Explore the vast range of titles available.

Pomegranate cultivation faces significant challenges due to fruit diseases that significantly impact crop yield and farmer income. Traditional methods for disease detection are often slow and prone to errors, delaying timely intervention. This paper proposes a deep learning-based system for automatic, multi-class disease classification in pomegranates using transfer learning. A dataset comprising 5099 annotated images was used to train and evaluate several CNN models, including DenseNet121, EfficientNetB0V2, MobileNetV2, ResNet50, VGG16, and InceptionV3. DenseNet121 emerged as the top performer, achieving an accuracy of 99.35%. To enhance practical value, a novel Healthy-Based Deviation Scoring (HBDS) method was developed to estimate disease severity using Grad-CAM ++ for lesion localization and Mahalanobis distance-based scoring, followed by Gaussian Mixture Model clustering. The severity predictions of the system were verified against manually labeled images, and the system has shown superior accuracy compared to pixel-based methods. Also, a recommendation module was integrated using a retrieval-augmented language model, which provides disease-specific treatment suggestions based on the predicted severity. The complete pipeline is implemented as a user-friendly web application that delivers real-time diagnosis, severity estimation, and actionable treatment plans, which offer a practical and scalable solution for modern precision agriculture.

Hierarchical ZnO@metal–organic framework@polypyrrole (ZnO@MOF@PPy) heterostructure arrays have been prepared by combining hydrothermal and chemical polymerisation processes. The structural properties and morphology of samples were determined using XRD, SEM, TEM, EDX, XPS, and BET techniques. The energy storage capability of ZnO, ZnO@MOF, and ZnO@MOF@PPy was investigated using electrochemical techniques such as cyclic voltammetry, galvanostatic charge–discharge analysis (GCD), and electrochemical impedance spectroscopy. ZnO@MOF@PPy displayed an excellent capacitance of 1181 Fg −1 at 2 Ag −1 and a respectable rate capacity of 463 Fg −1 at 10 Ag −1 in a 2 M potassium hydroxide electrolyte. According to the GCD cycles’ stability investigation, the ZnO@MOF@PPy has 92%t excellent stability after 10,000 cycles at 10 Ag −1 . Additionally, the use of ZnO@MOF@PPy as the positive electrode and activated carbon as the negative electrode in an asymmetric supercapacitor was studied. The device demonstrated a high capacity of 172 Fg −1 at 2 Ag −1 in the potential window of 0.0–1.6 V with suitable cycling stability of 81% and columbic efficiency of 72% after 10,000 continuous cycles in the current density of 10 Ag −1 , according to the results. This study demonstrated ZnO@MOF@PPy supercapacitive potential as a potential electrode for supercapacitors.

Despite the growing impacts of environmental changes due to smart city development, sustainable Internet of Health Things (IoHT) retains improved public health. Containment of contagious diseases is one of the prime factors as the population density in the smart city environment is growing exponentially. This work focuses on the prognosis of pediatric pneumonia through the IoHT framework. In the world population, nearly 15% of children under five years of mortality are caused by a lung infection called pediatric pneumonia. It kills approximately 800 thousand children every year, and 2200 children daily mortality rate due to pediatric pneumonia. The disease is caused by viral or bacterial infections in the lungs. Chest X-ray (CXR) is the predominant method for diagnosing and severity analysis of pneumonia by pediatricians. However, the CXR images are low-quality images, demanding the intelligence for accurate analysis and interpretation. Hence, researchers developed different machine learning and deep learning methods to diagnose pneumonia from CXR images in recent years. However, it lacks the accuracy of interpretations. This paper proposes a deep transfer learning-based neural network-based IoHT framework to diagnose pneumonia due to viral and bacterial infections. The proposed model is twofold: the first is the deep transfer learning network for discriminating normal CXR from pneumonia-affected lung CXR images. The second is that the deep transfer learning network is trained by an optimized training method called Adaptive Movement Estimation and deployed in IoHT. The performance of the proposed system is analyzed in terms of accuracy, sensitivity, specificity, and Area Under the Curve (AUC). It yields the highest sensitivity of 98.2% and a precision of 98.8%. The proposed system also yields a validation accuracy of 97.88, which is high compared to other state-of-the-art transfer learning methods for diagnosing pediatric pneumonia.

The research on bird classification from sound gains momentum in ornithology. This sound-based bird classification uses perceptual features with a filter bank in BARK, MEL and Equivalent rectangular bandwidth (ERB) frequency scales and log-energy in Time–frequency (T–F) unit features by taking Cochleagram on the responses of the Gammatone filter bank in BARK, MEL and ERB frequency scales. 80% of these features are given to the modelling technique for developing bootstrap aggregation for an ensemble of decision trees. 20% of the features are given to the model for classifying the bird by predicting the class for each feature vector. The performance of the system is assessed using recognition accuracy as a metric. Decision-level fusion of perceptual features with filters in different frequency scales has provided a maximum accuracy of 93%, besides delivering 100% accuracy for some bird species. Decision-level fusion of Cochleagram features with Gammatone filters in different frequency scales has yielded a maximum accuracy of 99%. This work has considered 42 bird species spanning diverse regions across the world. Ornithologists use this automated system to determine the ecosystem's health.

This paper numerically characterizes a photonic crystal fiber (PCF) surface plasmon resonance (SPR) biosensor using finite element method (FEM)-based COMSOL Multiphysics simulation software. The modal refractive index (RI) of the core-guided mode is successfully controlled by modifying the PCF’s air hole diameters, and the sensor is seen to have significant birefringence. Analyte channels are employed in the center hole, surrounded by graphene–Au bimetallic layers to reduce surrounding interference. The features of graphene allow it to minimize oxidation and improve molecular absorption. By analysis of wavelength and amplitude characteristics, the performance of the developed sensor is examined; when the analyte refractive index is 1.34, the sensor’s wavelength sensitivity is 15,000 nm/RIU for x -polarization and 10,000 nm/RIU for y -polarization. The amplitude sensitivity of x -polarization is 88 RIU −1 , while that of y-polarization is 425 RIU −1 . Both polarizations have a resolution of 1 × 10 –3 RIU. Thus, the proposed PCF can be used as the basis for building a real-time, higher-sensitivity, and fast-responding sensor.

Biometrics is the common method of securely and efficiently identifying and authenticating individuals by using unique biological features. Some common biometrics is fingerprint, speech, iris and signature. In this paper, the cryptosystem is proposed to enhance security and conserve the transmission bandwidth in implementing an authentication system The design of speech based secured authentication systems include the extraction of features from speech, creation of templates and testing procedures to authenticate the persons. The speaker recognition system is formed using Mel Frequency Cepstral coefficients (MFCC) and Recurrent Neural network (RNN) based machine learning technique. For developing a training system, MFCC features are extracted from the training data set. The RNN network is trained with features and a speakers’ template is created for each speaker. In testing phase to ensure security in speech based authentication, MFCC features are extracted from the test speech set and these features are encrypted before it gets transmitted through the unsecured channel. The proposed crypto system is developed based on 3D logistic chaotic map and DNA operation. Firstly, MFCC features derived from the test speech set are concatenated and subjected to first level diffusion and confused using 3D logistic map. The resultant is encoded as a DNA sequence E(n), using any one of the eight rules for encoding DNA. The DNA XOR operation is performed between E(n) and 3D logistic map DNA sequence L(n). Finally, the encrypted feature set is attained by DNA decoding. In the test phase, the proposed system decrypts the features and is matched with a stored trained model to locate the identity of the speaker. Overall accuracy is 88% for the text independent and 96% for the text dependent person authentication system tested with genuine utterances. This research is extended to estimate the performance against attacks utterance and propose system is assessed with respect to rejection rate.

To assist an investigation process, forensic experts compare and analyze audio recordings. Speech utterances are compared by humans and/or machines for use in court for investigation. Scientific research community insists for specific automatic or human-based approach to identify uniquxy2e audio features from identical twins group. Filters can be employed to enhance an audio recording for improving clarity. This may entail removal of unnecessary noise to enrich the intelligibility of speech. Forensic audio experts can examine a variety of characteristics of the audio recording to decide the possibility of alterations in the collected evidences. This includes confirming the integrity and authenticating that the content is what it purports to be. Thiswork named as FIST(Forensic Investigation for Twin Identification from Speech: Perceptual and Gamma-tone Features and Models) proposes an automated system to identify a twin from identical twin pairs by the use of gamma-tone features and perceptual features.The proposed features are excerpted from the set of training speeches and templates are created for each twin based on vector quantisation (VQ), Fuzzy C means clustering (FCM) and multivariate hidden Markov modelling (MHMM) techniques. For testing, features are extracted from the set of test utterances and worked out to the templates for classification. Based on the type of classifier used, classification of twin is carried out with minimum distance and maximum loglikelihood value. The proposed features are examinedfor sub-optimal and true success rates as key performance metrics to assess the system and also a comparative analysis is made across the proposed features. Among the inspected features, Gammatone energy features expose better performance in comparison to perceptual features by attaining the overall sub-optimal success rate and true success rate as97.8375% and 92.75% for Gammatone energy features with VQ based modelling technique. This work FIST has also been analysed by inducing disturbance in the form of speech interference from their own twin pairs and Gamma-tone energy feature with VQ based modelling technique performs better for twin identification. A high claim of 99.625% and 95.0625% accuracy has been achieved by employing decision level fusion classifier.

Biometrics is a common method of securely and efficiently identifying and authenticating individuals by using unique biological features. In this work, speech based person authentication system has been developed based on a different set of features such as Mel frequency cepstral coefficient (MFCC), Linear predictive cepstral coefficient (LPCC) and Perceptual features with critical band analysis done in Bark scale (PLPC), Mel Frequency (MF-PLPC) scale, and Equivalent rectangular bandwidth (ERB-PLPC) scale respectively. The implementation of a speech-based authentication system assimilates feature extraction from speech, modeling methods, and testing procedures for validating the person. Efficiency of the text dependent and text independent system is compared with respect to different features for the database encompassing genuine and attack utterances. In this work, AVSpoof and TIMIT database speech utterances are used for validating 44 and 104 speakers respectively. Efficiency of the proposed system is investigated with different metrics such as correlation coefficient, PSNR, F-ratio, recognition accuracy and ROC curve. ERBPLPC has provided better overall recognition accuracy for the system with genuine set of utterances considered for training and testing. This work is also extended to evaluate the authentication system against attack utterances chosen from speech synthesis attack, replay attack, and voice conversion attack in AVSpoof database and the performance of the system is assessed in terms of rejection rate. ERB-PLPC provides the high rejection rate as 91% on an average against all attacks and ensures that the feature selection is more robust by performing testing against spoofing attack set and genuine set of test utterances.

Classification of respiratory diseases using X-ray and CT scan images of lungs is currently practised and used by many medical practitioners for clinical diagnosis. Respiratory disease classification, using breathing and wheezing sounds, remains scarce in the research field and is slowly upcoming. In this work, robust respiratory disease classification using breathing sounds ( RRDCBS) is implemented by extracting multiple features from sounds, creating multiple modelling techniques, and experimental identification of diseases using appropriate testing procedures for multi-class and binary classification of respiratory diseases. Decision level fusion of features for Vector quantisation (VQ) modelling technique has provided 100% accuracy for classifying five respiratory diseases and healthy subjects. Decision level fusion of indices on the features has provided 100% accuracy for VQ, support vector machine (SVM), and K-nearest neighbour (KNN) modelling techniques to perform binary classification of the respiratory disease against healthy data sound. Deep recurrent and convolutional neural networks are also evaluated for multiple/binary classification of respiratory diseases.

Worldwide, recycling of bio-waste into fixable self-healing materials is a big challenge, in this paper, a feasible approach used for to synthesis partially bio-based benzoxazine from cashew nut shells. Melamine, cardanol and paraformaldehyde have been used to develop bio-based benzoxazine monomer through Mannich condensation reaction, which is then thermally co-polymerized with 2-Mercaptoethanol and Polyurethane. The structure of the monomer and composites have been confirmed by using traditional spectroscopy techniques. Our modified bio-based polybenzoxazine showed good self-healing and shape memory properties. The duration of the recovery process was significantly influenced by the thiol and polyurethane unit. The self-healing property of poly(sh- co -cdl-m), poly(u1/sh- co -cdl-m) and poly(u2/sh- co -cdl-m) was investigated with scanning electronic microscopy analysis and optical microscopy analysis. Finally, when compared with reported bio-based self-healing materials thiol and polyurethane cross-linked bio-based composites can be recyclable and heal themselves successfully up to 5 times and maximum healing time is 2 h at 50 °C. Later on, the bio-based polybenzoxazine is 100% recyclable with high T g , which is achieved the goal of cleaner production by using biomass and reducing petrochemical recourses, production and energy emission.