Browse Articles

Stacked convolutional neural network for emotion recognition using multi feature speech analysis

Abstract Remote diagnosis is increasingly incorporating emotion recognition, enabling clinicians to assess patients’ emotional states during teleconsultations through analysis of vocal and acoustic characteristics. This study proposes a refined deep learning framework for emotion recognition from speech signals, designed to enhance the reliability of remote medical assessments. Several deep learning architectures, including convolutional neural networks (CNN), recurrent neural networks (RNN), and long short-term memory (LSTM) models, were evaluated using three publicly available emotional speech datasets: RAVDESS, TESS, and SAVEE. The primary contribution of this work is the Stacked Convolutional Network (SCoNN), a deep neural architecture developed to hierarchically extract and integrate complex audio features for improved emotion classification. The model comprises multiple Conv1D blocks incorporating batch normalization, dropout, and activation layers, followed by a dense softmax output layer for final classification. SCoNN achieved accuracies of 99.93% on the TESS dataset using combined MFCC and Mel Spectrogram features; 91.51%, 90.63%, and 93.30% on the RAVDESS dataset for Mel Spectrogram, MFCC, and combined features, respectively; and 91.43%, 94.76%, and 95.00% on the SAVEE dataset for the same feature configurations. The novelty of SCoNN lies in its hierarchical stacking mechanism and adaptive multi-feature fusion, enabling superior capture of emotional variations in speech compared to conventional deep CNNs. The proposed framework demonstrates high efficiency and reliability for emotion recognition in remote healthcare applications.

An expansive global oxygenation of Earth’s surface environments 1.4 billion years ago

MethPy: a new software for analyzing non-CpG methylation after bisulfite assay and Sanger sequencing

Transformer-based deep learning for adaptive pedagogy under uncertain student preferences

Multiscale aperture synthesis imager

Coupled coordination study of inter-provincial carbon emission efficiency and ecological infrastructure in China

Discovery of anti-inflammatory compounds from the stem bark of Garcinia latissima through in vitro and in silico approaches

Flow Photo-Oxidative Dearomative Cyclization Enables Unified Asymmetric Synthesis of Trichodermamides A–F

Subgroup of meningiomas involving FOS and FOSB gene fusions

Comparison of the analytical model with neural network model on the case of heat exchanger behavior under fouling

Synthesis, structural elucidation, and molecular docking of diclofenac-derived hydrazone metal complexes with anti-inflammatory and anticancer potential

Abstract A novel Schiff base ligand, (E)-2-(2-((2,6-dichlorophenyl)amino)phenyl)-N’-(2-hydroxynaphthalen-1-yl)methylene)acetohydrazide, was synthesized and complexed with Cu(II), Ni(II), Co(II), Gd(III), La(III), and Ag(I) ions. The resulting metal complexes were characterized using FTIR, UV–Vis, 1H NMR, ESR, XRD, thermal analysis, and magnetic susceptibility measurements. Spectroscopic and magnetic data confirmed octahedral geometries for Co(II), Ni(II), and Cu(II), square antiprismatic and tricapped trigonal prismatic geometries for Gd(III) and La(III), respectively, and a tetrahedral geometry for Ag(I). The anti-inflammatory activity was evaluated using ELISA-based COX-1/COX-2 inhibition assays. The HDN ligand exhibited potent COX-2 inhibition (IC₅₀ = 0.06 µM) with a selectivity index (SI) of 174.5, outperforming diclofenac sodium (SI = 4.52) and indomethacin (SI = 1.25), and approaching rofecoxib (SI = 725). The Cu(II) and Gd(III) complexes showed strong cytotoxicity in MTT assays against MCF-7 (IC₅₀ = 0.65 and 0.80 µM) and HepG-2 (IC₅₀ = 1.00 and 2.47 µM) cell lines, significantly surpassing the HDN ligand and standard drugs such as 5-fluorouracil (IC₅₀ = 3.95 µM, MCF-7) and cisplatin (IC₅₀ = 15.24 µM, MCF-7). In silico studies including molecular docking and ADME profiling supported the experimental findings. The HDN ligand and Cu(II) complex exhibited strong binding affinities to COX-2 (− 22.87 kcal/mol), HepG-2 (− 32.56 and − 31.76 kcal/mol), and MCF-7 (− 25.29 and − 20.20 kcal/mol) receptors. SwissADME and BOILED-Egg models predicted high gastrointestinal absorption for the HDN ligand and favorable pharmacokinetic profiles for the metal complexes. The present results provide a preclinical proof-of-concept for diclofenac-derived Schiff base metal complexes as dual anti-inflammatory and anticancer agents. However, their therapeutic potential remains to be validated through stability, mechanistic, and in vivo investigations.

Acoustic spin skyrmion molecule lattices enabling stable transport and flexible manipulation

Mental health assessment model for college students based on facial expression recognition

Enhanced optical visible light positioning via RSS-based evolutionary optimization under LOS and NLOS conditions

A nickel gallium oxide chlorophyll mimic for green methanol synthesis

Characterisation of pharmacogenomic variation in the Shetland and Orkney Isles in Scotland

Abstract Genetic variation is partly responsible for variability in drug response across populations. However, the full catalogue of pharmacogenetic variants and their distribution are yet to be established, thus posing challenges in implementing individualised medicine in understudied populations. This study aimed to characterise variation in key drug response genes across founder populations from the Northern Isles of Scotland. We analysed whole genome sequence datasets from 498 Shetlanders and 1372 Orcadians, the majority of whom are research participants in the Viking Genes programme, and compared the genetic variation in 41 selected pharmacogenes with observed distributions in other European datasets. From this gene-set, we present frequencies of known and potentially novel star alleles (haplotypes and structural variants) for 18 core pharmacogenes analysed using StellarPGx, and variant distributions in 23 other selected pharmacogenes with existing clinical annotations in ClinPGx ( https://www.clinpgx.org ). Despite important differences in the frequencies of rare and/or novel potentially high-impact variants, the distributions of the well-studied common actionable pharmacogene star alleles do not vary dramatically across Shetland, Orkney, and the European populations represented in the 1000 Genomes Project or allele frequency meta-analyses in ClinPGx. Importantly, for gene-drug pairs with Clinical Pharmacogenetics Implementation Consortium Guidelines, we estimated (based on diplotypes alone) that the proportion of participants in the combined dataset that may benefit from a change in dose/drug ranged from 0 to 50.5%, depending on the gene-drug pair. Overall, understanding the landscape of pharmacogenomic variation in Shetland and Orkney is an important step towards implementation of precision medicine across rural Scotland.

Effectiveness of YOLO variants for small object detection in SAR images using a new dataset

Abstract Small object detection in SAR imagery remains challenging due to limited availability of specialized datasets. The article presents a new SAR dataset designed for small-object detection. Due to the absence of publicly available datasets dedicated to vehicle detection on satellite radar imagery, a custom dataset containing 23,644 manually labelled vehicles was created using Capella and ICEYE imagery Also the results of an extensive comparative analysis of three YOLO architectures (versions 7, 8, and 12) in the task of detecting small vehicles in radar imagery were presented. The study also considers the influence of image filtering on detection effectiveness. Experimental results provided new insights into fine-tuning YOLO architectures specifically for detecting small objects in synthetic aperture radar (SAR) images. In addition, the SIVED (SAR Image dataset for VEhicle Detection) dataset (high-resolution airborne imagery) was used in the study. Model performance was tested under various configurations and with Lee, Frost, and GammaMAP filters. Furthermore, a detailed analysis of model stability was performed. The experimental results revealed notable differences in performance among the tested models. The YOLOv8 model achieved the highest detection performance on the SIVED dataset, with an F1-score of 0.958 and mAP@[0.5:0.95] of 0.838 in the unfiltered scenario, along with high stability with respect to changes in threshold parameters. The YOLOv12 model demonstrated its best performance after Lee filtering (F1 score = 0.951, mAP@[0.5:0.95] = 0.774), indicating a greater sensitivity to the quality of the input data. On the contrary, the YOLOv7 model exhibited high sensitivity to changes in confidence thresholds, necessitating precise parameter tuning. The conducted research has shown that YOLOv8 achieves superior detection performance on satellite radar imagery samples despite not incorporating advanced self-attention mechanisms. This work contributes significantly to automatic object detection in radar images, providing practical guidelines for selecting and configuring YOLO models according to the characteristics of the SAR data.

Chromosome compartment assembly is essential for subtelomeric gene silencing in trypanosomes

Abstract Genome three-dimensional organization is essential for eukaryotic gene expression. The chromosomes of the pathogen Trypanosoma brucei contain hundreds of silent variant surface glycoprotein (VSG) genes in subtelomeric regions. T. brucei transcribes a single VSG gene and periodically changes the VSG expressed, altering its surface coat to escape host antibodies by antigenic variation. We show that T. brucei core and subtelomeric chromosome compartments are separated by distinct boundaries and display topologically associating domains and loops. Chromosomes co-interact through compartment boundaries, which insulate silent subtelomeric from transcribed core compartments. We uncover chromatin-associating factors at the boundaries, including repressor-activator protein 1 (RAP1), which spreads over silent compartments. Inactivation of the RAP1 regulator, phosphatidylinositol phosphate 5-phosphatase, removes RAP1 from boundaries and subtelomeric compartments, disrupting chromatin compartment contacts and activating all VSG genes. The data show spatial segregation of repressed from transcribed chromatin and phosphoinositide regulation of compartment assembly and genome organization.

Habitual violent media exposure does not bias facial emotional processing: a comparison of interactive vs. non-interactive content

Abstract The relationship between violent media exposure and aggression remains widely debated. The General Aggression Model (GAM) suggests that repeated exposure fosters aggression through the formation of aggressive scripts and emotional desensitization. However, empirical findings have been mixed, and a growing body of research has criticized the GAM for overstating the strength and consistency of this association. In light of this, the present study investigated whether habitual exposure to violent media biases emotional information processing and whether such effects differ by mode of media interactivity. A total of 54 violent media users participated, comprising violent video gamers (VVGs; M age = 20.07 years, SD  = 1.26) and non-video gamers (NVGs; M age = 21.29 years, SD  = 1.10). Results showed a robust happy-face advantage and lower recognition capacity for negative emotions across both groups. These findings challenge the script theory and desensitization assumptions. Interestingly, when trait aggression was included as a covariate in the ANCOVA, the emotional effects were consistently attenuated or rendered nonsignificant. This pattern suggests that media effects are shaped by personal predispositions rather than acting in isolation. While the findings challenge the claim that violent media increases aggression, caution is needed when generalizing to individuals with higher levels of dispositional aggression.

Smart Strategies for Improving Electric Vehicle Battery Performance and Efficiency

Abstract The increasing demand for Electric Vehicles (EVs) necessitates accurate range prediction and optimization of driving parameters to address range anxiety and improve user experience. This study proposes a machine learning-based framework for predicting EV range, optimum acceleration, and velocity using a synthetically generated dataset of 2,000 samples designed to reflect real-world driving scenarios. Four models—Random Forest (RF), Extra Trees (ET), Linear Regression (LR), and Long Short-Term Memory (LSTM)—were evaluated individually and in ensemble combinations. To ensure statistical reliability, all models were trained and tested over ten independent runs with randomized data partitions, and the results were reported as average performance with standard deviations. The ensembles consistently outperformed individual models, with the full ensemble (RF + ET + LSTM + LR) achieving the most robust performance across all metrics (MAE, MSE, and R²). Furthermore, a real-time web application was developed using the trained models to dynamically estimate driving parameters. The findings highlight the potential of integrating AI-driven predictive modelling into EV systems to support efficient driving behaviour and energy management.