only 40 subjects in DEAP dataset. To reduce the number of EEG electrodes from thirty-two to five for more practical applications, the frontal region gives the best results reaching up to 99.17\% CRR (from CNN-GRU). Amongst the two deep learning models, we find CNN-GRU to slightly outperform CNN-LSTM, while having faster training time. \textcolor{red}{Furthermore, CNN-GRU overcomes the influence of affective states in EEG-Based PI reported in the previous works.

Low-Frequency Noise Investigation of GaN/AlGaN Metal–Oxide–Semiconductor High-Electron-Mobility Field-Effect Transistor With Different Gate Length and Orientation

In this article, GaN/AlGaN metal–oxide–semiconductor high-electron-mobility field-effect transistors (MOSHEMTs) fabricated on high-resistivity Si (111) substrates have been evaluated using low-frequency (LF) noise measurement. The noise power spectral density (PSD) of devices with different lengths and channel orientations has been characterized in linear operation. No noticeable differences in the electrical and noise PSD characteristics have been observed between the GaN [ $1\overline {1}00$ ] and [ $11\overline {2}0$ ] channel orientations. While most devices are dominated by 1/ ${f}$ noise, originating from number fluctuations, for long devices ( ${L} \ge 1.1~\mu \text{m}$ ), additional generation–recombination (GR) noise has been observed, originating from traps in the GaN layer.

Condition monitoring of hydraulic cylinder seals using acoustic emissions

In recent years, there has been a growing concern for fluid spill from hydraulic cylinders in the offshore oil and gas industry. To diagnose the leakage from hydraulic cylinders, there have been attempts made in literature using fluid and pressure-based condition monitoring techniques. However, there have been limited attempts to monitor leakage from hydraulic cylinders using acoustic emissions. Therefore, in this paper, an attempt has been made to understand the fluid leakage in the hydraulic cylinder based on acoustic emissions. An experimental study was performed using a test rig (with a water-glycol as hydraulic fluid) which closely replicates the operation of a hydraulic cylinder. As piston rod seal failure is the foremost cause for leakage, experiments were performed using unworn, semi-worn, and worn piston rod seals. For each seal condition, experiments were performed for five strokes at pressure conditions of 10, 20, 30, and 40 bar. In this study, the continuous acoustic emission signal was observed for each hydraulic cylinder stroke. Acoustic emission data was analysed using different techniques such as time domain, frequency domain, and time-frequency technique. By using acoustic emission features such as root mean square (RMS), peak, skewness, median frequency, and mean frequency, it was possible to identify and separate non-leakage and leakage conditions in the test rig. By using AE bandpower and power spectral density features, it is also possible to identify the leakage due to semi-worn seal and worn seal in the test rig. This study lays a strong basis to develop a real-time monitoring technique based on acoustic emissions to monitor the health of piston rod seals used in the hydraulic cylinder in the offshore industry.

Partial directed coherence based graph convolutional neural networks for driving fatigue detection.

The mental state of a driver can be accurately and reliably evaluated by detecting the driver's electroencephalogram (EEG) signals. However, traditional machine learning and deep learning methods focus on the single electrode feature analysis and ignore the functional connection of the brain. In addition, the recent brain function connection network method needs to manually extract substantial brain network features, which results in cumbersome operation. For this reason, this paper introduces graph convolution combined with brain function connection theory into the study of mental fatigue and proposes a method for driving fatigue detection based on the partial directed coherence graph convolutional neural network (PDC-GCNN), which can analyze the characteristics of single electrodes while automatically extracting the topological features of the brain network. We designed a fatigue driving simulation experiment and collected the EEG signals. In the present work, the PDC method constructs the adjacency matrix to describe the relationship between EEG channels, and the GCNN combines single-electrode local brain area information and brain area connection information to further improve the performance of detecting fatigue states. Based on the features of differential entropy (DE) and power spectral density (PSD), the average recognition accuracy of ten-fold cross validation is 84.32% and 83.84%, respectively. For further experiments on each subject, the average recognition results are 95.24%/5.10% (PSD) and 96.01%/3.81% (DE). This research can be embedded in the vehicle driving fatigue detection system, which has practical application value.

Wet gas metering by cone throttle device with machine learning

Abstract The online measurement of gas and liquid flow rates in wet gas is of great significance. This paper presents a new method to measure gas and liquid flow rates of wet gas by combining the cone throttle device and machine learning techniques. The equivalent diameter ratio of the cone device is 0.45. Experiments are carried out in a horizontal pipe of diameter 50 mm and the operating pressure ranges from 100 kPa to 250 kPa. The working fluids are the mixture of air and water with the Lockhart-Martinelli parameter (XLM) less than 0.3. The multilayer feedforward neural network is used for developing the measurement model. The model requires representative features as inputs and uses the gas and liquid flow rates as outputs. In addition to the mean values of the permanent pressure loss and the upstream-throat differential pressure, the probability density function (PDF) and power spectral density (PSD) of the upstream-throat differential pressure fluctuation are also extracted as representative features. With the principal component analysis method, the independent PDF and PSD features are obtained. The untrained dataset is used to evaluate the performance of the neural network model. Predictions of the flow rates are in good agreement with the experiments. The mean relative errors of the gas and liquid flow rates are 0.05% and −3.66%, respectively. The results show that the proposed method is capable of establishing the implicit correlations between the characteristic parameters of wet gas and the corresponding flow rates.

Driving Drowsiness Detection with EEG Using a Modified Hierarchical Extreme Learning Machine Algorithm with Particle Swarm Optimization: A Pilot Study

Driving fatigue accounts for a large number of traffic accidents in modern life nowadays. It is therefore of great importance to reduce this risky factor by detecting the driver’s drowsiness condition. This study aimed to detect drivers’ drowsiness using an advanced electroencephalography (EEG)-based classification technique. We first collected EEG data from six healthy adults under two different awareness conditions (wakefulness and drowsiness) in a virtual driving experiment. Five different machine learning techniques, including the K-nearest neighbor (KNN), support vector machine (SVM), extreme learning machine (ELM), hierarchical extreme learning machine (H-ELM), and the proposed modified hierarchical extreme learning machine algorithm with particle swarm optimization (PSO-H-ELM), were applied to classify the subject’s drowsiness based on the power spectral density (PSD) feature extracted from the EEG data. The mean accuracies of the five classifiers were 79.31%, 79.31%, 74.08%, 81.67%, and 83.12%, respectively, demonstrating the superior performance of our new PSO-H-ELM algorithm in detecting drivers’ drowsiness, compared to the other techniques.

EEG Signal Reconstruction Using a Generative Adversarial Network With Wasserstein Distance and Temporal-Spatial-Frequency Loss.

Applications based on electroencephalography (EEG) signals suffer from the mutual contradiction of high classification performance vs. low cost. The nature of this contradiction makes EEG signal reconstruction with high sampling rates and sensitivity challenging. Conventional reconstruction algorithms lead to loss of the representative details of brain activity and suffer from remaining artifacts because such algorithms only aim to minimize the temporal mean-squared-error (MSE) under generic penalties. Instead of using temporal MSE according to conventional mathematical models, this paper introduces a novel reconstruction algorithm based on generative adversarial networks with the Wasserstein distance (WGAN) and a temporal-spatial-frequency (TSF-MSE) loss function. The carefully designed TSF-MSE-based loss function reconstructs signals by computing the MSE from time-series features, common spatial pattern features, and power spectral density features. Promising reconstruction and classification results are obtained from three motor-related EEG signal datasets with different sampling rates and sensitivities. Our proposed method significantly improves classification performances of EEG signals reconstructions with the same sensitivity and the average classification accuracy improvements of EEG signals reconstruction with different sensitivities. By introducing the WGAN reconstruction model with TSF-MSE loss function, the proposed method is beneficial for the requirements of high classification performance and low cost and is convenient for the design of high-performance brain computer interface systems.

Deep Learning for EEG-Based Preference Classification in Neuromarketing

The traditional marketing methodologies (e.g., television commercials and newspaper advertisements) may be unsuccessful at selling products because they do not robustly stimulate the consumers to purchase a particular product. Such conventional marketing methods attempt to determine the attitude of the consumers toward a product, which may not represent the real behavior at the point of purchase. It is likely that the marketers misunderstand the consumer behavior because the predicted attitude does not always reflect the real purchasing behaviors of the consumers. This research study was aimed at bridging the gap between traditional market research, which relies on explicit consumer responses, and neuromarketing research, which reflects the implicit consumer responses. The EEG-based preference recognition in neuromarketing was extensively reviewed. Another gap in neuromarketing research is the lack of extensive data-mining approaches for the prediction and classification of the consumer preferences. Therefore, in this work, a deep-learning approach is adopted to detect the consumer preferences by using EEG signals from the DEAP dataset by considering the power spectral density and valence features. The results demonstrated that, although the proposed deep-learning exhibits a higher accuracy, recall, and precision compared with the k-nearest neighbor and support vector machine algorithms, random forest reaches similar results to deep learning on the same dataset.

PSDRNN: An Efficient and Effective HAR Scheme Based on Feature Extraction and Deep Learning

Recently, in this article, due to pervasive usability, smartphone-based human activity recognition (HAR) has witnessed significant development in smart health. Meanwhile, deep recurrent neural network (DRNN) shows a strong ability to automatically extract features from the time-series data, and therefore, DRNN-based HAR schemes have achieved more effective recognition (i.e., recognition accuracy) than those adopting the traditional machine learning. However, the efficiency in training and recognition (in terms of running time) has not fully taken into account, especially for resource-constraint smartphones. To solve the above issue, we propose the PSDRNN and tri-PSDRNN schemes that employ the explicit feature extraction before DRNN. Specifically, considering that the power spectral density (PSD) feature can capture the frequency characteristics and meanwhile retain the successive time characteristics of data gathered from smartphone accelerometer, PSD feature vectors are, respectively, extracted from linear accelerations and triaxle accelerations and explicitly used as the input to the following DRNN classification model. Thorough experiments based on a real dataset demonstrate that the PSDRNN can achieve the comparable effectiveness as the xyz -DRNN (the most accurate DRNN-based HAR scheme only using acceleration data), and the average recognition and training time were reduced by 56% and 80%, respectively. Moreover, tri-PSDRNN advantages over the xyz -DRNN in terms of recognition accuracy, and the running time is still lower than the xyz -DRNN. Besides, our proposed PSDRNN scheme achieved superiority in the recognition of complex transition activities.

Odor-induced emotion recognition based on average frequency band division of EEG signals

BackgroundEmotion recognition plays a key role in multimedia. To enhance the sensation of reality, smell has been incorporated into multimedia systems because it can directly stimulate memories and trigger strong emotions. New methodFor the recognition of olfactory-induced emotions, this study explored a combination method using a support vector machine (SVM) with an average frequency band division (AFBD) method, where the AFBD method was proposed to extract the power-spectral-density (PSD) features from electroencephalogram (EEG) signals induced by smelling different odors. The so-called AFBD method means that each PSD feature was calculated based on equal frequency bandwidths, rather than the traditional EEG rhythm-based bandwidth. Thirteen odors were used to induce olfactory EEGs and their corresponding emotions. These emotions were then divided into two types of emotions, pleasure and disgust, or five types of emotions that were very unpleasant, slightly unpleasant, neutral, slightly pleasant, and very pleasant. ResultsComparison between the proposed SVM plus AFBD method and other methods found average accuracies of 98.9 % and 88.5 % for two- and five-emotion recognition, respectively. These values were considerably higher than those of other combination methods, such as the combinations of AFBD or EEG rhythm-based features with naive Bayesian, k-nearest neighbor classification, voting-extreme learning machine, and backpropagation neural network methods. ConclusionsThe SVM plus AFBD method represents a useful contribution to olfactory-induced emotion recognition. Classification of the five-emotion categories was generally inferior to the classification of the two-emotion categories, suggesting that the recognition performance decreased as the number of emotions in the category increased.

EEG signal analysis and classification on P300 speller-based BCI performance in ALS patients

Objective: the objective of the presented work is to analyse the electroencephalography signal based on brain computer interface by using P300 speller for amyotrophic lateral sclerosis (ALS) patients and perform classification on extracted features to get accuracy. Analysis/methods: amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease that involves the degeneration and death of nerve cell in the brain. It affects the process related to speech and loss of motor function in the patient. BCI technology is a communication solution for all amyotrophic lateral sclerosis (ALS) patients. The P300 speller included in the BNCI Horizon 2020 data is an application allows calculating the accuracy of classifier, which is necessary for the user to spell letters or sentences in a BCI speller paradigm. In this paper, we have extracted wavelet and power spectral density features. Association rule mining and ranking method is used for feature selection. For the classification, we have used multiple techniques and different classifiers and out of those, ten best techniques are selected based on their good performance. Finding: as a result, we get maximum 75% accuracy when we used random committee classifier.

EEG signal analysis and classification on P300 speller-based BCI performance in ALS patients

Objective: the objective of the presented work is to analyse the electroencephalography signal based on brain computer interface by using P300 speller for amyotrophic lateral sclerosis (ALS) patients and perform classification on extracted features to get accuracy. Analysis/methods: amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease that involves the degeneration and death of nerve cell in the brain. It affects the process related to speech and loss of motor function in the patient. BCI technology is a communication solution for all amyotrophic lateral sclerosis (ALS) patients. The P300 speller included in the BNCI Horizon 2020 data is an application allows calculating the accuracy of classifier, which is necessary for the user to spell letters or sentences in a BCI speller paradigm. In this paper, we have extracted wavelet and power spectral density features. Association rule mining and ranking method is used for feature selection. For the classification, we have used multiple techniques and different classifiers and out of those, ten best techniques are selected based on their good performance. Finding: as a result, we get maximum 75% accuracy when we used random committee classifier.

Electroencephalogram (EEG) stress analysis on alpha/beta ratio and theta/beta ratio

<span>This paper presents an analysis of stress feature using the power ratio of frequency bands including Alpha to Beta and Theta to Beta. In this study, electroencephalography (EEG) acquisition tool was utilized to collect brain signals from 40 subjects and objectively reflected stress features induced by virtual reality (VR) technology. The EEG signals were analyzed using Welch’s fast Fourier transform (FFT) to extract power spectral density (PSD) features which represented the power of a signal distributed over a range of frequencies. Slow wave versus fast wave (SW/FW) of EEG has been studied to discriminate stress from resting baseline. The results showed the Alpha/Beta ratio and Theta/Beta ratio are negatively correlated with stress and indicated that the power ratios can discriminate the data characteristics of brainwaves for stress assessment.</span>

Multi-objective particle swarm optimisation for mental task classification using hybrid features and hierarchical neural network classifier

Recognition of mental tasks using electroencephalograph (EEG) signals is of prime importance in man machine interface and assistive technologies. Considerably low recognition rate of mental tasks is still an issue. This work combines power spectral density (PSD) features and lazy wavelet transform (LWT) coefficients to present a new approach to feature extraction from EEG signals. A simple but novel neural network classifier called hierarchical neural network is proposed for the task recognition. A novel methodology based on multi objective particle swarm optimisation (MOPSO) to select discriminative features and the number of hidden layer nodes is proposed to improve the classification accuracy. The extracted features are presented to the hierarchical classifier to discriminate left-hand movement, right-hand movement and word generation task. The results are verified on standard brain computer interface (BCI) database and our own B-alert experimental system database. The benchmarking indicates that the proposed work outperforms the state of the art.

Recognizing Hazard Perception in a Visual Blind Area Based on EEG Features

Many potential hazards are encountered during daily driving in mixed traffic situations, and the anticipatory activity of a driver to a hazard is one of the key factors in many crashes. In a previous study using eye-tracking data, it was reliably recognized whether the eyes of a driver had become fixated or pursued hazard cues. A limitation of using eye-tracking data is that it cannot be identified whether the anticipatory activity of a driver to hazards has been activated. This study aimed to propose a method to recognize whether the psychological anticipation of a driver had been activated by a hazard cue using electroencephalogram (EEG) signals as input. Thirty-six drivers participated in a simulated driving task designed according to a standard psychological anticipatory study paradigm. Power spectral density (PSD) features were extracted from raw EEG data, and feature dimensions were reduced by principal component analysis (PCA). The results showed that when a driver detected a hazard cue, the alpha band immediately decreased, and the beta band increased approximately 300 ms after the cue appeared. Based on performance evaluation of the support vector machine (SVM), k-nearest neighbor (KNN) method, and linear discriminant analysis (LDA), SVM could detect the anticipatory activity of the driver to a potential hazard in a timely manner with an accuracy of 81%. The findings demonstrated that the hazard anticipatory activity of a driver could be recognized with EEG data as input.