Hand-crafted Feature Extraction Research Articles

Deep Transfer Learning Based Parkinson’s Disease Detection Using Optimized Feature Selection

Parkinson’s disease (PD) is one of the chronic neurological diseases whose progression is slow and symptoms have similarities with other diseases. Early detection and diagnosis of PD is crucial to prescribe proper treatment for patient’s productive and healthy lives. The disease’s symptoms are characterized by tremors, muscle rigidity, slowness in movements, balancing along with other psychiatric symptoms. The dynamics of handwritten records served as one of the dominant mechanisms which support PD detection and assessment. Several machine learning methods have been investigated for the early detection of this disease. But most of these handcrafted feature extraction techniques predominantly suffer from low performance accuracy issues. This cannot be tolerable for dealing with detection of such a chronic ailment. To this end, an efficient deep learning model is proposed which can assist to have early detection of Parkinson’s disease. The significant contribution of the proposed model is to select the most optimum features which have the effect of getting the high-performance accuracies. The feature optimization is done through genetic algorithm wherein <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$K$ </tex-math></inline-formula> -Nearest Neighbour technique. The proposed novel model results into detection accuracy higher than 95%, precision of 98%, area under curve of 0.90 with a loss of 0.12 only. The performance of proposed model is compared with some state-of-the-art machine learning and deep learning-based PD detection approaches to demonstrate the better detection ability of our model.

Wavelet Packet Decomposition-Based Multiscale CNN for Fault Diagnosis of Wind Turbine Gearbox.

This article presents an intelligent fault diagnosis method for wind turbine (WT) gearbox by using wavelet packet decomposition (WPD) and deep learning. Specifically, the vibration signals from the gearbox are decomposed using WPD and the decomposed signal components are fed into a hierarchical convolutional neural network (CNN) to extract multiscale features adaptively and classify faults effectively. The presented method combines the multiscale characteristic of WPD with the strong classification capacity of CNNs, and it does not need complex manual feature extraction steps as usually adopted in existing results. The presented CNN with multiple characteristic scales based on WPD (WPD-MSCNN) has three advantages: 1) the added WPD layer can legitimately process the nonstationary vibration data to obtain components at multiple characteristic scales adaptively, it takes full advantage of WPD and, thus, enables the CNN to extract multiscale features; 2) the WPD layer directly sends multiscale components to the hierarchical CNN to extract rich fault information effectively, and it avoids the loss of useful information due to hand-crafted feature extraction; and 3) even if the scale changes, the lengths of components remain the same, which shows that the proposed method is robust to scale uncertainties in the vibration signals. Experiments with vibration data from a production wind farm provided by a company using condition monitoring system (CMS) show that the presented WPD-MSCNN method is superior to traditional CNN and multiscale CNN (MSCNN) for fault diagnosis.

A Comprehensive Survey of RGB-Based and Skeleton-Based Human Action Recognition

With the advancement of computer vision, human action recognition (HAR) has shown its broad research worth and application prospects in a wide range of fields such as intelligent security, automatic driving and human-machine interaction. Based on the type of data captured by cameras and sensors, e.g., RGB, depth, skeleton, and infrared data, HAR methods can be classified into RGB-based and skeleton-based. RGB data is easy and inexpensive to obtain, but RGB-based methods need to cope with a large amount of irrelevant background information and are easily affected by factors such as lighting and shooting angle. The skeleton-based methods eliminate the impact of background variables and require little computational work due to their skeleton-focused features, but they lack the context data necessary for HAR. This paper gives a thorough survey of these two approaches, covering deep learning methods, handcrafted feature extraction methods, common datasets, challenges, and future research directions. The skeleton-based action recognition methods Section specifically presents the most well-liked 2D and 3D pose estimation algorithms. This survey aims to give researchers new to the area or engaged in a long-term study a selection of datasets and algorithms, as well as an overview of the present issues and expected future directions in the field.

An Automatic Premature Ventricular Contraction Recognition System Based on Imbalanced Dataset and Pre-Trained Residual Network Using Transfer Learning on ECG Signal

The development of automatic monitoring and diagnosis systems for cardiac patients over the internet has been facilitated by recent advancements in wearable sensor devices from electrocardiographs (ECGs), which need the use of patient-specific approaches. Premature ventricular contraction (PVC) is a common chronic cardiovascular disease that can cause conditions that are potentially fatal. Therefore, for the diagnosis of likely heart failure, precise PVC detection from ECGs is crucial. In the clinical settings, cardiologists typically employ long-term ECGs as a tool to identify PVCs, where a cardiologist must put in a lot of time and effort to appropriately assess the long-term ECGs which is time consuming and cumbersome. By addressing these issues, we have investigated a deep learning method with a pre-trained deep residual network, ResNet-18, to identify PVCs automatically using transfer learning mechanism. Herein, features are extracted by the inner layers of the network automatically compared to hand-crafted feature extraction methods. Transfer learning mechanism handles the difficulties of required large volume of training data for a deep model. The pre-trained model is evaluated on the Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) Arrhythmia and Institute of Cardiological Technics (INCART) datasets. First, we used the Pan-Tompkins algorithm to segment 44,103 normal and 6423 PVC beats, as well as 106,239 normal and 9987 PVC beats from the MIT-BIH Arrhythmia and IN-CART datasets, respectively. The pre-trained model employed the segmented beats as input after being converted into 2D (two-dimensional) images. The method is optimized with the using of weighted random samples, on-the-fly augmentation, Adam optimizer, and call back feature. The results from the proposed method demonstrate the satisfactory findings without the using of any complex pre-processing and feature extraction technique as well as design complexity of model. Using LOSOCV (leave one subject out cross-validation), the received accuracies on MIT-BIH and INCART are 99.93% and 99.77%, respectively, suppressing the state-of-the-art methods for PVC recognition on unseen data. This demonstrates the efficacy and generalizability of the proposed method on the imbalanced datasets. Due to the absence of device-specific (patient-specific) information at the evaluating stage on the target datasets in this study, the method might be used as a general approach to handle the situations in which ECG signals are obtained from different patients utilizing a variety of smart sensor devices.

Complex active sonar targets recognition using variable length deep convolutional neural network evolved by biogeography-based optimizer

Due to heterogeneous sound propagation conditions and fluctuating ambient noises, conventional handcrafted feature extraction methods represent poor results and high complexity in underwater sonar wave recognition tasks. In order to address these shortcomings, this paper proposes a hybrid metaheuristic deep learning-based approach. However, model depth may vary under different underwater ocean conditions. The deeper the model, the greater the number of hyperparameters, challenging the search space. It is crucial to have an efficient algorithm that can obtain an accurate model in a reasonable time. Therefore, this paper proposes the Variable-Length Habitat Biogeography-Based Optimizer (VLHBBO) to tune the hyperparameters of a deep conventional neural network. Given that there is no appropriate dataset for training the proposed model, experimental underwater scattering measurement is conducted on several target and non-target objects of the same size in the east of the Persian Gulf and the west of the Oman Sea. Furthermore, this study uses the benchmark datasets obtained from the New Array Technology III program as test datasets. The performance of the proposed model is compared to other underwater target classifiers in terms of eight metrics. The classification results indicate that the proposed VLBBO-DCNN classifier can effectively classify underwater sonar waves into relevant categories.

GRU-INC: An inception-attention based approach using GRU for human activity recognition

Human Activity Recognition (HAR) is very useful for the clinical applications, and many machine learning algorithms have been successfully implemented to achieve high-performance results. Although handcrafted feature extraction techniques were used in the past, Artificial Neural Network (ANN) is now more popular. In this work, a model has been proposed called Gated Recurrent Unit-Inception (GRU-INC) model has been proposed, which is an Inception-Attention based approach using Gated Recurrent Unit (GRU) that effectively makes use of the temporal and spatial information of the time-series data. The proposed model achieved an F1-score of 96.27%, 90.05%, 90.30%, 99.12%, and 95.99% on the publicly available datasets such as, UCI-HAR, OPPORTUNITY, PAMAP2, WISDM, and Daphnet, respectively. GRU along with Attention Mechanism (AM) was utilized for the temporal part, and Inception module along with Convolutional Block Attention Module (CBAM) was exploited for the spatial part of the model. The proposed architecture was evaluated against state-of-the-art models and similar works. It has been proved that the GRU-INC model has a higher recognition rate as well as lower computational cost. Thus our framework could be applicable in activity associated clinical and rehabilitation applications.

X-ray image analysis for osteoporosis diagnosis: From shallow to deep analysis

Recently, automated disease diagnosis based on medical images has become an integral component of digital pathology packages. Texture analysis is commonly used to address this issue, particularly in the context of estimating the osteoporosis progression in bone samples. Most research in this context uses handcrafted methods to directly extract bones image features despite the substantial correlation between sick and healthy bones, which explains the limited results. In this work, the handcrafted feature extraction method (e.g. HOG and/or LPQ) will be applied to a set of descriptors obtained from a deep analysis of bone texture images using Gabor's filter bank. In addition, the classifier automatically adjusts the Gabor filters settings, using the bat-inspired algorithm based optimization, to achieve deep analysis behavior and optimal performance. Using a typically osteoporosis database, our experimental results reveal a significant improvement over the state-of-the-art deep/handcrafted techniques, resulting in an excellent performance of 89.66% for osteoporosis diagnosis.

Diagnosis of COVID-19 Disease in Chest CT-Scan Images Based on Combination of Low-Level Texture Analysis and MobileNetV2 Features.

Since two years ago, the COVID-19 virus has spread strongly in the world and has killed more than 6 million people directly and has affected the lives of more than 500 million people. Early diagnosis of the virus can help to break the chain of transmission and reduce the death rate. In most cases, the virus spreads in the infected person's chest. Therefore, the analysis of a chest CT scan is one of the most efficient methods for diagnosing a patient. Until now, various methods have been presented to diagnose COVID-19 disease in chest CT-scan images. Most recent studies have proposed deep learning-based methods. But handcrafted features provide acceptable results in some studies too. In this paper, an innovative approach is proposed based on the combination of low-level and deep features. First of all, local neighborhood difference patterns are performed to extract handcrafted texture features. Next, deep features are extracted using MobileNetV2. Finally, a two-level decision-making algorithm is performed to improve the detection rate especially when the proposed decisions based on the two different feature set are not the same. The proposed approach is evaluated on a collected dataset of chest CT scan images from June 1, 2021, to December 20, 2021, of 238 cases in two groups of patient and healthy in different COVID-19 variants. The results show that the combination of texture and deep features can provide better performance than using each feature set separately. Results demonstrate that the proposed approach provides higher accuracy in comparison with some state-of-the-art methods in this scope.

Performance Evaluation of Feature Extraction Algorithms for Vehicle Shape Classification

Vehicle classification is a classic application of automotive image processing that is necessary for a variety of modern vehicle safety and comfort features. While machine learning-based solutions are effective in these fields and are currently employed extensively in various automotive applications. However, the most challenging aspect of automotive image processing with machine learning methods is gathering adequate quality and quantity datasets to develop such applications. Additionally, imbalanced datasets are common in multiclass automotive image processing, as is the case with the current topic of vehicle platoon management. The effectiveness of available handcrafted feature extractors and classifiers employed for vehicle class categorization varies greatly due to the effect of dataset imbalance. This study aims to examine how the performance of four prominent feature extractors alters when used with the imbalance dataset for vehicle shape classification. Also, the use of image augmentation techniques to increase the dataset size for three vehicle classes: car, bus, and truck, has been presented. Further, using the Support Vector Machine (SVM) classifier, experimental analysis was performed using feature extractors such as Histogram of Oriented Gradient (HOG), Scaled Invariant Feature Transform (SIFT), Speeded-Up Robust Feature (SURF), and Haar. Vehicle shape classification, which is an important characteristic in vehicle platoon management, has been evaluated using Receiver Operating Characteristic (ROC) for both the unbalanced dataset and the augmented dataset. The experimental results demonstrate that using the HOG feature extractor performs better when compared to SIFT, SURF, and HAAR feature extractors on the imbalance dataset. After using an image augmentation technique to add images, output performance improved significantly, with HOG output of 95%, SIFT output of 91%, SURF output of 91%, and HAAR output of 96%.

Skin Lesion Detection Using Hand-Crafted and DL-Based Features Fusion and LSTM.

The abnormal growth of cells in the skin causes two types of tumor: benign and malignant. Various methods, such as imaging and biopsies, are used by oncologists to assess the presence of skin cancer, but these are time-consuming and require extra human effort. However, some automated methods have been developed by researchers based on hand-crafted feature extraction from skin images. Nevertheless, these methods may fail to detect skin cancers at an early stage if they are tested on unseen data. Therefore, in this study, a novel and robust skin cancer detection model was proposed based on features fusion. First, our proposed model pre-processed the images using a GF filter to remove the noise. Second, the features were manually extracted by employing local binary patterns (LBP), and Inception V3 for automatic feature extraction. Aside from this, an Adam optimizer was utilized for the adjustments of learning rate. In the end, LSTM network was utilized on fused features for the classification of skin cancer into malignant and benign. Our proposed system employs the benefits of both ML- and DL-based algorithms. We utilized the skin lesion DermIS dataset, which is available on the Kaggle website and consists of 1000 images, out of which 500 belong to the benign class and 500 to the malignant class. The proposed methodology attained 99.4% accuracy, 98.7% precision, 98.66% recall, and a 98% F-score. We compared the performance of our features fusion-based method with existing segmentation-based and DL-based techniques. Additionally, we cross-validated the performance of our proposed model using 1000 images from International Skin Image Collection (ISIC), attaining 98.4% detection accuracy. The results show that our method provides significant results compared to existing techniques and outperforms them.

MULTforAD: Multimodal MRI Neuroimaging for Alzheimer’s Disease Detection Based on a 3D Convolution Model

Alzheimer’s disease (AD) is a neurological disease that affects numerous people. The condition causes brain atrophy, which leads to memory loss, cognitive impairment, and death. In its early stages, Alzheimer’s disease is tricky to predict. Therefore, treatment provided at an early stage of AD is more effective and causes less damage than treatment at a later stage. Although AD is a common brain condition, it is difficult to recognize, and its classification requires a discriminative feature representation to separate similar brain patterns. Multimodal neuroimage information that combines multiple medical images can classify and diagnose AD more accurately and comprehensively. Magnetic resonance imaging (MRI) has been used for decades to assist physicians in diagnosing Alzheimer’s disease. Deep models have detected AD with high accuracy in computing-assisted imaging and diagnosis by minimizing the need for hand-crafted feature extraction from MRI images. This study proposes a multimodal image fusion method to fuse MRI neuroimages with a modular set of image preprocessing procedures to automatically fuse and convert Alzheimer’s disease neuroimaging initiative (ADNI) into the BIDS standard for classifying different MRI data of Alzheimer’s subjects from normal controls. Furthermore, a 3D convolutional neural network is used to learn generic features by capturing AlD biomarkers in the fused images, resulting in richer multimodal feature information. Finally, a conventional CNN with three classifiers, including Softmax, SVM, and RF, forecasts and classifies the extracted Alzheimer’s brain multimodal traits from a normal healthy brain. The findings reveal that the proposed method can efficiently predict AD progression by combining high-dimensional MRI characteristics from different public sources with an accuracy range from 88.7% to 99% and outperforming baseline models when applied to MRI-derived voxel features.

Modulation spectral features for speech emotion recognition using deep neural networks

This work explores the use of constant-Q transform based modulation spectral features (CQT-MSF) for speech emotion recognition (SER). The human perception and analysis of sound comprise of two important cognitive parts: early auditory analysis and cortex-based processing. The early auditory analysis considers spectrogram-based representation whereas cortex-based analysis includes extraction of temporal modulations from the spectrogram. This temporal modulation representation of spectrogram is called modulation spectral feature (MSF). As the constant-Q transform (CQT) provides higher resolution at emotion salient low-frequency regions of speech, we find that CQT-based spectrogram, together with its temporal modulations, provides a representation enriched with emotion-specific information. We argue that CQT-MSF when used with a 2-dimensional convolutional network can provide a time-shift invariant and deformation insensitive representation for SER. Our results show that CQT-MSF outperforms standard mel-scale based spectrogram and its modulation features on two popular SER databases, Berlin EmoDB and RAVDESS. We also show that our proposed feature outperforms the shift and deformation invariant scattering transform coefficients, hence, showing the importance of joint hand-crafted and self-learned feature extraction instead of reliance on complete hand-crafted features. Finally, we perform Grad-CAM analysis to visually inspect the contribution of constant-Q modulation features over SER.

AntiMF: A deep learning framework for predicting anticancer peptides based on multi-view feature extraction.

In recent years, anticancer peptides have emerged as a new viable option in cancer therapy, with the ability to overcome the considerable side effects and poor outcomes of standard cancer therapies. Accurate anticancer peptide identification can facilitate its finding and speed up its application in treating cancer. However, many recent approaches are based on machine learning, which not only restricts the representation ability of the models but also requires a complex hand-crafted feature extraction process. Here, we propose AntiMF, a deep learning model that utilizes multi-view mechanism based on different feature extraction models. Comparative results show that our model has a better performance than the state-of-the-art methods in the prediction of anticancer peptides. By using an ensemble learning framework to extract representation, AntiMF can capture the different dimensional information, which can make model representation more complete. Moreover, we visualize what AntiMF learns on one of its ensemble models to intuitively show the effectivity of our model, indicating that AntiMF has the great potential ability to be an effective and useful model to identify anticancer peptides accurately.

Multimodal detection of epilepsy with deep neural networks

Epilepsy constitutes a chronic noncommunicable disease of the brain affecting approximately 50 million people around the world. Most of the existing research initiatives propose methods for detecting and predicting epilepsy, which rely on the extraction of handcrafted features and the train of traditional machine learning classifiers. In this paper, we present two new methods to distinguish healthy, interictal, and ictal cases without the time-consuming procedure of feature extraction. Firstly, we apply the short-time fourier transform (STFT) to the single-channel electroencephalogram (EEG) signals and construct an image consisting of three channels. This image is passed through pretrained models, including AlexNet, DenseNet201, EfficientNet, ResNet18, etc. Secondly, we introduce a multimodal deep neural network. Specifically, we pass each single-channel EEG signal through two branches of convolutional neural networks (CNNs), which can extract low and high frequency features. Also, we apply the short-time fourier transform (STFT) to the EEG signals and create an image consisting of three channels. The image is passed through a pretrained EfficientNet-B7 model. Finally, we employ a gated multimodal unit to control the importance of each modality. We evaluate the performance of the proposed model on five different cases on the EEG database of the University of Bonn and show that our introduced model achieves comparable performance to state-of-the-art approaches.

Comparison between handcraft feature extraction and methods based on Recurrent Neural Network models for gesture recognition by instrumented gloves: A case for Brazilian Sign Language Alphabet

Sign language is the communication mode used by hearing-impaired people. The development of tools for recognition of sign language is important to improve the interaction of these people in the society. This work proposes a comparison between models of gesture recognition based on instrumented glove with and without handcrafted feature extraction approach, which is a classical method for pattern recognition. For manual feature extraction approach, Forward Sequential Feature Selection (SFS) method was employed to select the best features and different classifiers were analyzed. For the non-manual feature extraction approach, Recurrent Neural Networks as Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU) models and, additionally, a proposed Ensemble GRU model were employed. The proposed Ensemble GRU is composed of three layers, and each layer is responsible to acquire knowledge of gestures’ periods: initial, gesture, and final periods. The classifiers LSTM and GRU can receive the data segments of the gestures directly, without manual feature extraction. A dataset composed by 26 gestures from alphabet of Brazilian Sign Language was explored. In the handcrafted feature extraction approach, the best result was of 88.5% accuracy, obtained with K-Nearest Neighbors model. On the opposite model, GRU and Ensemble GRU models reached 94.6% and 96.8% accuracy. It can be observed that non-manual feature extraction approach can be implemented for gesture recognition based on instrumented glove and the results can be better than conventional methods. Furthermore, the proposed Ensemble GRU model presented better result than the standard models of GRU and LSTM.

Fault diagnosis for induction generator-based wind turbine using ensemble deep learning techniques

Fault diagnosis in induction machines (IM) require effective detection at early stages to prevent permanent machine failure. This requires fast sensing of disturbances as well as efficient diagnosis methods. Conventional signal measuring techniques are usually intrusive, sensitive, with a low signal-to-noise ratio, and may show limited performance at incipient fault stages. Moreover, traditional machine learning (ML) diagnosis techniques based on handcrafted feature extraction methods suffer from limitations that can be overcome by deep learning (DL) with automatic feature extraction capabilities. In this paper, an online, non-intrusive method is proposed for fault diagnosis in a three-phase IM based on infrared imaging. The proposed method integrates three convolutional neural network (CNN)-based DL models (Inception, Xception, and MobileNet) to identify IM health status, fault types, and stator Inter-turn faults (ITF) location and severity. Deep features obtained from these CNNs are merged through Discrete Wavelet Transform (DWT) resulting in spatial-time-frequency features. This fusion reduces the input features’ size and improves diagnostic accuracy. Merged features’ length is further reduced using the Principal Component Analysis (PCA) where these reduced features are then used for classification. Compared to DL-based diagnosis methods introduced in previous work, the proposed method shows superior performance. First, it uses the IR images directly without the need for clustering and segmentation steps, thus saving time and effort. Moreover, ensembled DL is applied via combining the three CNNs benefits altogether rather than applying an individual DL method. Finally, unlike the huge features’ number exhibited by previous studies, the proposed approach uses minimal features, thus reducing classification complexity and time while maintaining 100% classification accuracy.

Image-Processing-Based Intelligent Defect Diagnosis of Rolling Element Bearings Using Spectrogram Images

Due to the excellent image recognition characteristics of convolutional neural networks (CNN), they have gained significant attention among researchers for image-processing-based defect diagnosis tasks. The use of deep CNN models for rolling element bearings’ (REBs’) defect diagnosis may be computationally expensive, and therefore may not be suitable for some applications where hardware and resources limitations exist. However, instead of using CNN models as end-to-end image classifiers, they can also be used to extract the deep features from images and those features can further be used as input to machine learning (ML) models for defect diagnosis tasks. In addition to extracting deep features using CNN models, there are also other methods for feature extraction from vibration characteristic images, such as the extraction of handcrafted features using the histogram of oriented gradients (HOG) and local binary pattern (LBP) descriptors. These features can also be used as input to classical ML models for image classification tasks. In this study, a performance comparison between all these image-processing-based defect diagnosis techniques was carried out in terms of fault detection accuracy and computational expense. Moreover, based upon the detailed comparison, a hybrid-ensemble method involving decision-level fusion is proposed, which is far less computationally expensive compared to CNN models while using them as end-to-end classifiers. The performance of all these models is also compared in the case of minimal training data availability and for diagnosis under slightly different operating conditions to ascertain their generalizability and ability to correctly diagnose despite the minimal availability of training data. The performance of the proposed hybrid-ensemble method remained outstanding for the REBs’ defect diagnosis despite the minimal of availability training data as well as the slight variation under operating conditions.

Deep features optimization based on a transfer learning, genetic algorithm, and extreme learning machine for robust content-based image retrieval.

The recent era has witnessed exponential growth in the production of multimedia data which initiates exploration and expansion of certain domains that will have an overwhelming impact on human society in near future. One of the domains explored in this article is content-based image retrieval (CBIR), in which images are mostly encoded using hand-crafted approaches that employ different descriptors and their fusions. Although utilization of these approaches has yielded outstanding results, their performance in terms of a semantic gap, computational cost, and appropriate fusion based on problem domain is still debatable. In this article, a novel CBIR method is proposed which is based on the transfer learning-based visual geometry group (VGG-19) method, genetic algorithm (GA), and extreme learning machine (ELM) classifier. In the proposed method, instead of using hand-crafted features extraction approaches, features are extracted automatically using a transfer learning-based VGG-19 model to consider both local and global information of an image for robust image retrieval. As deep features are of high dimension, the proposed method reduces the computational expense by passing the extracted features through GA which returns a reduced set of optimal features. For image classification, an extreme learning machine classifier is incorporated which is much simpler in terms of parameter tuning and learning time as compared to other traditional classifiers. The performance of the proposed method is evaluated on five datasets which highlight the better performance in terms of evaluation metrics as compared with the state-of-the-art image retrieval methods. Its statistical analysis through a nonparametric Wilcoxon matched-pairs signed-rank test also exhibits significant performance.

Nonlocal feature learning based on a variational graph auto-encoder network for small area change detection using SAR imagery

Synthetic aperture radar (SAR) image change detection is a challenging task due to inherent speckle noise, imbalanced class occurrence and the requirement for discriminative feature learning. The traditional handcrafted feature extraction and current convolution-based deep learning techniques have some advantages, but suffer from being limited to neighborhood-based spatial information. The nonlocally observable imbalance phenomenon that exists naturally in small area change detection has presented a huge challenge to methods that focus on local features only. In this paper, an unsupervised method based on a variational graph auto-encoder (VGAE) network was developed for object-based small area change detection using SAR images, with the advantages of alleviating the negative impact of class imbalance and suppressing speckle noise. The main steps include: 1) Three types of difference image (DI) are combined to establish a three-channel fused DI (TCFDI), which lays the data-level foundation for subsequent analysis. 2) Simple linear iterative clustering (SLIC) is used to divide the TCFDI into superpixels regarded as nodes. Two functions are proposed and developed to measure the similarity between nodes to build a weighted undirected graph. 3) A VGAE network is designed and trained using the graph and nodes, and high-level nonlocal feature representations of each node are extracted. The network, with a Gaussian Radial Basis Function constrained by geospatial distances, establishes the connection among nonlocal, but similar superpixels in the process of feature learning, which leads to speckle noise suppression and distinguishable features learned in latent space. The nodes are then identified as changed or unchanged classes via k-means clustering. Five real SAR datasets were used in comparative experiments. Up to 99.72% accuracy was achieved, which is superior to state-of-the-art methods that pay attention only to local information, thus, demonstrating the effectiveness and robustness of the proposed approach.

SchizoGoogLeNet: The GoogLeNet-Based Deep Feature Extraction Design for Automatic Detection of Schizophrenia.

Schizophrenia (SZ) is a severe and prolonged disorder of the human brain where people interpret reality in an abnormal way. Traditional methods of SZ detection are based on handcrafted feature extraction methods (manual process), which are tedious and unsophisticated, and also limited in their ability to balance efficiency and accuracy. To solve this issue, this study designed a deep learning-based feature extraction scheme involving the GoogLeNet model called “SchizoGoogLeNet” that can efficiently and automatically distinguish schizophrenic patients from healthy control (HC) subjects using electroencephalogram (EEG) signals with improved performance. The proposed framework involves multiple stages of EEG data processing. First, this study employs the average filtering method to remove noise and artifacts from the raw EEG signals to improve the signal-to-noise ratio. After that, a GoogLeNet model is designed to discover significant hidden features from denoised signals to identify schizophrenic patients from HC subjects. Finally, the obtained deep feature set is evaluated by the GoogleNet classifier and also some renowned machine learning classifiers to find a sustainable classification method for the obtained deep feature set. Experimental results show that the proposed deep feature extraction model with a support vector machine performs the best, producing a 99.02% correct classification rate for SZ, with an overall accuracy of 98.84%. Furthermore, our proposed model outperforms other existing methods. The proposed design is able to accurately discriminate SZ from HC, and it will be useful for developing a diagnostic tool for SZ detection.