Long Short-term Memory Classifier Research Articles

Arabic Spam Tweets Classification: A Comprehensive Machine Learning Approach

Nowadays, one of the most common problems faced by Twitter (also known as X) users, including individuals as well as organizations, is dealing with spam tweets. The problem continues to proliferate due to the increasing popularity and number of users of social media platforms. Due to this overwhelming interest, spammers can post texts, images, and videos containing suspicious links that can be used to spread viruses, rumors, negative marketing, and sarcasm, and potentially hack the user’s information. Spam detection is among the hottest research areas in natural language processing (NLP) and cybersecurity. Several studies have been conducted in this regard, but they mainly focus on the English language. However, Arabic tweet spam detection still has a long way to go, especially emphasizing the diverse dialects other than modern standard Arabic (MSA), since, in the tweets, the standard dialect is seldom used. The situation demands an automated, robust, and efficient Arabic spam tweet detection approach. To address the issue, in this research, various machine learning and deep learning models have been investigated to detect spam tweets in Arabic, including Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (NB) and Long-Short Term Memory (LSTM). In this regard, we have focused on the words as well as the meaning of the tweet text. Upon several experiments, the proposed models have produced promising results in contrast to the previous approaches for the same and diverse datasets. The results showed that the RF classifier achieved 96.78% and the LSTM classifier achieved 94.56%, followed by the SVM classifier that achieved 82% accuracy. Further, in terms of F1-score, there is an improvement of 21.38%, 19.16% and 5.2% using RF, LSTM and SVM classifiers compared to the schemes with same dataset.

Research on error detection in English translation texts using machine learning algorithms

This paper provides a concise overview of a grammatical error correction algorithm based on the encoder-decoder structure. The traditional unidirectional long short-term memory (LSTM) in the encoder was transformed into a bidirectional LSTM. Subsequently, the grammatical error correction algorithm was simulated and experimented with. In the experiments, it was compared with two other error correction algorithms: one based on the LSTM classification model and the other on the traditional unidirectional LSTM translation model. The results indicated that the three error correction algorithms exhibited little difference in detection performance when faced with distinct corpus databases. Furthermore, when dealing with the same corpus database, the bidirectional LSTM algorithm demonstrated the most robust detection performance, followed by the one based on the traditional unidirectional LSTM translation model, and lastly, the one based on the LSTM classification model performed the least effectively.

Forecasting relative returns for S&P 500 stocks using machine learning

Forecasting changes in stock prices is extremely challenging given that numerous factors cause these prices to fluctuate. The random walk hypothesis and efficient market hypothesis essentially state that it is not possible to systematically, reliably predict future stock prices or forecast changes in the stock market overall. Nonetheless, machine learning (ML) techniques that use historical data have been applied to make such predictions. Previous studies focused on a small number of stocks and claimed success with limited statistical confidence. In this study, we construct feature vectors composed of multiple previous relative returns and apply the random forest (RF), support vector machine (SVM), and long short-term memory (LSTM) ML methods as classifiers to predict whether a stock can return 2% more than its index in the following 10 days. We apply this approach to all S&P 500 companies for the period 2017–2022. We assess performance using accuracy, precision, and recall and compare our results with a random choice strategy. We observe that the LSTM classifier outperforms RF and SVM, and the data-driven ML methods outperform the random choice classifier (p = 8.46e−17 for accuracy of LSTM). Thus, we demonstrate that the probability that the random walk and efficient market hypotheses hold in the considered context is negligibly small.

Enhancing Smart City Safety and Utilizing AI Expert Systems for Violence Detection

Violent attacks have been one of the hot issues in recent years. In the presence of closed-circuit televisions (CCTVs) in smart cities, there is an emerging challenge in apprehending criminals, leading to a need for innovative solutions. In this paper, the propose a model aimed at enhancing real-time emergency response capabilities and swiftly identifying criminals. This initiative aims to foster a safer environment and better manage criminal activity within smart cities. The proposed architecture combines an image-to-image stable diffusion model with violence detection and pose estimation approaches. The diffusion model generates synthetic data while the object detection approach uses YOLO v7 to identify violent objects like baseball bats, knives, and pistols, complemented by MediaPipe for action detection. Further, a long short-term memory (LSTM) network classifies the action attacks involving violent objects. Subsequently, an ensemble consisting of an edge device and the entire proposed model is deployed onto the edge device for real-time data testing using a dash camera. Thus, this study can handle violent attacks and send alerts in emergencies. As a result, our proposed YOLO model achieves a mean average precision (MAP) of 89.5% for violent attack detection, and the LSTM classifier model achieves an accuracy of 88.33% for violent action classification. The results highlight the model’s enhanced capability to accurately detect violent objects, particularly in effectively identifying violence through the implemented artificial intelligence system.

InfusedHeart: A Novel Knowledge-Infused Learning Framework for Diagnosis of Cardiovascular Events

In the undertaken study, we have used a customized dataset termed ``Cardiac-200'' and the benchmark dataset ``PhysioNet.'' which contains 1500 heartbeat acoustic event samples (without augmentation) and 1950 samples (with augmentation) heartbeat acoustic events such as normal, murmur, extrasystole, artifact, and other unlabeled heartbeat acoustic events. The primary reason for designing a customized dataset, ``cardiac-200,'' is to balance the total number of samples into categories such as normal and abnormal heartbeat acoustic events. The average duration of the recorded heartbeat acoustic events is 10-12 s. In the undertaken study, we have analyzed and evaluated various heartbeat acoustic events using audio processing libraries such as Chromagram, Chroma-cq, Chroma-short-time Fourier transform (STFT), Chroma-cqt, and Chroma-cens to extract more information from the recorded heartbeat sound signals. The noise removal process has been carried out using local binary pattern (LBP) methodology. The noise-robust heartbeat acoustic images are classified using long short-term memory (LSTM)-convolutional neural network (CNN), recurrent neural network (RNN), LSTM, Bi-LSTM, CNN, K-means Clustering, and support vector machine (SVM) methods. The obtained results have shown that the proposed InfusedHeart Framework had outclassed all the other customized machine learning and deep learning approaches such as RNN, LSTM, Bi-LSTM, CNN, K-means Clustering, and SVM-based classification methodologies. The proposed Knowledge-infused Learning Framework has achieved an accuracy of 89.36% (without augmentation), 93.38% (with augmentation), and a standard deviation of 10.64 (without augmentation), and 6.62 (with augmentation). Furthermore, the proposed framework has been tested for various signal-to-noise ratio conditions such as SignaltoNoiseRatio0, SignaltoNoiseRatio3, SignaltoNoiseRatio6, SignaltoNoiseRatio9, SignaltoNoiseRatio12, SignaltoNoiseRatio15, and SignaltoNoiseRatio18. In the end, we have shown a detailed comparison of texture and without texture approaches and have discussed future enhancements and prospective ways for future directions.

EMG and IMU Data Fusion for Locomotion Mode Classification in Transtibial Amputees

Despite recent advancements in prosthetic technology, lower-limb amputees often remain limited to passive prostheses, which leads to an asymmetric gait and increased energy expenditure. Developing active prostheses with effective control systems is important to improve mobility for these individuals. This study presents a machine-learning-based approach to classify five distinct locomotion tasks: ground-level walking (GWL), ramp ascent (RPA), ramp descent (RPD), stairs ascent (SSA), and stairs descent (SSD). The dataset comprises fused electromyographic (EMG) and inertial measurement unit (IMU) signals from twenty non-amputated and five transtibial amputated participants. EMG sensors were strategically positioned on the thigh muscles, while IMU sensors were placed on various leg segments. The performance of two classification algorithms, support vector machine (SVM) and long short-term memory (LSTM), were evaluated on segmented data. The results indicate that SVM models outperform LSTM models in accuracy, precision, and F1 score in the individual evaluation of amputee and non-amputee datasets for 80–20 and 50–50 data distributions. In the 80–20 distribution, an accuracy of 95.46% and 95.35% was obtained with SVM for non-amputees and amputees, respectively. An accuracy of 93.33% and 93.30% was obtained for non-amputees and amputees by using LSTM, respectively. LSTM models show more robustness and inter-population generalizability than SVM models when applying domain-adaptation techniques. Furthermore, the average classification latency for SVM and LSTM models was 19.84 ms and 37.07 ms, respectively, within acceptable limits for real-time applications. This study contributes to the field by comprehensively comparing SVM and LSTM classifiers for locomotion tasks, laying the foundation for the future development of real-time control systems for active transtibial prostheses.

DETECTION MODEL FOR FAKE NEWS ON COVID-19 IN INDONESIA

Today, fake information has become a significant problem, exacerbated by the acceleration of access to information. The spread of fake information has a dangerous impact, especially regarding global health issues, for example COVID-19. People can access various resources to obtain information, including online sites and social media. One of the methods to control the spread of false information is detecting hoaxes. Many methods have been developed to identify hoaxes; most previous studies have focused on developing hoax detection methods using data from a single source in English. The present study is carried out to detect fake news in Indonesian language using multiple data sources, including traditional and social media in the context of COVID-19. The study uses Long Short-Term Memory (LSTM) and the Robustly Optimised Bidirectional Encoder Representations from Transformers Pre-Training Approach (RoBERTa). The LSTM approach is used to develop four different architectures that varied based on: (1) the use of text-only versus the use of both title and text; (2) the number of LSTM and dense layers; and (3) the activation function. The LSTM model with text-only data, a single LSTM layer and two dense layers, outperformed other LSTM architectures, achieving the highest accuracy of 92.17%. The LSTM models require a considerably short training time of 23–27 minutes for 3,847 articles and has a detection time of 3.8–4.1 ms per article. The RoBERTa classifiers outperformed all LSTM models with an accuracy of over 97% and a significantly better training time, with a margin of more than 50% compared to LSTM classifiers, although it had a slightly longer test time. Both LSTM and RoBERTa models outperformed the Naïve Bayes and SVM benchmark methods in terms of accuracy, precision, and recall. Therefore, this study shows that both LSTM and RoBERTa methods are reliable and can be reasonably implemented for real-time fake news detection.

A customized deep learning-based framework for classification and analysis of social media posts to enhance the Hajj and Umrah services

Each year, millions of pilgrims come to Saudi Arabia to perform Hajj and Umrah. To help pilgrims with their difficulties and make plans to enhance the Hajj and Umrah services, this article offers a deep learning-based framework to categorize and analyze pilgrims’ posts on social media networks X (formerly Twitter), Facebook and Instagram. We extracted Arabic posts related to Hajj/Umrah then tokenized and pre-processed the dataset to remove unnecessary parts such as punctuation. Posts are manually labeled into five classes: Health, Organization, Security, Services, and Worship. Labeled posts are collected to build a Long Short-Term Memory (LSTM) network classifier for deep learning. To increase the prediction accuracy, we customized the LSTM classification layer and used the sum of squares error (SSE) loss function instead of the default cross-entropy loss function. A rigorous simulation is run to assess the system’s effectiveness. Every time a simulation round occurs, the trained network is tasked with identifying the classes of unlabeled posts (testing data). Similar tests are performed using the K-Nearest Neighbors (KNN) and Linear Discriminant Analysis (LDA) algorithms for comparative analysis. Accuracy, confusion matrix, precision, sensitivity, specificity, and F1 score are used to quantify the system performance. The maximum mean accuracy of the proposed framework, KNN and LDA are 85.65%, 68.47%, and 59.06% respectively. The proposed framework needs to be integrated into a larger system called the Intelligent Pilgrim Service System (IPSS). We described the IPSS architecture, which comprises modules for data generation, data storage (on the cloud), and a control center. We described a hashtag mechanism as a part of the IPSS. This hashtag mechanism will provide both general and focused advice to millions of pilgrims related to their social media posts. We discussed scenarios of how the IPSS can enhance the Hajj and Umrah services.

Arrhythmia classification using ECG signal: A meta-heuristic improvement of optimal weighted feature integration and attention-based hybrid deep learning model

In healthcare facilities, the most common and least expensive diagnostic tool for monitoring electrical signals in the heart is the Electrocardiogram (ECG). Arrhythmia is nothing but abnormal rhythms in the heart. Arrhythmias can be dangerous and fatal. An ECG test can identify the type of arrhythmia that's causing the abnormal heartbeat. These necessitate lengthy, high-quality ECG segments, which may result in a few important episode omissions. Because of this, a low-cost, dependable, and automatic system for screening and treating arrhythmia becomes highly necessary. A novel method for the classification of arrhythmias is suggested on the basis of attention-based deep learning to satisfy the above-mentioned requirement. Signal collection, signal analysis, feature extraction, weighted fused features, and classification are the various phases of the proposed work. The input signal is initially obtained from benchmark data sources. The RR interval and QRS complex signal analysis are carried out next. A One-Dimensional Convolution Neural Network (1DCNN) is used to extract the deep features from the resulting RR interval and QRS complex, as well as the Higher-Order Statistical (HOS) features from the raw signal. The weighted fused feature selection is carried out after the above-mentioned three features have been obtained, where the weight is adjusted with the Recommended Adaptive Risk Rate-based Lemurs Optimization Algorithm (ARR-LO). Finally, the fused features are fed into the hybrid Attention model through the Long Short-Term Memory (LSTM) and Deep Temporal Convolution Network (DTCN), where ARR-LO is used to optimally determine some of the hyperparameters. The objectives of the study include developing an arrhythmia classification framework, constructing an adaptive attention-based hybrid deep learning model (HA-LSTM-DTCN), creating a new meta-heuristic algorithm (ARR-LO), and validating the suggested model against existing classifiers. The performance is compared with the conventional models and evaluated using a variety of metrics and the results from the simulation, the generated deep learning-based arrhythmia classification framework outperformed conventional prediction models such as DTCN, LSTM-DTCN, DNN, and LSTM classifiers in classifying the arrhythmia by 81.79%, 55.46%, and 80.73%, respectively and the overall accuracy of the proposed framework performs 94.425%. From the results, it was shown that generated AR-LO-HA-LSTM-DTCN model for arrhythmia classification improved the classification performance, thus making it useful in the quick diagnosis of arrhythmia disease. The findings show that the suggested arrhythmia classification model improves the classification performance, which can be used to quickly diagnose arrhythmia disease.

Computer aided progression detection model based on optimized deep LSTM ensemble model and the fusion of multivariate time series data

Alzheimer’s disease (AD) is the most common form of dementia. Early and accurate detection of AD is crucial to plan for disease modifying therapies that could prevent or delay the conversion to sever stages of the disease. As a chronic disease, patient’s multivariate time series data including neuroimaging, genetics, cognitive scores, and neuropsychological battery provides a complete profile about patient’s status. This data has been used to build machine learning and deep learning (DL) models for the early detection of the disease. However, these models still have limited performance and are not stable enough to be trusted in real medical settings. Literature shows that DL models outperform classical machine learning models, but ensemble learning has proven to achieve better results than standalone models. This study proposes a novel deep stacking framework which combines multiple DL models to accurately predict AD at an early stage. The study uses long short-term memory (LSTM) models as base models over patient’s multivariate time series data to learn the deep longitudinal features. Each base LSTM classifier has been optimized using the Bayesian optimizer using different feature sets. As a result, the final optimized ensembled model employed heterogeneous base models that are trained on heterogeneous data. The performance of the resulting ensemble model has been explored using a cohort of 685 patients from the University of Washington's National Alzheimer’s Coordinating Center dataset. Compared to the classical machine learning models and base LSTM classifiers, the proposed ensemble model achieves the highest testing results (i.e., 82.02, 82.25, 82.02, and 82.12 for accuracy, precision, recall, and F1-score, respectively). The resulting model enhances the performance of the state-of-the-art literature, and it could be used to build an accurate clinical decision support tool that can assist domain experts for AD progression detection.

Text classification in fair competition law violations using deep learning

IntroductionEnsuring fair competition through manual review is a complex undertaking. This paper introduces the utilization of Long Short-Term Memory (LSTM) neural networks and TextCNN to establish a text classifier for classifying and reviewing normative documents.MethodsThe experimental dataset used consists of policy measure samples provided by the antitrust division of the Guangdong Market Supervision Administration. We conduct a comparative analysis of the performance of LSTM and TextCNN classification models.ResultsIn three classification experiments conducted without an enhanced experimental dataset, the LSTM classifier achieved an accuracy of 95.74%, while the TextCNN classifier achieved an accuracy of 92.7% on the test set. Conversely, in three classification experiments utilizing an enhanced experimental dataset, the LSTM classifier demonstrated an accuracy of 96.36%, and the TextCNN classifier achieved an accuracy of 96.19% on the test set.DiscussionThe experimental results highlight the effectiveness of LSTM and TextCNN in classifying and reviewing normative documents. The superior accuracy achieved with the enhanced experimental dataset underscores the potential of these models in real-world applications, particularly in tasks involving fair competition review.

Design of recognition algorithm for multiclass digital display instrument based on convolution neural network

Digital display instrument identification is a crucial approach for automating the collection of digital display data. In this study, we propose a digital display area detection CTPNpro algorithm to address the problem of recognizing multiclass digital display instruments. We developed a multiclass digital display instrument recognition algorithm by combining the character recognition network constructed using a convolutional neural network and bidirectional variable-length long short-term memory (LSTM). First, the digital display region detection CTPNpro network framework was designed based on the CTPN network architecture by introducing feature fusion and residual structure. Next, the digital display instrument identification network was constructed based on a convolutional neural network using two-way LSTM and Connectionist temporal classification (CTC) of indefinite length. Finally, an automatic calibration system for digital display instruments was built, and a multiclass digital display instrument dataset was constructed by sampling in the system. We compared the performance of the CTPNpro algorithm with other methods using this dataset to validate the effectiveness and robustness of the proposed algorithm.

Accurate detection of vehicle, pedestrian, cyclist and wheelchair from roadside light detection and ranging sensors

Accurate detection plays a critical role in improving the safety situation of vulnerable road users. This study extends infrastructure-based LiDAR application to all three major vulnerable road user groups including pedestrians, cyclists, and wheelchair users. Two critical problems for accurate detection of small-sized road users are scanning angle variability and feature fluctuation. To address these issues, a feature-based classification method combined with prior LiDAR trajectory information is developed. Effective dimension-related features are proposed and five classifiers including artificial neural network (ANN), random forest (RF), adaptive boosting (AdaBoost), random under-sampling boosting (RUSBoost), and long short-term memory (LSTM) are tested with a novel feature engineering process. A total of seven features are selected from the point cloud of clusters for vehicle/pedestrian/cyclist/wheelchair classification. By updating these significant features based on prior information of the entire trajectory, the performance of road user classification (imbalanced datasets) has been significantly improved. Experimental study is conducted to examine the recall rate, F1-score, and AUC of vehicles, pedestrians, cyclists, and wheelchairs before and after integration with prior trajectory information. The result shows the trained AdaBoost, RUSBoost, and LSTM classifiers with prior trajectory information can achieve recall/F1-score/AUC: (1) Low traffic volumes – vehicles (100%/99.96%/99.96%), pedestrians (99.96%/99.96%/99.97%), cyclists (99.74%/99.45%/99.67%), and wheelchairs (99.22%/99.68%/99.01%) and (2) Moderate traffic volumes – vehicles (99.39%/99.44%/99.69%), pedestrians (98.33%/97.99%/98.64%), and cyclists (95.41%/94.29%/94.40%), using 32-laser LiDAR sensors (10 Hz).

User depression and severity level prediction during COVID-19 epidemic from social network data

In recent times, depression becomes a major issue, which causes suicide, particularly among youngsters. During the Coronavirus disease (COVID-19) epidemic, many organizations suggested social distance and quarantine actions, which cause major attention to the mental health and depression of each individual. Most population express their emotions by using modern social media technologies like Twitter, Facebook, etc. By considering user tweets, a Long Short-Term Memory (LSTM)-based classifier was designed that learns rich attributes such as psychological, contextual, cognitive, person-level and distress-dependent n-gram attributes of each user for depression severity level prediction. But, it did not learn the social network structural properties of the most prominent communities, which influences the prediction outcomes. Hence in this article, a novel model is developed to predict the user’s depression severity levels by considering the social network structure of most prominent communities and influence measures. At first, it analyses the physical characteristics of the most prominent groups based on their balanced local and global power distribution. Then, the influential users and communities are identified along with the rich group of attributes. Moreover, those attributes are provided to the LSTM classifier for the user’s depression severity level prediction during the COVID-19 epidemic. Finally, the investigational outcomes exhibit that the presented model attains 93.53% accuracy and 0.4376 Root Mean Square Error (RMSE) contrasted with the conventional classifiers to estimate the user’s depression severity level.

Development of New Machine Learning Based Algorithm for the Diagnosis of Obstructive Sleep Apnea from ECG Data

In this study, a machine learning algorithm is proposed to be used in the detection of Obstructive Sleep Apnea (OSA) from the analysis of single-channel ECG recordings. Eighteen ECG recordings from the PhysioNet Apnea-ECG dataset were used in the study. In the feature extraction stage, dynamic time warping and median frequency features were obtained from the coefficients obtained from different frequency bands of the ECG data by using the wavelet transform-based algorithm. In the classification phase, OSA patients and normal ECG recordings were classified using Random Forest (RF) and Long Short-Term Memory (LSTM) classifier algorithms. The performance of the classifiers was evaluated as 90% training and 10% testing. According to this evaluation, the accuracy of the RF classifier was 82.43% and the accuracy of the LSTM classifier was 77.60%. Considering the results obtained, it is thought that it may be possible to use the proposed features and classifier algorithms in OSA classification and maybe a different alternative to existing machine learning methods. The proposed method and the feature set used are promising because they can be implemented effectively thanks to low computing overhead.

Enhancing CSI-Based Human Activity Recognition by Edge Detection Techniques

Human Activity Recognition (HAR) has been a popular area of research in the Internet of Things (IoT) and Human–Computer Interaction (HCI) over the past decade. The objective of this field is to detect human activities through numeric or visual representations, and its applications include smart homes and buildings, action prediction, crowd counting, patient rehabilitation, and elderly monitoring. Traditionally, HAR has been performed through vision-based, sensor-based, or radar-based approaches. However, vision-based and sensor-based methods can be intrusive and raise privacy concerns, while radar-based methods require special hardware, making them more expensive. WiFi-based HAR is a cost-effective alternative, where WiFi access points serve as transmitters and users’ smartphones serve as receivers. The HAR in this method is mainly performed using two wireless-channel metrics: Received Signal Strength Indicator (RSSI) and Channel State Information (CSI). CSI provides more stable and comprehensive information about the channel compared to RSSI. In this research, we used a convolutional neural network (CNN) as a classifier and applied edge-detection techniques as a preprocessing phase to improve the quality of activity detection. We used CSI data converted into RGB images and tested our methodology on three available CSI datasets. The results showed that the proposed method achieved better accuracy and faster training times than the simple RGB-represented data. In order to justify the effectiveness of our approach, we repeated the experiment by applying raw CSI data to long short-term memory (LSTM) and Bidirectional LSTM classifiers.

Using a novel convolutional neural network for plant pests detection and disease classification.

Early plant diseases and pests identification reduces social, economic, and environmental deficiencies entailing toxic chemical utilization on agricultural farms, thus posing a threat to global food security. An enhanced convolutional neural network (CNN) along with long short-term memory (LSTM) using a majority voting ensemble classifier has been proposed to tackle plant pest and disease identification and classification. Within pre-trained models, deep feature extractions have been obtained from connected layers. Deep features have been extracted and are sent to the LSTM layer to build a robust, enhanced LSTM-CNN model for detecting plant pests and diseases. Experiments were carried out using a Turkey dataset, with 4447 apple pests and diseases categorized into 15 different classes. The study was evaluated in different CNNs using logistic regression (LR), LSTM, and extreme learning machine (ELM), focusing on plant disease detection problems. The ensemble majority voting classifier was used at the LSTM layer to detect and classify plant disease labels. Furthermore, an autonomous selection of the optimal LSTM layer network parameters was applied. Finally, the performance was validated based on sensitivity, F1 score, accuracy, and specificity using LSTM, ELM, and LR classifiers. The presented model attained 99.2% accuracy compared to the cutting-edge models on different classifiers such as LSTM, LR, and ELM, and performed better compared to transfer learning. Pre-trained models, such as VGG19, VGG18, and AlexNet, demonstrated better accuracy when the fc6 layer was compared with other layers. © 2023 Society of Chemical Industry.

Classification of arrhythmias using an LSTM- and GAN-based approach to ECG signal augmentation

Abstract Funding Acknowledgements Type of funding sources: Private grant(s) and/or Sponsorship. Main funding source(s): British Heart Foundation Introduction Automated classification of arrhythmias in ECGs is becoming increasingly important. Publicly available ECG datasets have been widely used by the research community to create novel artificial intelligence models that improve these detection rates. The development of these models requires access to large volume of labelled data. However, access to such databases is becoming increasingly limited. In addition, the datasets are often unbalanced because abnormal rhythms are far outweighed by normal samples. The unbalanced nature of the datasets can lead to less accurate models. Therefore, generating realistic synthetic signals can augment the real signals found in such databases and provide data that allows sophisticated model development. Purpose In this study, we propose a deep learning-based approach for synthetic ECG signal generation that uses long short-term memory (LSTM) autoencoder and generative adversarial networks (GAN) to generate signals that mimic the distribution of arrhythmia signals (Figure 1). Methods The LSTM autoencoder is composed of two parts: an encoder and a decoder (Figure 1b). The encoder takes original ECG signal as its input and uses LSTM layers to compress the signal into a set of features. The decoder is formed by reversing the encoding process, which uses the encoded features as its input and converts them back into the original signal. To generate synthetic signals, we inserted GANs between the LSTM encoder and the decoder. GANs are composed of a generator and a discriminator (Figure 1c). The generator produces synthetic ECG features based on noise, whereas the discriminator tries to distinguish between real features and results received from the generator. The pathological beats studies were: left bundle branch block (LBBB), right bundle branch block (RBBB), aberrated atrial premature (AA), and normal beats (N) from the MIT-BIH arrhythmia database, using lead II only. To evaluate the quality of our synthetic signals, we trained an LSTM classifier on a combination of our real and synthetic data and compared the testing results with a model trained on real data alone. Results The LSTM encoder, decoder and GAN were trained individually for each beat type, and examples of generated signals are illustrated in Figure 2. The average accuracy of the classification for the original dataset was 90%, with a recall of 98% for N, 36% for AA, 39% for LBBB and 97% for RBBB. Once synthetic signals were added to the training set, the average testing classification accuracy increased to 98%, with a recall of 99% for N, 83% for AA, 99% for LBBB and 99% for RBBB. For fair comparison, the testing set contained only real data and remained unchanged for both groups. Conclusion In this work, we proved the capability of GANs to generate realistic synthetic signals that helped to improve the detection rates of arrhythmias as measured by both increased overall accuracy and recall.

Using CCA-Fused Cepstral Features in a Deep Learning-Based Cry Diagnostic System for Detecting an Ensemble of Pathologies in Newborns

Crying is one of the means of communication for a newborn. Newborn cry signals convey precious information about the newborn's health condition and their emotions. In this study, cry signals of healthy and pathologic newborns were analyzed for the purpose of developing an automatic, non-invasive, and comprehensive Newborn Cry Diagnostic System (NCDS) that identifies pathologic newborns from healthy infants. For this purpose, Mel-frequency Cepstral Coefficients (MFCC) and Gammatone Frequency Cepstral Coefficients (GFCC) were extracted as features. These feature sets were also combined and fused through Canonical Correlation Analysis (CCA), which provides a novel manipulation of the features that have not yet been explored in the literature on NCDS designs, to the best of our knowledge. All the mentioned feature sets were fed to the Support Vector Machine (SVM) and Long Short-term Memory (LSTM). Furthermore, two Hyperparameter optimization methods, Bayesian and grid search, were examined to enhance the system's performance. The performance of our proposed NCDS was evaluated with two different datasets of inspiratory and expiratory cries. The CCA fusion feature set using the LSTM classifier accomplished the best F-score in the study, with 99.86% for the inspiratory cry dataset. The best F-score regarding the expiratory cry dataset, 99.44%, belonged to the GFCC feature set employing the LSTM classifier. These experiments suggest the high potential and value of using the newborn cry signals in the detection of pathologies. The framework proposed in this study can be implemented as an early diagnostic tool for clinical studies and help in the identification of pathologic newborns.

UWB Positioning System Based on LSTM Classification With Mitigated NLOS Effects

It is known that an ultrawideband (UWB)-based indoor positioning system (IPS) has superior positioning performance and can meet the requirements of location-based services (LBSs) as the Internet of Things (IoT) applications. However, there is a limitation of UWB positioning when it is conducted at the nonline-of-sight (NLOS) channels degrading the UWB ranging accuracy at indoor environments. In this article, we propose an artificial intelligence (AI) applied UWB positioning system that can enhance the positioning performance by classifying channel conditions with channel impulse response (CIR) of the received UWB signal. The proposed system mitigates the positioning degradation caused by the NLOS situations by performing extended Kalman filter (EKF) localization and long short-term memory (LSTM) training of the observed channel status. The main feature of the proposed UWB positioning method is that it can be used even at unknown locations not trained with the LSTM model learning the channel status, because of our training strategy of not the position coordinates, but the UWB ranging error between UWB devices corresponding to CIR of the received UWB signal. This article provides the experimental setup and performance evaluation results of the proposed system. The evaluation results showed that the proposed AI-applied UWB positioning method significantly improved its accuracy performance compared with the existing positioning methods.