Welch Power Spectral Density Estimate Research Articles

A fine-tuning deep residual convolutional neural network for emotion recognition based on frequency-channel matrices representation of one-dimensional electroencephalography

Emotion recognition (ER) plays a crucial role in enabling machines to perceive human emotional and psychological states, thus enhancing human-machine interaction. Recently, there has been a growing interest in ER based on electroencephalogram (EEG) signals. However, due to the noisy, nonlinear, and nonstationary properties of electroencephalography signals, developing an automatic and high-accuracy ER system is still a challenging task. In this study, a pretrained deep residual convolutional neural network model, including 17 convolutional layers and one fully connected layer with transfer learning technique in combination frequency-channel matrices (FCM) of two-dimensional data based on Welch power spectral density estimate from the one-dimensional EEG data has been proposed for improving the ER by automatically learning the underlying intrinsic features of multi-channel EEG data. The experiment result shows a mean accuracy of 93.61 ± 0.84%, a mean precision of 94.70 ± 0.60%, a mean sensitivity of 95.13 ± 1.02%, a mean specificity of 91.04 ± 1.02%, and a mean F1-score of 94.91 ± 0.68%, respectively using 5-fold cross-validation on the DEAP dataset. Meanwhile, to better explore and understand how the proposed model works, we noted that the ranking of clustering effect of FCM for the same category by employing the t-distributed stochastic neighbor embedding strategy is: softmax layer activation is the best, the middle convolutional layer activation is the second, and the early max pooling layer activation is the worst. These findings confirm the promising potential of combining deep learning approaches with transfer learning techniques and FCM for effective ER tasks.

Acoustic emissions in vertebral cortical shell failure

Understanding the initiation of bony failure is critical in assessing the progression of bone fracture and in developing injury criteria. Detection of acoustic emissions in bone can be used to identify fractures more sensitively and at an earlier inception time compared to traditional methods. However, high rate loading conditions, complex specimen-device interaction or geometry may cause other acoustic signals. Therefore, characterization of the isolated local acoustic emission response from cortical bone fracture is essential to distinguish its characteristics from other potential acoustic sources. This work develops a technique to use acoustic emission signals to determine when cortical bone failure occurs by characterization using both a Welch power spectral density estimate and a continuous wavelet transform. Isolated cortical shell specimens from thoracic vertebral bodies with attached acoustic sensors were subjected to quasistatic loading until failure. The resulting acoustic emissions had a wideband frequency response with peaks from 20 to 900 kHz, with the spectral peaks clustered in three bands of frequencies (166 ± 52.6 kHz, 379 ± 37.2 kHz, and 668 ± 63.4 kHz). Using these frequency bands, acoustic emissions can be used as a monitoring tool in biomechanical spine testing, distinguishing bone failure from structural response. This work presents a necessary set of techniques for effectively utilizing acoustic emissions to determine the onset of cortical bone fracture in biological material testing. Acoustic signatures can be developed for other cortical bone regions of interest using the presented methods.

Damage Detection in CFST Column by Simulation of Ultrasonic Waves Using STFT-Based Spectrogram and Welch Power Spectral Density Estimate

Damage Detection in CFST Column by Simulation of Ultrasonic Waves Using STFT-Based Spectrogram and Welch Power Spectral Density Estimate

State assessment for bearing rotor static unbalance based on Welch-PSD and SAE

Bearings are an important part of mechanical equipment and it will cause a series of mechanical failures once the malfunction of bearing occurs. Rotor unbalance is the most common type of bearing failure; thus the assessment of bearing rotor unbalance is essential to maintain the normal operation of mechanical. In this paper, a method based on Welch power spectral density estimate (Welch-PSD) and stacked automatic encoder (SAE) is proposed to achieve state assessment of bearing rotor static unbalance by processing the two-way vibration signals collected by the acceleration sensor installed in the vertical and horizontal directions of the bearing. Firstly, the Welch-PSD method is used to decompose the vibration signal to obtain the power spectral density, and the vibration power of the working frequency is taken as the feature. Then, the Stacked Auto-Encoder method is introduced to assessment the bearing rotor unbalance state. This paper designs an experiment of rotor unbalance fault in different degree to verify the accuracy of the designed method. The experimental results show that the Welch-PSD method can accurately extract the rotor unbalance fault feature. In addition, the SAE neural network can apply the fault feature to accurately assessment the bearing rotor unbalance degree.

Automatic seizure detection system with low complex PSD using TMS320C6713 DSP

Epilepsy is one of the most common neurological disorders characterised by a sudden and recurrent malfunction of the brain, a 'seizure'. An electroencephalogram (EEG) has been an especially valuable clinical tool for the evaluation and treatment of epilepsy. This paper proposes an automatic and feasible real-time seizure detection system for epileptic patients. The seizure detection system use Power Spectral Density (PSD) of EEG sub-bands as features. A low complex PSD estimate, with 30% reduction in computation time compared to traditional Welch PSD estimate is used in this work. Mahalanobis Distance Based Classifier (MDBC) is used in the classification stage to classify between seizures and non-seizures. The EEG dataset used in this work is obtained from Department of Epileptology, University of Bonn, Germany. The proposed automatic seizure detection system is implemented in TI's TMS320C6713 DSP processor and obtained the classification accuracy of higher than 95%.

English

A digestive system disorders are the very important problem in human being universally. As a panacea, a preliminary investigation method is proposed before the endoscopy procedure is called Electrogastrography. Electrogastrogram (EGG) is obtained from the human being cutaneously, which is non- invasive, easy procedure over endoscopy. EGG recorded for a 180 human being includes the normal individual and patients with digestive system disorder such us Nausea, Dyspepsia, Vomiting, ulcer, etc. A database is created for the analysis purpose. During the acquisition of EGG, the analog signal's numeric values are recorded as .bio file format using the data scope. Wavelet transform is used for the analysis of EGG to find the deviation in frequency and Power from the power spectrum estimate plot obtained for disorders patients compare to normal Individual. Wavelet Transform analysis includes the principle component analysis, denosing of the signal and power spectrum estimation. In the proposed wavelet analysis, the data of a patient or normal individual is included for analysis. The EGG signal is reconstructed with a data obtained from the data scope. This signal undergoes principle component analysis to remove the noise in the EGG signal to obtain de-noised signal and it is plotted for power spectral density estimation with Welch power spectral Density Estimation. As a result of the proposed method the power variation is found to be a range of 40-52dB and frequency is detected as 0.06 Hz to 0.075Hz for ulcer patient 0.02 Hz to 0.04 Hz for dyspepsia patient and 0.05Hz to 0.058 Hz for Normal Individual. The above said finding supports the physician in the diagnosis of digestive system disorders at fair amount of accuracy.

Automatic detection of buried utilities and solid objects with GPR using neural networks and pattern recognition

The task of locating buried utilities using ground penetrating radar is addressed, and a novel processing technique computationally suitable for on-site imaging is proposed. The developed system comprises a neural network classifier, a pattern recognition stage, and additional pre-processing, feature-extraction and image processing stages. Automatic selection of the areas of the radargram containing useful information results in a reduced data set and hence a reduction in computation time. A backpropagation neural network is employed to identify portions of the radar image corresponding to target reflections by training it to recognise the Welch power spectral density estimate of signal segments reflected from various types of buried target. This results in a classification of the radargram into useful and redundant sections, and further processing is performed only on the former. The Hough Transform is then applied to the edges of these reflections, in order to accurately identify the depth and position of the buried targets. This allows a high resolution reconstruction of the subsurface with reduced computation time. The system was tested on data containing pipes, cables and anti-personnel landmines, and the results indicate that automatic and effective detection and mapping of such structures can be achieved in near real-time.