Multi-class Classification Framework Research Articles

VisionDeep-AI: Deep learning-based retinal blood vessels segmentation and multi-class classification framework for eye diagnosis

Eye problems can lead to vision loss and have a significant impact on daily life, underscoring the critical importance of early diagnosis and treatment to prevent further damage and complications. A comprehensive segmentation-classification framework – VisionDeep-AI is developed in the proposed work for retinal vessel segmentation and multi-class classification on given fundus images. A weighted bi-directional feature pyramid network and U-Net backbone architecture-based customized deep-learning model is built to segment blood vessels which enhances feature extraction and multi-scale feature fusion. The architectural design comprises an end-to-end encoder-decoder network featuring six-depth layers with varying resolutions, enabling the extraction of high-level descriptors and lower-level, fine-grained characteristics. Furthermore, a multi-modal deep feature fusion architecture is developed for the multi-class classification of fundus images in four different categories by integrating features from segmented vessel images and raw fundus images, thereby accommodating more diversified information. A thorough analysis of VisionDeep-AI has been done on a comprehensive colour fundus image dataset. To ensure reliable results, data augmentation was done to prevent the over-fitting of the model and to enhance its generalization capability. The segmentation model performed exceptionally well and achieved a high accuracy of 97.73% and a dice coefficient of 89.90% in blood vessel segmentation from fundus images. Furthermore, multi-modal deep feature-fused classification architecture achieved a test accuracy of 81.50%, with a specificity of 93.83%. These findings showcase the robustness, generalizability and efficiency of the proposed VisionDeep-AI framework for fundus image segmentation and retinal disease classification, contributing towards advanced medical technologies and diagnosis precision.

MuLHiTA: A Novel Multiclass Classification Framework With Multibranch LSTM and Hierarchical Temporal Attention for Early Detection of Mental Stress.

Mental stress is an increasingly common psychological issue leading to diseases such as depression, addiction, and heart attack. In this study, an early detection framework based on electroencephalogram (EEG) data is developed for reducing the risk of these diseases. In existing frameworks, signals are often segmented into smaller sections prior to being input to a deep neural network. However, this approach ignores the fundamental nature of EEG signals as a carrier of valuable information (e.g., the integrity of frequency and phase, and temporal fluctuations of EEG components). As such, this type of segmenting may lead to information loss and a failure to effectively identify mental stress levels. Thus, we propose a novel multiclass classification framework termed multibranch LSTM and hierarchical temporal attention (MuLHiTA) for the early identification of mental stress levels. It specifically focuses on not only intraslice (within each slice) but also interslice (between different slices) samples in parallel. This was achieved by including two complementary branches, each of which integrated a specifically designed attention module into a bidirectional long short-term memory (BLSTM) network, enabling extraction of the most discriminative features from interslice and intraslice EEG signals simultaneously. The outputs of attention modules were then summed to obtain a feature representation that contributes to reduce overfitting and more effective multiclass classification. In addition, electrode positions were optimized using neural activity areas under high-stress conditions, thereby reducing computational costs by minimizing the number of critical electrodes. MuLHiTA was evaluated across one private [Montreal imaging stress task (MIST)] and two publicly available EEG datasets [EEG during mental arithmetic tasks (DMAT) and Simultaneous task EEG workload (STEW)]. These were divided into training and test sets using an 8:2 ratio, and the training data were further divided into training and validation sets using a fivefold cross-validation (CV) method, in which the model with the highest accuracy among the five was selected. The model was trained once more with the full training set, and the test data were then used to evaluate its performance. This approach achieved average classification accuracies of 93.58%, 91.80%, and 99.71% for the MIST, STEW, and DMAT datasets, respectively. Experimental results showed MuLHiTA was superior to state-of-the-art algorithms, including EEGNet, BLSTM, EEGLearn, convolutional neural network (CNN)-long short-term memory (LSTM), and convolutional recurrent attention model (CRAM), for multiclass classification. This demonstrates the viability of MuLHiTA for the early detection of mental stress.

Enhancing individual tree mortality mapping: The impact of models, data modalities, and classification taxonomy

Tree mortality is rapidly increasing as a result of more frequent and extensive droughts and forest fires across the globe. The increasing pace and scale of disturbance and resulting mortality necessitates the study of tree mortality at the scale at which it occurs (individual trees) to develop effective management strategies. Achieving this objective requires high-resolution aerial or satellite remote sensing data supported by field validation and efficient modeling. However, the optimal remote sensing data modalities and models for mapping overstory tree mortality remain uncertain and there is a lack of benchmark datasets for reproducible research, comparison, and model improvement. In this research, we propose a framework to generate tree mortality maps at the scale of individual trees and a labeled benchmark dataset using open-source high-resolution remote sensing data. Our benchmark dataset consists of over 15,000 image crops that have been manually labeled via multi-annotator majority for live, dead, mixed, and other classes in National Ecological Observatory Network (NEON) RGB, NEON hyperspectral, and National Agriculture Imagery Program (NAIP) RGBNIR imagery. Specifically, we present a multi-class classification framework that categorizes lidar-derived tree segments as live, dead, mixed, or other. We compared the performance of different machine learning models, remote sensing data modalities and their associated spatial resolution, spectral depth, and orthorectification using our benchmark dataset. Among various models, we found Convolutional Neural Network (CNN) models outperformed all other models with the highest accuracies and F1 scores across all data modalities. Similarly, among remote sensing data modalities, NEON RGB data at 0.1 m resolution performed best due to its high spatial resolution and high quality orthorectification. The overall accuracy of 87.27% with a macro F1 score of 0.76 was achieved using the RGB data with the CNN model. Our framework can be applied to any forest ecosystem in which lidar and raster imagery are available. We applied the framework using the best performing model and data modality to produce an individual tree mortality map over the Teakettle Experimental Forest in the southern Sierra Nevada. The framework and high-resolution mortality map can be valuable resources for forest managers and provide an invaluable basis for studying potential mortality drivers of dominant trees, forest inventory, selective logging, fire disturbance, and succession modeling.

ResNet and PCA-Based Deep Learning Scheme for Efficient Face Recognition

Face recognition is an emerging field of research in recent days. With the rise of deep learning, face recognition has become efficient and precise, creating new milestones. The performance, accuracy, and computational time of the existing schemes can be enhanced by devising a new scheme. In this context, a multiclass classification framework for face recognition using residual network (ResNet) and principal component analysis (PCA) schemes of deep learning with Dlib library is proposed in this paper. The proposed framework produces face recognition accuracy of 99.6% and a reduction of computational time with 68.03% using principal component analysis.

An adaptive multi-class imbalanced classification framework based on ensemble methods and deep network

An adaptive multi-class imbalanced classification framework based on ensemble methods and deep network

A hybrid learning framework for multi-modal facial prediction and recognition using improvised non-linear SVM classifier

Multi-modal classification plays a vital role for the real-time applications since most of the conventional models are independent of homogeneous features with single classifier. In addition, traditional directional gradient descriptors are difficult to find the heterogeneous features on different biometric classification. Feature extraction, segmentation, and multi-modal classification are the essential key factors that improve the true positive rate, error rate, and false positive rate of human-based recognition systems. In this work, a hybrid ensemble-based feature selection ranking measure, hybrid segmentation, and ensemble multi-class multi-modal classification framework on different biometric features is designed. Experimental results show that the proposed multi-class multi-modal ensemble classification framework has better optimization in terms of false positive rate, error rate, and precision than the conventional homogeneous local gradient feature extraction-based classification models on different features.

Characterization of Moving Sound Sources Direction-of-Arrival Estimation Using Different Deep Learning Architectures

Sound source localization is an important task for several applications and the use of deep learning for this task has recently become a popular research topic. While a number of previous works has focused on static sound sources, in this paper we evaluate the performance of a deep learning classification system for localization of moving sound sources. In particular, we evaluate the effect of key parameters at the levels of feature extraction (e.g., STFT parameters) and model training (e.g., neural network architectures). We evaluate the performance of different settings in terms of precision and F-score, in a multi-class multi-label classification framework. In our previous work for localization of moving sound sources, we investigated feedforward neural networks under different acoustic conditions and STFT parameters, and showed that the presence of some reverberation in the training dataset can help in achieving better detection for the direction of arrival of the sources. In this paper, we extend the work to show that (1) window size does not affect the performance of static sources but highly affects the performance of moving sources, (2) sequence length has a significant effect on the performance of recurrent architectures, and (3) a temporal convolutional neural network can outperform both recurrent and feedforward networks for moving sound sources.

A multi-class classification framework for disease screening and disease diagnosis of COVID-19 from chest X-ray images.

To accurately diagnose multiple lung diseases from chest X-rays, the critical aspect is to identify lung diseases with high sensitivity and specificity. This study proposed a novel multi-class classification framework that minimises either false positives or false negatives that is useful in computer aided diagnosis or computer aided detection respectively. To minimise false positives or false negatives, we generated respective stacked ensemble from pre-trained models and fully connected layers using selection metric and systematic method. The diversity of base classifiers was based on diverse set of false positives or false negatives generated. The proposed multi-class framework was evaluated on two chest X-ray datasets, and the performance was compared with the existing models and base classifiers. Moreover, we used LIME (Local Interpretable Model-agnostic Explanations) to locate the regions focused by the multi-class classification framework.

An investigation about the relationship between dysarthria level of speech and the neurological state of Parkinson’s patients

Parkinson’s disease (PD) is the most common neurological disorder that typically affects elderly people. In the earlier stage of disease, it has been seen that 90% of the patients develop voice disorders namely hypokinetic dysarthria. As time passes, the severity of PD increases, and patients have difficulty performing different speech tasks. During the progression of the disease, due to less control of articulatory organs such as the tongue, jaw, and lips, the quality of speech signals deteriorates. Periodic medical evaluations are very important for PD patients; however, having access to a medical appointment with a neurologist is a privilege in most countries. Considering that the speech recording process is inexpensive and very easy to do, we want to explore in this paper the suitability of mapping information of the dysarthria level into the neurological state of patients and vice versa. Three levels of severity are considered in a multiclass framework using time–frequency (TF) features and random-forest along with an Error-Correcting Output Code (ECOC) approach. The multiclass classification task based on dysarthria level is performed using the TF features with words and diadochokinetic (DDK) speech tasks. The developed model shows an unweighted average recall (UAR) of 68.49% with the DDK task /pakata/ based on m-FDA level, and 48.8% with the word /petaka/ based on the UPDRS level using the Random Forest classifier. With the aim, to evaluate the neurological states using the dysarthria level, the developed models are used to predict the MDS-UPDRS-III level of patients. The highest matching accuracy of 32% with the word /petaka/ is achieved. Similarly, the multiclass classification framework based on MDS-UPDRS-III is applied to predict the dysarthria level of patients. In this case, the highest matching accuracy of 18% was obtained with the DDK tasks /pataka/.

A Non-Exclusive Multi-Class Convolutional Neural Network for the Classification of Functional Requirements in AUTOSAR Software Requirement Specification Text

Software Requirement Specification (SRS) describes a software system to be developed that captures the functional, non-functional, and technical aspects of the stakeholder’s requirements. Retrieval and extraction of software information from SRS are essential to the development of software product line (SPL). Albeit Natural Language Processing (NLP) techniques, such as information retrieval and standard machine learning, have been advocated in the recent past as a semi-automatic means of optimising requirements specifications, they have not been widely embraced. The complexity in the organization’s information makes requirement analysis intricately a challenging task. The interdependence of subsystems and within an organisation drives this complexity. A plain multi-class classification framework may not address this issue. Hence, this paper propounds an automated non-exclusive approach for classification of functional requirements from SRS, using a deep learning framework. Specifically, Word2Vec and FastText word embeddings are utilised for document representation for training a convolutional neural network (CNN). The study was carried out by the compilation of manually categorised relevant enterprise data (AUTomotive Open System ARchitecture (AUTOSAR)), which were also employed for model training. Over a convolutional neural network, the impact of data trained with Word2Vec and FastText word embeddings from SRS documentation were compared to pre-trained word embeddings models, available online.

Analysis of French phonetic idiosyncrasies for accent recognition

Speech recognition systems have made tremendous progress since the last few decades. They have developed significantly in identifying the speech of the speaker. However, there is a scope of improvement in speech recognition systems in identifying the nuances and accents of a speaker. It is known that any specific natural language may possess at least one accent. Despite the identical word phonemic composition, if it is pronounced in different accents, we will have sound waves, which are different from each other. Differences in pronunciation, in accent and intonation of speech in general, create one of the most common problems of speech recognition. If there are a lot of accents in language we should create the acoustic model for each separately. We carry out a systematic analysis of the problem in the accurate classification of accents. We use traditional machine learning techniques and convolutional neural networks, and show that the classical techniques are not sufficiently efficient to solve this problem. Using spectrograms of speech signals, we propose a multi-class classification framework for accent recognition. In this paper, we focus our attention on the French accent. We also identify its limitation by understanding the impact of French idiosyncrasies on its spectrograms.

Generating diagnostic profiles of cognitive decline and dementia using magnetoencephalography

Accurate identification of the underlying cause(s) of cognitive decline and dementia is challenging due to significant symptomatic overlap between subtypes. This study presents a multi-class classification framework for subjects with subjective cognitive decline, mild cognitive impairment, Alzheimer's disease, dementia with Lewy bodies, fronto-temporal dementia and cognitive decline due to psychiatric illness, trained on source-localized resting-state magnetoencephalography data. Diagnostic profiles, describing probability estimates for each of the 6 diagnoses, were assigned to individual subjects.A balanced accuracy rate of 41% and multi-class area under the curve value of 0.75 were obtained for 6-class classification. Classification primarily depended on posterior relative delta, theta and beta power and amplitude-based functional connectivity in the beta and gamma frequency band. Dementia with Lewy bodies (sensitivity: 100%, precision: 20%) and Alzheimer's disease subjects (sensitivity: 51%, precision: 90%) could be classified most accurately. Fronto-temporal dementia subjects (sensitivity: 11%, precision: 3%) were most frequently misclassified.Magnetoencephalography biomarkers hold promise to increase diagnostic accuracy in a noninvasive manner. Diagnostic profiles could provide an intuitive tool to clinicians and may facilitate implementation of the classifier in the memory clinic.

Convolutional Neural Network-Based Terahertz Spectral Classification of Liquid Contraband for Security Inspection

Terahertz (THz) spectroscopy is now achieving increasing attention in security inspection owning to its non-destructiveness and deep penetrability of most packaging materials, such as leather, wood and wrapper. However, two major obstacles remain in spectral classification of liquid contraband: the complex components in some contraband and the spectral overlapping effect in similar types of contraband. In this paper, we establish two THz classification datasets and propose a real-time multi-class and multi-concentration liquid contraband spectral classification framework based on a convolutional neural network (CNN). The framework can not only identify contraband with complex components but also classify different concentrations of contraband. We also evaluate the robustness of the framework in different signal-to-noise ratio (SNR). Experimental results demonstrate that our algorithm (CNN) achieves the best performance compared with other deep learning and machine learning algorithms. It can be concluded that THz spectroscopy together with CNN is a promising technique for the classification of liquid contraband.

Predicting Learning Behavior Using Log Data in Blended Teaching

Online and offline blended teaching mode, the future trend of higher education, has recently been widely used in colleges around the globe. In the article, we conducted a study on students’ learning behavior analysis and student performance prediction based on the data about students’ behavior logs in three consecutive years of blended teaching in a college’s “Java Language Programming” course. Firstly, the data from diverse platforms such as MOOC, Rain Classroom, PTA, and cnBlog are integrated and preprocessed. Secondly, a novel multiclass classification framework, combining the genetic algorithm (GA) and the error correcting output codes (ECOC) method, is developed to predict the grade levels of students. In the framework, GA is designed to realize both the feature selection and binary classifier selection to fit the ECOC models. Finally, key factors affecting grades are identified in line with the optimal subset of features selected by GA, which can be analyzed for teaching significance. The results show that the multiclass classification algorithm designed in this article can effectively predict grades compared with other algorithms. In addition, the selected subset of features corresponding to learning behaviors is pedagogically instructive.

Learning Behavior Analysis Using Clustering and Evolutionary Error Correcting Output Code Algorithms in Small Private Online Courses

In recent years, online and offline teaching activities have been combined by the Small Private Online Course (SPOC) teaching activities, which can achieve a better teaching result. Therefore, colleges around the world have widely carried out SPOC-based blending teaching. Particularly in this year’s epidemic, the online education platform has accumulated lots of education data. In this paper, we collected the student behavior log data during the blending teaching process of the “College Information Technology Fundamentals” course of three colleges to conduct student learning behavior analysis and learning outcome prediction. Firstly, data collection and preprocessing are carried out; cluster analysis is performed by using k-means algorithms. Four typical learning behavior patterns have been obtained from previous research, and these patterns were analyzed in terms of teaching videos, quizzes, and platform visits. Secondly, a multiclass classification framework, which combines a feature selection method based on genetic algorithm (GA) with the error correcting output code (ECOC) method, is designed for training the classification model to achieve the prediction of grade levels of students. The experimental results show that the multiclass classification method proposed in this paper can effectively predict the grade of performance, with an average accuracy rate of over 75%. The research results help to implement personalized teaching for students with different grades and learning patterns.

Alzheimer’s disease multiclass diagnosis via multimodal neuroimaging embedding feature selection and fusion

Alzheimer’s disease (AD) will become a global burden in the coming decades according to the latest statistical survey. How to effectively detect AD or MCI (mild cognitive impairment) using reliable biomarkers and robust machine learning methods has become a challenging problem. In this study, we propose a novel AD multiclass classification framework with embedding feature selection and fusion based on multimodal neuroimaging. The framework has three novel aspects: (1) An l2,1-norm regularization term combined with the multiclass hinge loss is used to naturally select features across all the classes in each modality. (2) To fuse the complementary information contained in each modality, an lp-norm (1<p<∞) regularization term is introduced to combine different kernels to perform multiple kernel learning to avoid a sparse kernel coefficient distribution, thereby effectively exploiting complementary modalities. (3) A theorem that transforms the multiclass hinge loss minimization problem using the l2,1-norm and lp-norm regularizations to a previous solvable optimization problem and its proof are given. Additionally, it is theoretically proved that the optimization process converges to the global optimum. Extensive comparison experiments and analysis support the promising performance of the proposed method.

A Novel Supervised Cascaded Classifier System (SC²S) for Robust Remote Sensing Image Classification

Classification is one of the widely used techniques to quantify and mine the rich information present in hyperspectral imagery. However, the realization of trustworthy classification still continues to be a challenging task. This is due to the presence of various uncertainties, such as incomplete a priori knowledge on the actual number of classes and noise. Especially, mismatching between the number of spectral classes and the number of information classes leads to substantial omission or commission errors in the classification. In this letter, we present a new multiclass classification framework, named supervised cascaded classifier system (SC2S), that addresses the abovementioned problem by providing reliable results. The SC2S method is a two-stage cascaded classification procedure that involves quantifying the uncertainty and then classifying the samples. In the first stage, pixels for which no reliable training samples can be found in the training stage are detected. In the second stage, pixels that are matched with the respective training distribution are classified using a supervised learning algorithm, otherwise labeled as unknown. The proposed SC2S framework has been implemented on eight different classification scenarios with a varying number of unknown classes (UCs) using hyperspectral and multispectral imageries. The performance of the proposed SC2S method is compared with other widely used classifiers with and without reject-option. The experimental results indicate that our method offers superior classification results even in the case of data sets with a large number of unknown spectral classes.

Cognitive radar antenna selection via deep learning

Direction-of-arrival (DoA) estimation of targets improves with the number of elements employed by a phased array radar antenna. Since larger arrays have high associated cost, area and computational load, there is a recent interest in thinning the antenna arrays without loss of far-field DoA accuracy. In this context, a cognitive radar may deploy a full array and then select an optimal subarray to transmit and receive the signals in response to changes in the target environment. Prior works have used optimisation and greedy search methods to pick the best subarrays cognitively. In this study, deep learning is leveraged to address the antenna selection problem. Specifically, they construct a convolutional neural network (CNN) as a multi-class classification framework, where each class designates a different subarray. The proposed network determines a new array every time data is received by the radar, thereby making antenna selection a cognitive operation. Their numerical experiments show that the proposed CNN structure provides 22% better classification performance than a support vector machine and the resulting subarrays yield 72% more accurate DoA estimates than random array selections.

Online reconstruction-free single-pixel image classification

In single-pixel imaging, a series of illumination patterns are projected onto an object and the reflected or transmitted light from the object is integrated by a photodetector (the single-pixel detector). Then, from the set of received photodetector signals, the image of the object can ultimately be reconstructed. However, this reconstruction is not only computationally expensive, but also unnecessary for purposes such as image classification tasks. This work proposes a reconstruction-free multi-class image classification framework that, unlike most of the existing approaches, exploits the sequential nature of the problem. Indeed, by accumulating evidence of the sequence of scalar values, a decision is made after each measurement on whether already classifying the object being imaged, or waiting for more measurements. This online decision relies on a mechanism to achieve a recognition-delay trade-off that induces behaviours within the conservative-to-aggressive spectrum, which suit distinct requirements in different applications. Additionally, the presentation order of the illumination patterns makes a difference in terms of the reconstruction quality (if required) and classification performance when a limited number of patterns is used. Nevertheless, in many cases, simple data- and task-agnostic orders, such as random or frequency-based orders, are commonly used. To address this, a novel sparse-representation-based strategy is presented that sorts the patterns according to their general and discriminability utilities. Both, the online classification framework including the recognition-delay trade-off mechanism, and the data- and task-aware pattern ordering proposed, are experimentally assessed, with encouraging results, on the MNIST digits and CalTech 101 Silhouettes datasets.

Multi-class Support Vector Machine classifiers using intrinsic and penalty graphs

In this paper, a new multi-class classification framework incorporating geometric data relationships described in both intrinsic and penalty graphs in multi-class Support Vector Machine is proposed. Direct solutions are derived for the proposed optimization problem in both the input and arbitrary-dimensional Hilbert spaces for linear and non-linear multi-class classification, respectively. In addition, it is shown that the proposed approach constitutes a general framework for SVM-based multi-class classification exploiting geometric data relationships, which includes several SVM-based classification schemes as special cases. The power of the proposed approach is demonstrated in the problem of human action recognition in unconstrained environments, as well as in facial image and standard classification problems. Experiments indicate that by exploiting geometric data relationships described in both intrinsic and penalty graphs the SVM classification performance can be enhanced.