Wavelet Convolution Research Articles

Spatiotemporal denoising wavelet network for infrared thermography-based machine prognostics integrating ensemble uncertainty

Infrared thermography (IRT) is increasingly deployed in noncontact condition monitoring, and further facilitates the machinery remaining useful life (RUL) prediction. For IRT-based prognostics, the interpretability of degradation features, the spatiotemporal denoising capability, and the effectiveness of uncertainty quantification are crucial issues. Therefore, this paper proposes an ensemble of deep spatiotemporal denoising wavelet networks (EDSDWN). Firstly, a 4D wavelet convolution layer (WCL) is designed to capture crucial degradation-related features with meaningful physical interpretability based on multi-resolution analysis. Secondly, noises existent in features are further filtered out in a restoration process by a proposed deep image stream denoiser (DISD) block with residual learning. A base DSDWN consisting of the WCL, the DISD, and a regressor is comprehensively constructed. Ultimately, the effective uncertainty quantification is implemented by EDSDWN adaptively fitting a RUL density with a Student-t mixture distribution. EDSDWN is constructed by employing an automatically weighted algorithm to incorporate multiple base DSDWNs. Experimental results on a thermal image dataset acquired from rotating machinery and a simulated dataset have proven that EDSDWN is tailored to IRT-based machine prognostics. This proposed method implements considerably lower prediction errors and higher effectiveness of uncertainty quantification than several advanced prognostics models.

A Waveform-Independent Measure of Recurrent Neural Activity.

Rhythmic neural activity, so-called oscillations, plays a key role in neural information transmission, processing, and storage. Neural oscillations in distinct frequency bands are central to physiological brain function, and alterations thereof have been associated with several neurological and psychiatric disorders. The most common methods to analyze neural oscillations, e.g., short-time Fourier transform or wavelet analysis, assume that measured neural activity is composed of a series of symmetric prototypical waveforms, e.g., sinusoids. However, usually, the models generating the signal, including waveform shapes of experimentally measured neural activity are unknown. Decomposing asymmetric waveforms of nonlinear origin using these classic methods may result in spurious harmonics visible in the estimated frequency spectra. Here, we introduce a new method for capturing rhythmic brain activity based on recurrences of similar states in phase-space. This method allows for a time-resolved estimation of amplitude fluctuations of recurrent activity irrespective of or specific to waveform shapes. The algorithm is derived from the well-established field of recurrence analysis, which, in comparison to Fourier-based analysis, is still very uncommon in neuroscience. In this paper, we show its advantages and limitations in comparison to short-time Fourier transform and wavelet convolution using periodic signals of different waveform shapes. Furthermore, we demonstrate its application using experimental data, i.e., intracranial and noninvasive electrophysiological recordings from the human motor cortex of one epilepsy patient and one healthy adult, respectively.

Event-related oscillations differentiate between cognitive, motor and visual impairments.

Parkinson's disease (PD) and dementia with Lewy bodies (DLB) share pathological and clinical similarities while differing in the timing and severity of motor cognitive and visual impairment. Previous EEG studies found abnormal neural oscillations in PD, mild cognitive impairment (MCI) and Alzheimer's disease, however, the electrophysiological signature of clinical symptoms is still unclear. We assessed the specificity of event-related oscillations in distinguishing between cognitive, motor and visual involvement in patients with neurodegenerative conditions. EEG was recorded during a visual oddball task in 30 PD, 28 DLB, 30 MCI patients and 32 age-matched healthy controls. Target and non-target event-related power were examined in the time-frequency domain using complex Morlet wavelet convolution and compared within and between the study groups. MCI (z = -1.8, p = 0.04, Cohen's d = -0.5) and DLB (z = -3.1, p < 0.001, d = -1.0) patients showed decreased delta-band target event-related synchronization compared to participants with normal cognition. PD (z = 1.6, p = 0.05, d = 0.5) and DLB (z = 2.7, p < 0.01, d = 0.9) patients showed decreased beta suppression compared to MCI patients and controls. DLB patients with visual hallucinations (VH) showed decreased early-alpha suppression (z = 2.08, p = 0.019, d = 3.19, AUC = 1.0 ± 0.0) compared to DLB-VH-. Decreased event-related delta-band synchronization, reflecting a decline in information processing ability, was characteristic of cognitive impairment due to any cause. Decreased event-related beta suppression, reflecting impaired execution of motor action, was specific to PD and DLB. Decreased event-related early-alpha suppression was characteristic of the presence of VH in DLB. These findings show that specific oscillations may reflect specific clinical symptoms, being a marker of network dysfunction.

An Attention-Based Wavelet Convolution Neural Network for Epilepsy EEG Classification.

As a kind of non-invasive, low-cost, and readily available brain examination, EEG has attached significance to the means of clinical diagnosis of epilepsy. However, the reading of long-term EEG records has brought a heavy burden to neurologists and experts. Therefore, automatic EEG classification for epileptic patients plays an essential role in epilepsy diagnosis and treatment. This paper proposes an Attention Mechanism-based Wavelet Convolution Neural Network for epilepsy EEG classification. Attention Mechanism-based Wavelet Convolution Neural Network firstly uses multi-scale wavelet analysis to decompose the input EEGs to obtain their components in different frequency bands. Then, these decomposed multi-scale EEGs are input into the Convolution Neural Network with an attention mechanism for further feature extraction and classification. The proposed algorithm achieves 98.89% triple classification accuracy on the Bonn EEG database and 99.70% binary classification accuracy on the Bern-Barcelona EEG database. Our experiments prove that the proposed algorithm achieves a state-of-the-art classification effect on epilepsy EEG.

Bearing fault diagnosis based on wavelet sparse convolutional network and acoustic emission compression signals.

A bearing is an important and easily damaged component of mechanical equipment. For early fault diagnosis of ball bearings, acoustic emission signals are more sensitive and less affected by mechanical background noise. To cope with the large amount of data brought by the high sampling frequency and high sampling points of acoustic emission signals, a compressed sensing processing framework is introduced to research data compression and feature extraction, and a wavelet sparse convolutional network is proposed for resolved diagnosis and evaluation. The main research objective of this paper is to maximize the compression rate of the signal under the constraint of ensuring the reconstruction error of the acoustic emission signal, which can reduce the data volume of the acoustic emission signal and reduce the pressure of data analysis for subsequent fault diagnosis. At the same time, a wide convolution kernel based on a continuous wavelet is introduced when designing the neural network, and the energy information of different frequency bands of the signal is extracted by the wavelet convolution kernel to characterize the fault characteristics of the equipment. The energy pooling layer is designed to enhance the deep mining ability of compressed features, and the regularized loss function is introduced to improve the diagnostic accuracy and robustness through feature sparseness. The experimental results show that the method can effectively extract the fault characteristics of the bearing acoustic emission signal, improve the analysis efficiency and accurately classify the bearing faults.

Fault Location with High Precision of Flexible DC Distribution System Using Wavelet Transform and Convolution Neural Network

Though flexible DC distribution system (FDCDS) is becoming a new hotspot in power systems lately because of the rapid development of power electronic devices and massive use of renewable energy, the failure to realize accurate fault location with high precision restricts its further application. Thus, a novel precise pole-to-ground fault location method of FDCDS based on wavelet transform (WT) and convolution neural network (CNN) is proposed in this paper for the limitation on the number of measuring points and high difficulty in extracting characteristics of FDCDS. The fault voltage signal is decomposed with multi-resolution by discrete wavelet transform (DWT), and then the transient energy function is constructed to select the frequency bands containing rich fault characteristics for signal reconstruction. The reconstructed signal forms two-dimensional time-frequency images through continuous wavelet transform (CWT), which are used as the input of CNN classifier after image enhancement to form the mapping relation between the fault feature and fault position using the powerful generalization ability of CNN, so as to complete fault location with high precision. The sample data on PSCAD/EMTDC verifies the accuracy and reliability of the proposed method, which can achieve fault location with positioning precision of 30 m. The proposed method overcomes the influence of the control strategy of the converter and the number of input capacitors of the bridge arm in the time-domain analysis, and still has strong robustness in the case that FDCDS is connected with many distributed generations (DGs) with output fluctuation. Furthermore, four other methods for fault location as comparisons are given to reflect the validity and anti-interference ability of proposed methods in various noises.

On the forward simulation and cost functions for the ultrasonic material characterization of polymers

AbstractIn this contribution, we describe the identification of the elastic properties of polymers in use‐specific setups, in particular the high‐frequency behavior (compared to quasi‐static or low‐frequency behavior that is usually characterized). Such a parameter search can be expressed as an inverse problem by comparing experimental measurements with adequate simulated data. We employ an axisymmetric scaled boundary finite element formulation to model the propagation of ultrasonic guided waves through cylindrical polymeric samples as the so‐called forward model of an optimization procedure. The simulated dynamic response is compared to a reference signal, and gradient descent‐based optimization is used to find the best possible material parameter fit. We present a cost function based on the wavelet convolution that results in a locally convex cost surface well suited for gradient descent optimization.

The effect of aging on inter-limb coordination: Establishing a novel transcranial alternating current stimulation (tACS) protocol by means of electroencephalography (EEG)

Abstract Healthy aging is a societal challenge with a significant socioeconomic impact. Gaining knowledge about neurodegenerative changes that impact functional independence is thus of high scientific importance. Age-related structural and functional neural changes have consequences for inter-limb coordination, which forms an intrinsic part of everyday life (e.g., driving, buttoning a shirt). Research has demonstrated that transcranial alternating current stimulation (tACS) targeting neural oscillations can alter behavioral performance. Nevertheless, profound understanding of tACS in the context of inter-limb coordination is yet to be achieved. In order to accomplish this, a more thorough understanding of the neural oscillations underlying inter-limb coordination is of utmost importance. The current study used electroencephalography (EEG) in healthy right-handed participants during two inter-limb coordination tasks; the bimanual tracking task (BTT) and the multi-limb reaction time task (MLT). The BTT consists of rotating two dials with both hands in a coordinated fashion. Three conditions were used, differing in left- and right-hand movement frequency. The MLT requires participants to lift multiple limbs in a correct manner. Six conditions were used, differing in required motor response (e.g., lifting right arm and left leg, lifting left arm and both legs). Twenty healthy right-handed young adults (age: 22.3 ± 1.0 (mean ± standard deviation)) and nineteen older adults (age: 70.7 ± 3.0) participated. Laplacian filtered EEG data will be analyzed in the time-frequency domain using morlet wavelet convolution. EEG time-frequency data between 4 and 35 hertz will be analyzed by means of repeated measures ANOVA, investigating differences across conditions and groups for each task respectively. In addition, bivariate channel-level functional connectivity will be analyzed using inter-site phase clustering. Behavioral data will be analyzed with a mixed model approach. These insights will be used to establish novel tACS protocols. If significant differences between young and older adults are present, these findings can lead to age-adapted tACS protocols. Keywords: Electroencephalography, Transcranial alternating current stimulation, Ageing, Functional connectivity

Classification of Complex Emotions Using EEG and Virtual Environment: Proof of Concept and Therapeutic Implication.

During the last decades, neurofeedback training for emotional self-regulation has received significant attention from scientific and clinical communities. Most studies have investigated emotions using functional magnetic resonance imaging (fMRI), including the real-time application in neurofeedback training. However, the electroencephalogram (EEG) is a more suitable tool for therapeutic application. Our study aims at establishing a method to classify discrete complex emotions (e.g., tenderness and anguish) elicited through a near-immersive scenario that can be later used for EEG-neurofeedback. EEG-based affective computing studies have mainly focused on emotion classification based on dimensions, commonly using passive elicitation through single-modality stimuli. Here, we integrated both passive and active elicitation methods. We recorded electrophysiological data during emotion-evoking trials, combining emotional self-induction with a multimodal virtual environment. We extracted correlational and time-frequency features, including frontal-alpha asymmetry (FAA), using Complex Morlet Wavelet convolution. Thinking about future real-time applications, we performed within-subject classification using 1-s windows as samples and we applied trial-specific cross-validation. We opted for a traditional machine-learning classifier with low computational complexity and sufficient validation in online settings, the Support Vector Machine. Results of individual-based cross-validation using the whole feature sets showed considerable between-subject variability. The individual accuracies ranged from 59.2 to 92.9% using time-frequency/FAA and 62.4 to 92.4% using correlational features. We found that features of the temporal, occipital, and left-frontal channels were the most discriminative between the two emotions. Our results show that the suggested pipeline is suitable for individual-based classification of discrete emotions, paving the way for future personalized EEG-neurofeedback training.

Hypergraph wavelet neural networks for 3D object classification

Recently, hypergraph learning has shown great potential in a variety of classification tasks. However, existing hypergraph neural networks lack flexibility in modeling and extracting high-order relationships among data. To solve this problem, we propose a novel framework called hypergraph wavelet neural networks (HGWNN) to explore the high-order correlation in 3D data. Firstly, considering the non-uniformity of most data sets in the real world, we propose a “data-driven” hypergraph construction scheme, which is more efficient than some commonly used hypergraph construction methods. Secondly, in order to efficiently learn deep embeddings from the constructed hypergraph, we propose a hypergraph wavelet convolution operator. It enables efficient information aggregation by fully exploiting the localization property of wavelets. This convolution operator is suitable for both non-uniform and uniform hypergraphs. Finally, we design a new hypergraph regularizer based on the sparse prior of wavelet coefficients to promote local smoothness and avoid network overfitting. We have conducted experiments on object classification tasks on two 3D benchmark datasets: the National Taiwan University (NTU) 3D model dataset and the ModelNet40 dataset. Experimental results demonstrate the effectiveness of the proposed method compared with the state-of-the-art methods.

Morlet Wavelet Filtering and Phase Analysis to Reduce the Limit of Detection for Thin Film Optical Biosensors

The ultimate detection limit of optical biosensors is often limited by various noise sources, including those introduced by the optical measurement setup. While sophisticated modifications to instrumentation may reduce noise, a simpler approach that can benefit all sensor platforms is the application of signal processing to minimize the deleterious effects of noise. In this work, we show that applying complex Morlet wavelet convolution to Fabry-Pérot interference fringes characteristic of thin film reflectometric biosensors effectively filters out white noise and low-frequency reflectance variations. Subsequent calculation of the average difference in extracted phase between the filtered analyte and reference signals enables a significant reduction in the limit of detection (LOD). This method is applied on experimental data sets of thin film porous silicon sensors (PSi) in buffered solution and complex media obtained from two different laboratories. The demonstrated improvement in the LOD achieved using wavelet convolution and average phase difference paves the way for PSi optical biosensors to operate with clinically relevant detection limits for medical diagnostics, environmental monitoring, and food safety.

Fault identification of rolling bearing based on multi hidden layers wavelet convolution extreme learning neural network

Fault identification of rolling bearing based on multi hidden layers wavelet convolution extreme learning neural network

Sparse-view CT reconstruction based on multi-level wavelet convolution neural network.

Sparse-view computed tomography (CT) is a recent approach to reducing the radiation dose in patients and speeding up the data acquisition. Consequently, sparse-view CT has been of particular interest among researchers within the CT community. Advanced reconstruction algorithms for sparse-view CT, such as iterative algorithms with total-variation (TV), have been studied along with the problem of increasing computational burden and the blurring of artifacts in the reconstructed images. Studies on deep-learning-based approaches applying U-NET have recently achieved remarkable outcomes in various domains including low-dose CT. In this study, we propose a new method for sparse-view CT reconstruction based on a multi-level wavelet convolutional neural network (MWCNN). First, a filtered backprojection (FBP) was used to reconstruct a sparsely sampled sinogram from 60, 120, and 180 projections. Subsequently, the sparse-view data obtained from FBP were fed to a deep-learning network, i.e., the MWCNN. Our network architecture combines a wavelet transform and modified U-NET without pooling. By replacing the pooling function with the wavelet transform, the receptive field is enlarged to improve the performance. We qualitatively and quantitatively evaluated the interpolation, iterative TV method, and standard U-NET in terms of a reduction in the streaking artifacts and a preservation of the anatomical structures. When compared with other methods, the proposed method showed the highest performance based on various evaluation parameters such as the structural similarity, root mean square error, and resolution. These results indicate that the MWCNN possesses a powerful potential for achieving a sparse-view CT reconstruction.

Unsupervised phase learning and extraction from quasiperiodic multidimensional time-series data

Automatic phase extraction is commonly done by digital markers or handcrafted feature detection that are designed specifically for one particular type of signal. A few more generalized ways are to apply Hilbert transform or complex wavelet convolution. However, they are limited to single-dimensional signals, and they require the input window to cover multiple signal periods to extract a smooth phase sequence. In this work, we propose a learning-based phase extraction method for multi-dimensional signals, consuming only unlabeled signal examples for training. A neural network architecture is designed to map a window of signal directly to a phase value under a key constraint that encourages two consecutive phase values to progress forward. The concept has been generalized to a broad range of applications with an adjustable input window, a flexible phase progression penalty design, and better training stability. The proposed method consistently outperforms the complex wavelet convolution in three synthetic signals by reducing the phase extraction error by 63% while shortening the size of the observation window by over 90%. An experiment with an augmented electrocardiogram (ECG) signal also produces a clean phase sequence with an input window that is over thirty times smaller. A qualitative test on eleven-dimensional kinematic signals from three repetitive upper limb movements has shown that the method can also capture the fluctuation in phase progression. With the flexibility, the labelless training, the accuracy, and the short observation window, this technique has a potential to be used in a variety of applications that need instantaneous phase extraction.

Some Characterizations of Wavelet Transform

In this paper, by using the theory of wavelet convolution associated with Fourier convolution, operational properties are discussed on wavelet transform of $$S_{\omega }({\mathbb {R}}), {\mathcal {D}}_{\omega }({\mathbb {R}})$$ and $${\mathcal {M}}_{c}$$ spaces. Further, Calderon’s formula is investigated for $${\mathcal {W}}_{M, a}$$ space.

Robust and Blind Multiple Image Watermarking Using CNN and DWT in Video

Digital watermarking was introduced as a result of rapid advancement of networked multimedia systems. It had been developed to enforce copyright technologies for cover of copyright possession. Due to increase in growth of internet users of networks are increasing rapidly. It has been concluded that to minimize distortions and to increase capacity, techniques in frequency domain must be combined with another technique which has high capacity and strong robustness against different types of attacks. In this paper, a robust multiple watermarking which combine Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT)and Convolution Neural Network techniques on selected middle band of the video frames is used. This methodology is considered to be robust blind watermarking because it successfully fulfills the requirement of imperceptibility and provides high robustness against a number of image-processing attacks such as Mean filtering, Median filtering, Gaussian noise, salt and pepper noise, poison noise and rotation attack. The proposed method embeds watermark by decomposing the host image. Convolution neural network calculates the weight factor for each wavelet coefficient. The watermark bits are added to the selected coefficients without any perceptual degradation for host image. The simulation is performed on MATLAB platform. The result analysis is evaluated on PSNR and MSE which is used to define robustness of the watermark that means that the watermark will not be destroyed after intentional or involuntary attacks and can still be used for certification. The analysis of the results was made with different types of attacks concluded that the proposed technique is approximately 14% efficient as compared to existing work.

Wavelet-based optical velocimetry for oil spill flow rate estimation

Flow rate estimation of an underwater oil spill is challenging due to the complex flow dynamics of an oil spill. Existing techniques such as particle image velocimetry (PIV) and feature tracking velocimetry (FTV) were unable to accurately estimate the flow rate of the oil spill in the Macondo well incident, with the most accurate technique being optical plume velocimetry (OPV). OPV estimates the velocity field based on temporal averaging algorithm, which has a better temporal resolution compared to PIV and FTV. This is useful for flow estimation where the temporal evolution of turbulent flow is significant. However, the classical temporal cross correlation algorithm used in OPV is not appropriate to deal with multi-scale flow behaviour associated with a turbulent flow such as that of oil spills. Several methods were previously used to multi-scale representation, such as Short-Time Fourier Transform, Wavelet Transform (WT) and Empirical Mode Decomposition. Among all of these methods, wavelet transform (WT) is more appropriate for turbulent flow applications, since the interesting point is the presentation of flow spectral changes over time. This is because the wavelet convolution has the ability to balance between temporal and frequency information. This paper proposed an alternative technique called wavelet-based optical velocimetry (WOV). WOV estimates the velocity field based on cross correlation of wavelet coefficients. Continuous wavelet transform was first utilized to transform two intensity signals into the wavelet domain. The Fast Fourier Transform (FFT) algorithm was then used for wavelet coefficient correlation, from which the velocity was estimated. To validate the accuracy of the WOV technique, a turbulent jet flow was experimentally simulated at Reynold's number ranging from 1847 to 11,656 at the nozzle. In our experimental work, WOV estimated the nozzle flow rate with an error of 8.3%, while the OPV estimated the nozzle flow rate with an error of 19.7%. The better estimation of WOV technique is due to its advantage to deal with the multi-scale behaviour of turbulent flow, since the WT has a multi-resolution scheme. The accuracy of WOV algorithm, however is sensitive to the selection of wavelet scaling, as well as the wavelet function. A significant influence was observed for the wavelet scaling, due to the difference of the frequency bandwidth width of WT, while the influence of wavelet function is less significant.

Precise detection of speech endpoints dynamically: A wavelet convolution based approach

Precise detection of speech endpoints is an important factor which affects the performance of the systems where speech utterances need to be extracted from the speech signal such as Automatic Speech Recognition (ASR) system. Existing endpoint detection (EPD) methods mostly uses Short-Term Energy (STE), Zero-Crossing Rate (ZCR) based approaches and their variants. But STE and ZCR based EPD algorithms often fail in the presence of Non-speech Sound Artifacts (NSAs) produced by the speakers. Pattern recognition and classification techniques are also applied but those methods require labeled data for training. In this article, a novel approach is proposed to extract speech endpoints and the algorithm is termed as Wavelet Convolution based Speech Endpoint Detection (WCSED). WCSED decomposes the speech signal into high-frequency and low-frequency components using wavelet convolution and then computes information-entropy based thresholds for the two frequency components. The low-frequency thresholds are used to extract voiced speech segments, whereas the high-frequency thresholds are used to extract the unvoiced speech segments by filtering out the NSAs. WCSED does not require any labeled data for training and can automatically extract speech segments. Experiments are carried out on two speech databases and the results are promising even in the presence of NSAs.

2190 Longitudinal changes in EEG power envelope connectivity are proportional to motor recovery in chronic stroke patients

OBJECTIVES/SPECIFIC AIMS: The objective of this study is to determine the degree to which the use of a contralesionally-controlled brain-computer interface for stroke rehabilitation drives change in interhemispheric motor cortical activity. METHODS/STUDY POPULATION: Ten chronic stroke patients were trained in the use of a brain-computer interface device for stroke recovery. Patients perform motor imagery to control the opening and closing of a motorized hand orthosis. This device was sent home with patients for 12 weeks, and patients were asked to use the device 1 hour per day, 5 days per week. The Action Research Arm Test (ARAT) was performed at 2-week intervals to assess motor function improvement. Before the active motor imagery task, patients were asked to quietly rest for 90 seconds before the task to calibrate recording equipment. EEG signals were acquired from 2 electrodes—one each centered over left and right primary motor cortex. Signals were preprocessed with a 60 Hz notch filter for environmental noise and referenced to the common average. Power envelopes for 1 Hz frequency bands (1–30 Hz) were calculated through Gabor wavelet convolution. Correlations between electrodes were then calculated for each frequency envelope on the first and last 5 runs, thus generating one correlation value per subject, per run. The chosen runs approximately correspond to the first and last week of device usage. These correlations were Fisher Z-transformed for comparison. The first and last 5 run correlations were averaged separately to estimate baseline and final correlation values. A difference was then calculated between these averages to determine correlation change for each frequency. The relationship between beta-band correlation changes (13–30 Hz) and the change in ARAT score was determined by calculating a Pearson correlation. RESULTS/ANTICIPATED RESULTS: Beta-band inter-electrode correlations tended to decrease more in patients achieving greater motor recovery (Pearson’s r=−0.68, p=0.031). A similar but less dramatic effect was observed with alpha-band (8–12 Hz) correlation changes (Pearson’s r=−0.42, p=0.22). DISCUSSION/SIGNIFICANCE OF IMPACT: The negative correlation between inter-electrode power envelope correlations in the beta frequency band and motor recovery indicates that activity in the motor cortex on each hemisphere may become more independent during recovery. The role of the unaffected hemisphere in stroke recovery is currently under debate; there is conflicting evidence regarding whether it supports or inhibits the lesioned hemisphere. These findings may support the notion of interhemispheric inhibition, as we observe less in common between activity in the 2 hemispheres in patients successfully achieving recovery. Future neuroimaging studies with greater spatial resolution than available with EEG will shed further light on changes in interhemispheric communication that occur during stroke rehabilitation.

Cortical Power-Density Changes of Different Frequency Bands in Visually Guided Associative Learning: A Human EEG-Study.

The computer-based Rutgers Acquired Equivalence test (RAET) is a widely used paradigm to test the function of subcortical structures in visual associative learning. The test consists of an acquisition (pair learning) and a test (rule transfer) phase, associated with the function of the basal ganglia and the hippocampi, respectively. Obviously, such a complex task also requires cortical involvement. To investigate the activity of different cortical areas during this test, 64-channel EEG recordings were recorded in 24 healthy volunteers. Fast-Fourier and Morlet wavelet convolution analyses were performed on the recordings. The most robust power changes were observed in the theta (4–7 Hz) and gamma (>30 Hz) frequency bands, in which significant power elevation was observed in the vast majority of the subjects, over the parieto-occipital and temporo-parietal areas during the acquisition phase. The involvement of the frontal areas in the acquisition phase was remarkably weaker. No remarkable cortical power elevations were found in the test phase. In fact, the power of the alpha and beta bands was significantly decreased over the parietooccipital areas. We conclude that the initial acquisition of the image pairs requires strong cortical involvement, but once the pairs have been learned, neither retrieval nor generalization requires strong cortical contribution.