A Spectral Feature Based CNN Long Short-Term Memory Approach for Classification

Robots are becoming everyday devices, increasing their interaction with humans. To make human-machine interaction more natural, cognitive features like Visual Voice Activity Detection (VVAD), which can detect whether a person is speaking or not, given visual input of a camera, need to be implemented. Neural networks are state of the art for tasks in Image Processing, Time Series Prediction, Natural Language Processing and other domains. Those Networks require large quantities of labeled data. Currently there are not many datasets for the task of VVAD. In this work we created a large scale dataset called the VVAD-LRS3 dataset, derived by automatic annotations from the LRS3 dataset. The VVAD-LRS3 dataset contains over 44K samples, over three times the next competitive dataset (WildVVAD). We evaluate different baselines on four kinds of features: facial and lip images, and facial and lip landmark features. With a Convolutional Neural Network Long Short Term Memory (CNN LSTM) on facial images an accuracy of 92% was reached on the test set. A study with humans showed that they reach an accuracy of 87.93% on the test set.

The study presents the results of the development and testing of deep learning neural network architectures, which demonstrate high accuracy rates in classifying neurophysiological data, in particular hemodynamic brain activation patterns obtained by functional near-infrared spectroscopy, during solving mathematical problems on spatial-numerical associations. The analyzed signal represents a multidimensional time series of oxyhemoglobin and deoxyhemoglobin dynamics. Taking the specificity of the fNIRS signal into account, a comparative analysis of 2 types of neural network architectures was carried out: (1) architectures based on recurrent neural networks: recurrent neural network with long short-term memory, recurrent neural network with long short-term memory with fully connected layers, bidirectional recurrent neural network with long short-term memory, convolutional recurrent neural network with long short-term memory; (2) architectures based on convolutional neural networks with 1D convolutions: convolutional neural network, fully convolutional neural network, residual neural network. Trained long short-term memory recurrent neural network architectures showed worse results in accuracy in comparison with 1D convolutional neural network architectures. Residual neural network (model_Resnet) demonstrated the highest accuracy rates in three experimental conditions more than 88% in detecting age-related differences in brain activation during spatial-numerical association tasks considering the individual characteristics of the respondents’ signal.

This paper implements a deep learning‐based modulation pattern recognition algorithm for communication signals using a convolutional neural network architecture as a modulation recognizer. In this paper, a multiple‐parallel complex convolutional neural network architecture is proposed to meet the demand of complex baseband processing of all‐digital communication signals. The architecture learns the structured features of the real and imaginary parts of the baseband signal through parallel branches and fuses them at the output according to certain rules to obtain the final output, which realizes the fitting process to the complex numerical mapping. By comparing and analyzing several commonly used time‐frequency analysis methods, a time‐frequency analysis method that can well highlight the differences between different signal modulation patterns is selected to convert the time‐frequency map into a digital image that can be processed by a deep network. In order to fully extract the spatial and temporal characteristics of the signal, the CLP algorithm of the CNN network and LSTM network in parallel is proposed. The CNN network and LSTM network are used to extract the spatial features and temporal features of the signal, respectively, and the fusion of the two features as well as the classification is performed. Finally, the optimal model and parameters are obtained through the design of the modulation recognizer based on the convolutional neural network and the performance analysis of the convolutional neural network model. The simulation experimental results show that the improved convolutional neural network can produce certain performance gains in radio signal modulation style recognition. This promotes the application of machine learning algorithms in the field of radio signal modulation pattern recognition.

This paper applies deep learning models to predict Bitcoin price directions and the subsequent profitability of trading strategies based on these predictions. The study compares the performance of the convolutional neural network–long short-term memory (CNN–LSTM), long- and short-term time-series network, temporal convolutional network, and ARIMA (benchmark) models for predicting Bitcoin prices using on-chain data. Feature-selection methods—i.e., Boruta, genetic algorithm, and light gradient boosting machine—are applied to address the curse of dimensionality that could result from a large feature set. Results indicate that combining Boruta feature selection with the CNN–LSTM model consistently outperforms other combinations, achieving an accuracy of 82.44%. Three trading strategies and three investment positions are examined through backtesting. The long-and-short buy-and-sell investment approach generated an extraordinary annual return of 6654% when informed by higher-accuracy price-direction predictions. This study provides evidence of the potential profitability of predictive models in Bitcoin trading.

Urdu language uses cursive script which results in connected characters constituting ligatures. For identifying characters within ligatures of different scales (font sizes), Convolution Neural Network (CNN) and Long Short Term Memory (LSTM) Network are used. Both network models are trained on formerly extracted ligature thickness graphs, from which models extract Meta features. These thickness graphs provide consistent information across different font sizes. LSTM and CNN are also trained on raw images to compare performance on both forms of inputs. For this research, two corpora, i.e. Urdu Printed Text Images (UPTI) and Centre for Language Engineering (CLE) Text Images are used. Overall performance of networks ranges between 90% and 99.8%. Average accuracy on Meta features is 98.08% while using raw images, 97.07% average accuracy is achieved.

Identification of brain tumors and accurate grading at an early stage are crucial in cancer diagnosis, as a timely diagnosis can increase the chances of survival. Considering the challenges and risks of tumor biopsies, noninvasive imaging procedures such as Magnetic Resonance Imaging (MRI) are extensively used in analyzing brain tumors. Recent advances in the field of medical imaging with deep learning using three dimensional (3D) MRI is aiding the clinical experts significantly in the diagnosis of brain tumor. In this study, three BraTS MRI datasets named BraTS 2018, BraTS 2019 and BraTS 2020 are employed to classify brain tumor into high-grade glioma (HGG) and low-grade glioma (LGG) where each of the datasets contains four different sequences of 3D MRI brain images named T1-weighted MRI (T1), T1-weighted MRI with contrast enhancement (T1ce), T2-weighted MRI (T2), and Fluid Attenuated Inversion Recovery (FLAIR) for a single patient. This research is composed of two approaches where, in the first part, we propose a hybrid deep learning model named TimeDistributed-CNN-LSTM (TD-CNN-LSTM) combining 3D Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) where each layer of the architecture is wrapped with a TimeDistributed function. The objective of developing the hybrid model is to consider all the four 3D MRI sequences of each patient as a single input data because every sequence contains necessary information on tumor that can have an impact on improving cancer detection performance. However, interpreting all the MRI sequences together with optimal performance especially in 3D is quite challenging. Therefore, the model is developed with optimal configuration based on highest accuracy performing ablation study for layer architecture and hyper-parameters. In the second part, a 3D CNN model is trained respectively with each of the MRI sequences to compare the performance with TD-CNN-LSTM model. In this regard, for both of the models, BraTS 2018 and BraTS 2019 are combined to increase the number of images and used as train dataset where BraTS 2020 dataset is employed as the test dataset. Moreover, before training the models the datasets is preprocessed to ensure the highest performance. Our results of these two approaches demonstrate that the TD-CNN-LSTM network outperforms 3D CNN achieving the highest test accuracy of 98.90%. Later, to evaluate the performance consistency, the TD-CNN-LSTM model is evaluated with K-fold cross validation using different k values. The approach of putting together all the MRI sequences at a time using hybrid CNN-LSTM approach with good generalization capability to classify brain tumor can be used effectively in future Computer Aided Diagnosis (CAD) based research which can aid radiologists in medical diagnostics.

Purpose Low-velocity impacts may cause barely visible internal damage in carbon fiber-reinforced polymer (CFRP) laminates, which can significantly reduce the load-carrying capacity of Type IV hydrogen storage cylinders. This study aims to propose an acoustic-emission (AE) sensing framework to quantitatively evaluate low-velocity impact damage in CFRP laminates relevant to Type IV cylinder structures, and to establish severity grading criteria based on an interpretable scalar indicator. Design/methodology/approach Drop-weight impact tests were conducted from 10–80 J while AE signals were continuously recorded. AE features were extracted using empirical mode decomposition (EMD) and principal component analysis (PCA). A hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) classifier was trained to identify damage-related AE signals, and wavelet entropy was calculated to quantify impact damage severity. Nonlinear ultrasonic testing was additionally used to validate damage progression. Findings The CNN–LSTM model achieved 93.3% classification accuracy for AE signals associated with matrix cracking, delamination and fiber fracture. Wavelet entropy showed strong correlation with impact damage severity and identified three damage stages. Two critical thresholds were found: wavelet entropy 0.275 (transition to moderate damage) and 0.55 (transition to severe damage), corresponding to impact energies of 40 and 60 J, respectively. Originality/value This study links AE signal characteristics to impact damage severity through an interpretable scalar metric (wavelet entropy) and provides quantitative thresholds (0.275 and 0.55) for three-stage damage grading. It further combines EMD–PCA feature extraction with a CNN–LSTM classifier (93.3% accuracy) to identify damage-related AE signals, and validates damage progression using nonlinear ultrasonic testing. The framework offers a practical sensing strategy for composite pressure-vessel monitoring.

Considerable attention has recently been focused on classification and regression-based convolutional neural network (CNN) and long short-term memory (LSTM) due to their excellent performance in capturing complex spatial and temporal information characteristics for structural damage identification. However, few studies have considered structural damage identification as a classification and regression problem. In addition, bridges in practical engineering are vulnerable to various environmental and vehicle loading conditions. Hence, this study proposed a new two-stage CNN–LSTM configuration for bridge damage identification using vibration data considering the influence of temperatures. First, a classification-based CNN–LSTM is designed to perform multiclass damage detection tasks, and then a regression-based CNN–LSTM is developed for damage localization and severity prediction tasks. The performance of the proposed damage identification method was evaluated through a simulation dataset of a concrete highway bridge model and a field experiment dataset of Z24-bridge (Switzerland). In addition, a set of statistical evaluation metrics such as sparse categorical cross-entropy loss, accuracy, confusion matrix, mean squared loss, mean absolute error, mean absolute percentage error, and coefficient of determination were used to compare the damage identification performance of the proposed CNN–LSTM configuration with a regular CNN model and conventional machine learning (ML) algorithms. Prediction results indicate that the proposed CNN–LSTM model outperforms the regular CNN model and conventional ML algorithms for bridge damage identification.

Low-rate Distributed Denial of Service (LDDoS) attacks are complex and common security issues, which disrupt the essential services of the varied and emerging networks. However, the traditional detection models are normally in a dilemma condition since LDDoS attacks have the characteristics of concealment and persistence in space and time dimensions. In this paper, we propose a novel, practical and fast ensemble model (FastCBLA-EM) to detect the LDDoS attacks. In the FastCBLA-EM, a dilated padding-based dual sliding method is worked out for producing fixed-length chronological order samples from flows per well-tuned packet step. Next, a competitive network composed of 1-D Convolutional Neural Network (1-D CNN) and a bidirectional Long Short Term Memory (bi-LSTM) network is employed to learn both spatial and temporal representations of samples, parallelly. Finally, an arbitration mechanism based on the enhanced attention method is used to assemble and weigh these representations for detecting LDDoS attacks. We compare the performances of the FastCBLA-EM with other models by obtaining a group of well-tuned hyperparameters in the revised ISCX-2016-SlowDos flow-based dataset. Experimental results demonstrate that the detection accuracy of the FastCBLA-EM can be up to 99.7% and the time complexity is <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$O(n)$</tex-math></inline-formula> .

In recent years, channel state information (CSI) in WiFi 802.11n has been increasingly used to collect data pertaining to human activity. Such raw data are then used to enhance human activity recognition. Activities such as lying down, falling, walking, running, sitting down, and standing up can now be detected with the use of information collected through CSI. Human activity recognition has a multitude of applications, such as home monitoring of patients. Four deep learning models are presented in this paper, namely: a convolution neural network (CNN) with a Gated Recurrent Unit (GRU); a CNN with a GRU and attention; a CNN with a GRU and a second CNN, and a CNN with Long Short-Term Memory (LSTM) and a second CNN. Those models were trained to perform Human Activity Recognition (HAR) using CSI amplitude data collected by a CSI tool. Experiments conducted to test the efficacy of these models showed superior results compared with other recent approaches. This enhanced performance of our models may be attributable the ability of our models to make full use of available data and to extract all data features, including high dimensionality and time sequence. The highest average recognition accuracy reached by the proposed models was achieved by the CNN-GRU, and the CNN-GRU with attention models, standing at 99.31% and 99.16%, respectively. In addition, the performance of the models was evaluated for unseen CSI data by training our models using a random split-of-dataset method (70% training and 30% testing). Our models achieved impressive results with accuracies reaching 100% for nearly all activities. For the lying down activity, accuracy obtained from the CNN-GRU model stood at 99.46%; slightly higher than the 99.05% achieved by the CNN-GRU with attention model. This confirmed the robustness of our models against environmental changes.

<title>Abstract</title> In this study, a hybrid deep learning model is presented that integrates Bidirectional Long Short-Term Memory (BiLSTM) with the Informer architecture enhanced by ProbSparse attention to improve photovoltaic (PV) power forecasting across multiple time horizons. This integrated design aims to enhance temporal feature representation and reduce long-horizon error accumulation, providing a more reliable and efficient approach to PV power prediction. Existing hybrid models, including Convolutional Neural Network with Long Short-Term Memory (CNN–LSTM) and Gated Recurrent Unit with Long Short-Term Memory (GRU–LSTM), achieved coefficients of determination (R²) of 0.771 and 0.892, respectively; however, their predictive accuracy deteriorated over extended forecasting windows due to cumulative errors and limited sensitivity to long-range temporal dependencies. The proposed BiLSTM–Informer model addresses these limitations through advanced feature engineering, incorporating Fourier transformation, cyclic time encoding, and autoregressive lag selection to improve the representation of periodic and environmental variations. In addition, a comprehensive sensitivity analysis was conducted to examine the effects of varying input sequence lengths (look-back windows) and forecast horizons on model accuracy and temporal learning behaviour. The results show that the BiLSTM–Informer consistently maintained superior predictive stability across extended horizons, achieving up to a 12% reduction in mean absolute error compared with benchmark hybrid models, thereby demonstrating enhanced robustness and generalisation for both short- and long-term forecasting. When evaluated on a five-year hourly PV dataset, the model achieved an R² of 0.952, a mean absolute error of 1.22 kWh, and a root mean square error of 2.21 kWh, outperforming all benchmark approaches. To assess practical applicability, the trained model was deployed via a Streamlit-based web application on the Orender platform, where real-time inference achieved forecasting accuracies ranging from 89% to 97.3% across hourly to monthly resolutions. The dataset and pre-trained model are publicly available at https://github.com/kamilkenny/EDA, and live inference can be accessed at https://kamil-deployment-of-edgehill-durning.onrender.com/.

Online leakage current classification using convolutional neural network long short-term memory for high voltage insulators on web-based service

Diabetes is a persistent health condition led by insufficient use or inappropriate use of insulin in the body. If left undetected, it can lead to further complications involving organ damage such as heart, lungs, and eyes. Timely detection of diabetes helps obtain the right medication, diet, and exercise plan to lead a healthy life. ML approach has been utilized to obtain rapid and reliable diabetes detection, however, existing approaches suffer from the use of limited datasets, lack of generalizability, and lower accuracy. This study proposes a novel feature extraction approach to overcome these limitations by using an ensemble of convolutional neural network (CNN) and long short-term memory (LSTM) models. Multiple datasets are combined to make a larger dataset for experiments and multiple features are utilized for investigating the efficacy of the proposed approach. Features from the extra tree classifier, CNN, and LSTM are also considered for comparison. Experimental results reveal the superb performance of CNN-LSTM-based features with random forest model obtaining a 0.99 accuracy score. This performance is further validated by comparison with existing approaches and k-fold cross-validation which shows the proposed approach provides robust results.

The rapid transmission and mutation rates of viruses necessitate efficient and precise classification methods to aid public health responses. Often constrained by culturing requirements and narrow taxonomic scopes, traditional diagnostic techniques struggle to keep pace with the rapidly expanding genomic landscape. This study introduces an advanced multi-virus genome sequence classification approach utilizing a k-mer-based dictionary for rapid genomic pattern extraction, enhanced by sophisticated natural language processing (NLP) and machine learning (ML) models. Our contributions are threefold: first, we introduce a k-mer dictionary approach that significantly enhances virus classification speed and accuracy by capturing essential genomic patterns; second, we implement a suite of optimized classifiers—Multinomial Naive Bayes, Random Forest, K-Nearest Neighbor, and nu-SVM—that achieve high performance metrics, with accuracy at 0.95, precision at 0.96, specificity at 0.94, and sensitivity at 0.99; third, we provide a comprehensive comparative analysis with advanced deep learning architectures, such as CNN, CNN-LSTM, and CNN-Bidirectional LSTM, wherein our method demonstrates superior performance. These findings validate the effective integration of ML and NLP techniques in viral genome classification, setting new benchmarks in bioinformatics and offering scalable solutions for rapid virus detection and identification.

One tool for promoting mental health is human stress detection through multitasks of electroencephalography (EEG) recordings. This study proposed a short-term stress detection approach using VGGish as a feature extraction and convolution neural network (CNN) as a classifier based on EEG signals from the SAM 40 dataset. This database was recently available and was collected from 40 patients using 32 channels to identify performance on four tasks including Stroop color-word test (SCWT), answering arithmetic problems, finding mirror-identical images, and relaxing. Each task took 25 s to complete and was then repeated three times to record three trials. This means that the total EEG data contain 480 signals for four tasks recorded using 120 trials per task. The primary objective of this research was to track the amount of short-term stress that patients experienced while they engaged in the four mental tasks. Moreover, the VGGish-CNN model is applied to the SAM 40 dataset using five stages including signal preprocessing, segmentation, filtration, spectrogram, and classification process. We compared the VGGish-CNN model and the VGGish model for stress-based EEG classification to determine the best classification accuracy. The proposed approach for stress detection is the preliminary study that achieved an accuracy of 99.25% using the VGGish-CNN model on the SAM 40 dataset. Next, k-fold cross validation is performed to verify the efficiency of the VGGish-CNN model. This study can advance the application of brain–computer interface (BCI) and its use to identify patterns in EEG data that invoke stress-related inferences to aid in the diagnosis of mental disorders. In the future, investigation of human stress using EEG data will be useful in neurorehabilitation.