RESUMEN
This paper introduces a method aiming at enhancing the efficacy of speaker identification systems within challenging acoustic environments characterized by noise and reverberation. The methodology encompasses the utilization of diverse feature extraction techniques, including Mel-Frequency Cepstral Coefficients (MFCCs) and discrete transforms, such as Discrete Cosine Transform (DCT), Discrete Sine Transform (DST), and Discrete Wavelet Transform (DWT). Additionally, an Artificial Neural Network (ANN) serves as the classifier for this method. Reverberation is modeled using varying-length comb filters, and its impact on pitch frequency estimation is explored via the Auto Correlation Function (ACF). This paper also contributes to the field of cancelable speaker identification in both open and reverberation environments. The proposed method depends on comb filtering at the feature level, deliberately distorting MFCCs. This distortion, incorporated within a cancelable framework, serves to obscure speaker identities, rendering the system resilient to potential intruders. Three systems are presented in this work; a reverberation-affected speaker identification system, a system depending on cancelable features through comb filtering, and a novel cancelable speaker identification system within reverbration environments. The findings revealed that, in both scenarios with and without reverberation effects, the DWT-based features exhibited superior performance within the speaker identification system. Conversely, within the cancelable speaker identification system, the DCT-based features represent the top-performing choice.
Asunto(s)
Redes Neurales de la Computación , Ruido , Acústica , Análisis de OndículasRESUMEN
Automatic seizure detection and prediction using clinical Electroencephalograms (EEGs) are challenging tasks due to factors such as low Signal-to-Noise Ratios (SNRs), high variance in epileptic seizures among patients, and limited clinical data constraints. To overcome these challenges, this paper presents two approaches for EEG signal classification. One of these approaches depends on Machine Learning (ML) tools. The used features are different types of entropy, higher-order statistics, and sub-band energies in the Hilbert Marginal Spectrum (HMS) domain. The classification is performed using Support Vector Machine (SVM), Logistic Regression (LR), and K-Nearest Neighbor (KNN) classifiers. Both seizure detection and prediction scenarios are considered. The second approach depends on spectrograms of EEG signal segments and a Convolutional Neural Network (CNN)-based residual learning model. We use 10000 spectrogram images for each class. In this approach, it is possible to perform both seizure detection and prediction in addition to a 3-state classification scenario. Both approaches are evaluated on the Children's Hospital Boston and the Massachusetts Institute of Technology (CHB-MIT) dataset, which contains 24 EEG recordings for 6 males and 18 females. The results obtained for the HMS-based model showed an accuracy of 100%. The CNN-based model achieved accuracies of 97.66%, 95.59%, and 94.51% for Seizure (S) versus Pre-Seizure (PS), Non-Seizure (NS) versus S, and NS versus S versus PS classes, respectively. These results demonstrate that the proposed approaches can be effectively used for seizure detection and prediction. They outperform the state-of-the-art techniques for automatic seizure detection and prediction. Block diagram of proposed epileptic seizure detection method using HMS with different classifiers.
Asunto(s)
Epilepsia , Convulsiones , Masculino , Niño , Femenino , Humanos , Convulsiones/diagnóstico , Epilepsia/diagnóstico , Redes Neurales de la Computación , Electroencefalografía/métodos , Máquina de Vectores de Soporte , Procesamiento de Señales Asistido por Computador , AlgoritmosRESUMEN
Recently, biometrics has become widely used in applications to verify an individual's identity. To address security issues, biometrics presents an intriguing window of opportunity to enhance the usability and security of the Internet of Things (IoT) and other systems. It can be used to secure a variety of newly emerging IoT devices. However, biometric scenarios need more protection against different hacking attempts. Various solutions are introduced to secure biometrics. Cryptosystems, cancelable biometrics, and hybrid systems are efficient solutions for template protection. The new trend in biometric authentication systems is to use bio-signals. In this paper, two proposed authentication systems are introduced based on bio-signals. One of them is unimodal, while the other is multimodal. Protected templates are obtained depending on encryption. The deoxyribonucleic acid (DNA) encryption is implemented on the obtained optical spectrograms of bio-signals. The authentication process relies on the DNA sensitivity to variations in the initial values. In the multimodal system, the singular value decomposition (SVD) algorithm is implemented to merge bio-signals. Different evaluation metrics are used to assess the performance of the proposed systems. Simulation results prove the high accuracy and efficiency of the proposed systems as the equal error rate (EER) value is close to 0 and the area under the receiver operator characteristic curve (AROC) is close to 1. The false accept rate (FAR), false reject rate (FRR), and decidability (D) are also estimated with acceptable results of 1.6 × 10-8, 9.05 × 10-6, and 29.34, respectively. Simulation results indicate the performance stability of the proposed systems in the presence of different levels of noise.
Asunto(s)
Identificación Biométrica , Biometría , Biometría/métodos , Identificación Biométrica/métodos , Algoritmos , Simulación por Computador , ADNRESUMEN
The brain tumor is an urgent malignancy caused by unregulated cell division. Tumors are classified using a biopsy, which is normally performed after the final brain surgery. Deep learning technology advancements have assisted the health professionals in medical imaging for the medical diagnosis of several symptoms. In this paper, transfer-learning-based models in addition to a Convolutional Neural Network (CNN) called BRAIN-TUMOR-net trained from scratch are introduced to classify brain magnetic resonance images into tumor or normal cases. A comparison between the pre-trained InceptionResNetv2, Inceptionv3, and ResNet50 models and the proposed BRAIN-TUMOR-net is introduced. The performance of the proposed model is tested on three publicly available Magnetic Resonance Imaging (MRI) datasets. The simulation results show that the BRAIN-TUMOR-net achieves the highest accuracy compared to other models. It achieves 100%, 97%, and 84.78% accuracy levels for three different MRI datasets. In addition, the k-fold cross-validation technique is used to allow robust classification. Moreover, three different unsupervised clustering techniques are utilized for segmentation.
Asunto(s)
Neoplasias Encefálicas , Aprendizaje Profundo , Humanos , Neoplasias Encefálicas/diagnóstico por imagen , Redes Neurales de la Computación , Encéfalo/diagnóstico por imagen , Imagen por Resonancia Magnética/métodosRESUMEN
The security issue is essential in the Internet-of-Things (IoT) environment. Biometrics play an important role in securing the emerging IoT devices, especially IoT robots. Biometric identification is an interesting candidate to improve IoT usability and security. To access and control sensitive environments like IoT, passwords are not recommended for high security levels. Biometrics can be used instead, but more protection is needed to store original biometrics away from invaders. This paper presents a cancelable multimodal biometric recognition system based on encryption algorithms and watermarking. Both voice-print and facial images are used as individual biometrics. Double Random Phase Encoding (DRPE) and chaotic Baker map are utilized as encryption algorithms. Verification is performed by estimating the correlation between registered and tested models in their cancelable format. Simulation results give Equal Error Rate (EER) values close to zero and Area under the Receiver Operator Characteristic Curve (AROC) equal to one, which indicates the high performance of the proposed system in addition to the difficulty to invert cancelable templates. Moreover, reusability and diversity of biometric templates is guaranteed.
RESUMEN
Emotion recognition is one of the trending research fields. It is involved in several applications. Its most interesting applications include robotic vision and interactive robotic communication. Human emotions can be detected using both speech and visual modalities. Facial expressions can be considered as ideal means for detecting the persons' emotions. This paper presents a real-time approach for implementing emotion detection and deploying it in the robotic vision applications. The proposed approach consists of four phases: preprocessing, key point generation, key point selection and angular encoding, and classification. The main idea is to generate key points using MediaPipe face mesh algorithm, which is based on real-time deep learning. In addition, the generated key points are encoded using a sequence of carefully designed mesh generator and angular encoding modules. Furthermore, feature decomposition is performed using Principal Component Analysis (PCA). This phase is deployed to enhance the accuracy of emotion detection. Finally, the decomposed features are enrolled into a Machine Learning (ML) technique that depends on a Support Vector Machine (SVM), k-Nearest Neighbor (KNN), Naïve Bayes (NB), Logistic Regression (LR), or Random Forest (RF) classifier. Moreover, we deploy a Multilayer Perceptron (MLP) as an efficient deep neural network technique. The presented techniques are evaluated on different datasets with different evaluation metrics. The simulation results reveal that they achieve a superior performance with a human emotion detection accuracy of 97%, which ensures superiority among the efforts in this field.
Asunto(s)
Aprendizaje Automático , Máquina de Vectores de Soporte , Algoritmos , Teorema de Bayes , Emociones , Humanos , HablaRESUMEN
Two schemes for optical wireless modulation format recognition (MFR), based on the orthogonal-triangular decomposition (OTD) and Hough transform (HT) of the constellation diagrams, are proposed in this paper. Constellation diagrams are obtained at optical signal-to-noise ratios (OSNRs) ranging from 5 to 30 dB for seven different modulation formats (2/4/8/16-PSK and 8/16/32-QAM) as images. The first scheme depends on applying the HT of the obtained images; the second scheme is based on utilization of the decomposition of each of the obtained image matrices into an orthogonal matrix (Q) and an upper triangular matrix (R) followed by the HT. Different classifiers, including AlexNet, VGG16, and VGG19, are used for the MFR task. Model setups and results are provided to study the scheme efficiency at different levels of OSNR. The proposed schemes provide unique signatures for constellation diagrams. Moreover, it reveals that the main pattern corresponding to each constellation diagram is more distinguishable for both proposed schemes at different levels of OSNR. The obtained results achieve high accuracy at low OSNR values.
RESUMEN
The focus of most research nowadays in the field of communication technology is on increasing bandwidth in wireless connectivity. Adaptive modulation can be used to enhance efficiency of communication systems. Adaptive modulation requires modulation format identification (MFI) at the receiver side to avoid the overhead required to determine the modulation type at the receiver. We present an MFI algorithm based on fan-beam projection to generate patterns from the constellation diagrams that are more discriminative. The constellation diagrams are obtained as images for eight different modulation formats (2/4/8/16 - PSK and 8/16/32/64 - QAM). Different classifiers such as AlexNet, VGG16, and VGG19 are studied and compared for the task of MFI. Evaluation of this proposed algorithm is performed by estimating the classification accuracy at different optical signal-to-noise ratios (OSNRs) ranging from 5 to 30 dB. The simulation results reveal that the proposed algorithm succeeds in identifying the wireless optical modulation format blindly with a classification accuracy up to 100% even at low OSNR values less than 8 dB compared with the related work.
RESUMEN
INTRODUCTION: In humanity's ongoing fight against its common enemy of COVID-19, researchers have been relentless in finding efficient technologies to support mitigation, diagnosis, management, contact tracing, and ultimately vaccination. OBJECTIVES: Engineers and computer scientists have deployed the potent properties of deep learning models (DLMs) in COVID-19 detection and diagnosis. However, publicly available datasets are often adulterated during collation, transmission, or storage. Meanwhile, inadequate, and corrupted data are known to impact the learnability and efficiency of DLMs. METHODS: This study focuses on enhancing previous efforts via two multimodal diagnostic systems to extract required features for COVID-19 detection using adulterated chest X-ray images. Our proposed DLM consists of a hierarchy of convolutional and pooling layers that are combined to support efficient COVID-19 detection using chest X-ray images. Additionally, a batch normalization layer is used to curtail overfitting that usually arises from the convolution and pooling (CP) layers. RESULTS: In addition to matching the performance of standard techniques reported in the literature, our proposed diagnostic systems attain an average accuracy of 98% in the detection of normal, COVID-19, and viral pneumonia cases using corrupted and noisy images. CONCLUSIONS: Such robustness is crucial for real-world applications where data is usually unavailable, corrupted, or adulterated.
RESUMEN
Electroencephalography (EEG) is among the main tools used for analyzing and diagnosing epilepsy. The manual analysis of EEG must be conducted by highly trained clinicians or neuro-physiologists; a process that is considered to have a comparatively low inter-rater agreement. Furthermore, the new data interpretation consumes an excessive amount of time and resources. Hence, an automatic seizure detection and prediction system can improve the quality of patient care in terms of shortening the diagnosis period, reducing manual errors, and automatically detecting debilitating events. Moreover, for patient treatment, it is important to alert the patients of epilepsy seizures prior to seizure occurrence. Various distinguished studies presented good solutions for two-class seizure detection problems with binary classification scenarios. To deal with these challenges, this paper puts forward effective approaches for EEG signal classification for normal, pre-ictal, and ictal activities. Three models are presented for the classification task. Two of them are patient-specific, while the third one is patient non-specific, which makes it better for the general classification tasks. The two-class classification is implemented between normal and pre-ictal activities for seizure prediction and between normal and ictal activities for seizure detection. A more generalized three-class classification framework is considered to identify all EEG signal activities. The first model depends on a Convolutional Neural Network (CNN) with residual blocks. It contains thirteen layers with four residual learning blocks. It works on spectrograms of EEG signal segments. The second model depends on a CNN with three layers. It also works on spectrograms. On the other hand, the third model depends on Phase Space Reconstruction (PSR) to eliminate the limitations of the spectrograms used in the first models. A five-layer CNN is used with this strategy. The advantage of the PSR is the direct projection from the time domain, which keeps the main trend of different signal activities. The third model deals with all signal activities, and it was tested for all patients of the CHB-MIT dataset. It has a superior performance compared to the first models and the state-of-the-art models.
Asunto(s)
Aprendizaje Profundo , Epilepsia , Electroencefalografía , Epilepsia/diagnóstico , Humanos , Redes Neurales de la Computación , Convulsiones/diagnósticoRESUMEN
The new Coronavirus disease 2019 (COVID-19) is rapidly affecting the world population with statistics quickly falling out of date. Due to the limited availability of annotated Coronavirus X-ray and CT images, the detection of COVID-19 remains the biggest challenge in diagnosing this disease. This paper provides a promising solution by proposing a COVID-19 detection system based on deep learning. The proposed deep learning modalities are based on convolutional neural network (CNN) and convolutional long short-term memory (ConvLSTM). Two different datasets are adopted for the simulation of the proposed modalities. The first dataset includes a set of CT images, while the second dataset includes a set of X-ray images. Both of these datasets consist of two categories: COVID-19 and normal. In addition, COVID-19 and pneumonia image categories are classified in order to validate the proposed modalities. The proposed deep learning modalities are tested on both X-ray and CT images as well as a combined dataset that includes both types of images. They achieved an accuracy of 100% and an F1 score of 100% in some cases. The simulation results reveal that the proposed deep learning modalities can be considered and adopted for quick COVID-19 screening.
RESUMEN
Deep learning is one of the most promising machine learning techniques that revolutionalized the artificial intelligence field. The known traditional and convolutional neural networks (CNNs) have been utilized in medical pattern recognition applications that depend on deep learning concepts. This is attributed to the importance of anomaly detection (AD) in automatic diagnosis systems. In this paper, the AD is performed on medical electroencephalography (EEG) signal spectrograms and medical corneal images for Internet of medical things (IoMT) systems. Deep learning based on the CNN models is employed for this task with training and testing phases. Each input image passes through a series of convolution layers with different kernel filters. For the classification task, pooling and fully-connected layers are utilized. Computer simulation experiments reveal the success and superiority of the proposed models for automated medical diagnosis in IoMT systems.
Asunto(s)
Inteligencia Artificial , Redes Neurales de la Computación , Simulación por Computador , Internet , Aprendizaje AutomáticoRESUMEN
High-speed wireless communication is necessary in our personal lives, in both working and living spaces. This paper presents a scheme for wireless optical modulation format recognition (MFR) based on the Hough transform (HT). The HT is used to project constellation diagrams onto another space for efficient feature extraction. Constellation diagrams are obtained at optical signal-to-noise ratios (OSNR) ranging from 5 to 30 dB for eight different modulation formats (2/4/8/16 phase-shift keying and 8/16/32/64 QAM). Different classifiers are used for the task of MFR: AlexNet, VGG16, and VGG19. A study of the effect of varying the number of samples on the accuracy of the classifiers is provided for each modulation format. To evaluate the proposed scheme, the efficiency of the three classifiers is studied at different values of OSNR. The obtained results reveal that the proposed scheme succeeds in identifying the wireless optical modulation format blindly with a classification accuracy up to 100%, even at low OSNR values less than 10 dB.
RESUMEN
In this article, we study the statistical characteristics and examine the performance of original representation and mathematical modelling of deoxyribonucleic acid (DNA) sequences. The proposed mathematical modelling approach is presented to create closed formulas for the original DNA data sequences with different methods. Accuracy of representation is studied based on evaluation metric values. The root Mean Squared Error (RMSE) and correlation coefficient (R) are used for examining the accuracy of all mathematical models to select the optimum one for DNA representation. In addition, statistical parameters such as energy, entropy, standard deviation, variance, mean, range, Mean Absolute Deviation (MAD), skewness and kurtosis are also used for the selection of the optimum model for DNA representation. Finally, spectral estimation methods are used for exon prediction, which means determination of the coding region (exon) for actual sequences and selected mathematical model: Sum of Sinusoids (SoS) with 8 terms and Gaussian with 8 terms. The exon prediction results from original DNA sequences and mathematically modelled DNA sequences coincide and ensure the success of the proposed sum-of--sinusoids for modelling of DNA sequences, while the Gaussian model is not appropriate for this task.
Asunto(s)
ADN/química , Análisis de Secuencia de ADN/estadística & datos numéricos , Secuencia de Bases , Bases de Datos de Ácidos Nucleicos , Exones/genética , Modelos EstadísticosRESUMEN
This article is mainly concerned with COVID-19 diagnosis from X-ray images. The number of cases infected with COVID-19 is increasing daily, and there is a limitation in the number of test kits needed in hospitals. Therefore, there is an imperative need to implement an efficient automatic diagnosis system to alleviate COVID-19 spreading among people. This article presents a discussion of the utilization of convolutional neural network (CNN) models with different learning strategies for automatic COVID-19 diagnosis. First, we consider the CNN-based transfer learning approach for automatic diagnosis of COVID-19 from X-ray images with different training and testing ratios. Different pre-trained deep learning models in addition to a transfer learning model are considered and compared for the task of COVID-19 detection from X-ray images. Confusion matrices of these studied models are presented and analyzed. Considering the performance results obtained, ResNet models (ResNet18, ResNet50, and ResNet101) provide the highest classification accuracy on the two considered datasets with different training and testing ratios, namely 80/20, 70/30, 60/40, and 50/50. The accuracies obtained using the first dataset with 70/30 training and testing ratio are 97.67%, 98.81%, and 100% for ResNet18, ResNet50, and ResNet101, respectively. For the second dataset, the reported accuracies are 99%, 99.12%, and 99.29% for ResNet18, ResNet50, and ResNet101, respectively. The second approach is the training of a proposed CNN model from scratch. The results confirm that training of the CNN from scratch can lead to the identification of the signs of COVID-19 disease.
Asunto(s)
COVID-19 , Aprendizaje Profundo , Prueba de COVID-19 , Humanos , Redes Neurales de la Computación , Radiografía Torácica , SARS-CoV-2RESUMEN
Corona Virus Disease 19 (COVID-19) firstly spread in China since December 2019. Then, it spread at a high rate around the world. Therefore, rapid diagnosis of COVID-19 has become a very hot research topic. One of the possible diagnostic tools is to use a deep convolution neural network (DCNN) to classify patient images. Chest X-ray is one of the most widely-used imaging techniques for classifying COVID-19 cases. This paper presents a proposed wireless communication and classification system for X-ray images to detect COVID-19 cases. Different modulation techniques are compared to select the most reliable one with less required bandwidth. The proposed DCNN architecture consists of deep feature extraction and classification layers. Firstly, the proposed DCNN hyper-parameters are adjusted in the training phase. Then, the tuned hyper-parameters are utilized in the testing phase. These hyper-parameters are the optimization algorithm, the learning rate, the mini-batch size and the number of epochs. From simulation results, the proposed scheme outperforms other related pre-trained networks. The performance metrics are accuracy, loss, confusion matrix, sensitivity, precision, F 1 score, specificity, Receiver Operating Characteristic (ROC) curve, and Area Under the Curve (AUC). The proposed scheme achieves a high accuracy of 97.8 %, a specificity of 98.5 %, and an AUC of 98.9 %.
RESUMEN
This paper presents a new trend in biometric security systems, which is cancelable multi-biometrics. In general, traditional biometric systems depend on a single biometric for identification. These traditional systems are subject to different types of attacks. In addition, a biometric signature may be lost in hacking scenarios; for example, in the case of intrusion, biometric signatures can be stolen forever. To reduce the risk of losing biometric signatures, the trend of cancelable biometrics has evolved by using either deformed or encrypted versions of biometrics for verification. In this paper, several biometric traits for the same person are treated to obtain a single cancelable template. First, optical scanning holography (OSH) is applied during the acquisition of each biometric. The resulting outputs are then compressed simultaneously to generate a unified template based on the energy compaction property of the discrete cosine transform (DCT). Hence, the OSH is used in the proposed approach as a tool to generate deformed versions of human biometrics in order to get the unified biometric template through DCT compression. With this approach, we guarantee the possibility of using multiple biometrics of the same user to increase security, as well as privacy of the new biometric template through utilization of the OSH. Simulation results prove the robustness of the proposed cancelable multi-biometric approach in noisy environments.
Asunto(s)
Biometría/métodos , Seguridad Computacional , Compresión de Datos/métodos , Holografía/métodos , Simulación por Computador , Dermatoglifia , Mano , Humanos , Iris , Curva ROCRESUMEN
Optical wireless communication (OWC) technology is one of several alternative technologies for addressing the radio frequency limitations for applications in both indoor and outdoor architectures. Indoor optical wireless systems suffer from noise and intersymbol interference (ISI). These degradations are produced by the wireless channel multipath effect, which causes data rate limitation and hence overall system performance degradation. On the other hand, outdoor OWC suffers from several physical impairments that affect transmission quality. Channel coding can play a vital role in the performance enhancement of OWC systems to ensure that data transmission is robust against channel impairments. In this paper, an efficient framework for OWC in developing African countries is introduced. It is suitable for OWC in both indoor and outdoor environments. The outdoor scenario will be suitable to wild areas in Africa. A detailed study of the system stages is presented to guarantee the suitable modulation, coding, equalization, and quality assessment scenarios for the OWC process, especially for tasks such as image and video communication. Hamming and low-density parity check coding techniques are utilized with an asymmetrically clipped DC-offset optical orthogonal frequency-division multiplexing (ADO-OFDM) scenario. The performance versus the complexity of both utilized techniques for channel coding is studied, and both coding techniques are compared at different coding rates. Another task studied in this paper is how to perform efficient adaptive channel estimation and hence equalization on the OWC systems to combat the effect of ISI. The proposed schemes for this task are based on the adaptive recursive least-squares (RLS) and the adaptive least mean squares (LMS) algorithms with activity detection guidance and tap decoupling techniques at the receiver side. These adaptive channel estimators are compared with the adaptive estimators based on the standard LMS and RLS algorithms. Moreover, this paper presents a new scenario for quality assessment of optical communication systems based on the regular transmission of images over the system and quality evaluation of these images at the receiver based on a trained convolutional neural network. The proposed OWC framework is very useful for developing countries in Africa due to its simplicity of implementation with high performance.
RESUMEN
Brain tumor is a mass of anomalous cells in the brain. Medical imagining techniques have a vital role in the diagnosis of brain tumors. Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) techniques are the most popular techniques to localize the tumor area. Brain tumor segmentation is very important for the diagnosis of tumors. In this paper, we introduce a framework to perform brain tumor segmentation, and then localize the region of the tumor, accurately. The proposed framework begins with the fusion of MR and CT images by the Non-Sub-Sampled Shearlet Transform (NSST) with the aid of the Modified Central Force Optimization (MCFO) to get the optimum fusion result from the quality metrics perspective. After that, image interpolation is applied to obtain a High-Resolution (HR) image from the Low-Resolution (LR) ones. The objective of the interpolation process is to enrich the details of the fusion result prior to segmentation. Finally, the threshold and the watershed segmentation are applied sequentially to localize the tumor region, clearly. The proposed framework enhances the efficiency of segmentation to help the specialists diagnose brain tumors.
Asunto(s)
Algoritmos , Neoplasias Encefálicas , Encéfalo/diagnóstico por imagen , Neoplasias Encefálicas/diagnóstico por imagen , Humanos , Procesamiento de Imagen Asistido por Computador , Imagen por Resonancia Magnética , Tomografía Computarizada por Rayos XRESUMEN
Automatic detection of maculopathy disease is a very important step to achieve high-accuracy results for the early discovery of the disease to help ophthalmologists to treat patients. Manual detection of diabetic maculopathy needs much effort and time from ophthalmologists. Detection of exudates from retinal images is applied for the maculopathy disease diagnosis. The first proposed framework in this paper for retinal image classification begins with fuzzy preprocessing in order to improve the original image to enhance the contrast between the objects and the background. After that, image segmentation is performed through binarization of the image to extract both blood vessels and the optic disc and then remove them from the original image. A gradient process is performed on the retinal image after this removal process for discrimination between normal and abnormal cases. Histogram of the gradients is estimated, and consequently the cumulative histogram of gradients is obtained and compared with a threshold cumulative histogram at certain bins. To determine the threshold cumulative histogram, cumulative histograms of images with exudates and images without exudates are obtained and averaged for each type, and the threshold cumulative histogram is set as the average of both cumulative histograms. Certain histogram bins are selected and thresholded according to the estimated threshold cumulative histogram, and the results are used for retinal image classification. In the second framework in this paper, a Convolutional Neural Network (CNN) is utilized to classify normal and abnormal cases.