Small Input Research Articles

DFTSA-Net: Deep Feature Transfer-Based Stacked Autoencoder Network for DME Diagnosis

Diabetic macular edema (DME) is the most common cause of irreversible vision loss in diabetes patients. Early diagnosis of DME is necessary for effective treatment of the disease. Visual detection of DME in retinal screening images by ophthalmologists is a time-consuming process. Recently, many computer-aided diagnosis systems have been developed to assist doctors by detecting DME automatically. In this paper, a new deep feature transfer-based stacked autoencoder neural network system is proposed for the automatic diagnosis of DME in fundus images. The proposed system integrates the power of pretrained convolutional neural networks as automatic feature extractors with the power of stacked autoencoders in feature selection and classification. Moreover, the system enables extracting a large set of features from a small input dataset using four standard pretrained deep networks: ResNet-50, SqueezeNet, Inception-v3, and GoogLeNet. The most informative features are then selected by a stacked autoencoder neural network. The stacked network is trained in a semi-supervised manner and is used for the classification of DME. It is found that the introduced system achieves a maximum classification accuracy of 96.8%, sensitivity of 97.5%, and specificity of 95.5%. The proposed system shows a superior performance over the original pretrained network classifiers and state-of-the-art findings.

Influence of laser ignition on characteristics of an engine for hydrogen enriched CNG and iso-octane usage

The study has focused on determining the laser plug effects on engine characteristics and the laser plug usage results have compared with spark plug usage. The laser ignition technique is a type of new ignition technique and an important solution that can make combustion systems more efficient. The testing of an engine with a laser plug is the novelty of the study and the tests were carried out with reference to equivalence ratio and plug power ranges. The behaviors of the engine at full load were examined so experimentally for both ignition techniques at hydrogen enriched CNG and iso-octane mixture usage. The tests were carried out for variations of 0.4–2.0 equivalence ratio and 20–120 W plug power. A mixture that 90% iso-octane and 10% HCNG in mass was used at two ignition modes in tests for 3300 rpm maximum engine torque speed. Also, the flame formation and propagation for both ignition techniques were detected via a high-speed camera. The tests have shown the laser ignition leads to more energy consumption in the rich mixture conditions and also, less energy is required in the lean conditions. The laser ignition discharge has extended the engine's lean combustion limits via a small energy input at the tests. The high-speed camera images have shown that the laser ignition reduces the Kernel flame formation and propagation time. The laser ignition technique was produced less NOx than the conventional spark ignition method.

A Survey on Statistical, Information, and Estimation—Theoretic Views on Privacy

Privacy has become an emerging challenge in both information theory and computer science due to massive (centralized) collection of user data. In this article, we overview privacy-preserving mechanisms and metrics from the lenses of information theory, and unify different privacy metrics, including f-divergences, Rényi divergences, and differential privacy (DP), in terms of the probability likelihood ratio (and its logarithm). We review recent progress on the design of privacy-preserving mechanisms according to the privacy metrics in computer science, where DP is the standard privacy notion which controls the output shift given small input perturbation, and information theory, where the privacy is guaranteed by minimizing information leakage. In particular, for DP, we include its important variants (e.g., Rényi DP, Pufferfish privacy) and properties, discuss its connections with information-theoretic quantities, and provide the operational interpretations of its additive noise mechanisms. For information-theoretic privacy, we cover notable frameworks, including the privacy funnel, originated from rate-distortion theory and information bottleneck, to privacy guarantee against statistical inference/guessing, and information obfuscation on samples and features. Finally, we discuss the implementations of these privacy-preserving mechanisms in current data-driven machine learning scenarios, including deep learning, information obfuscation, federated learning, and dataset sharing.

Methods for Small Area Population Forecasts: State-of-the-Art and Research Needs.

Small area population forecasts are widely used by government and business for a variety of planning, research and policy purposes, and often influence major investment decisions. Yet, the toolbox of small area population forecasting methods and techniques is modest relative to that for national and large subnational regional forecasting. In this paper, we assess the current state of small area population forecasting, and suggest areas for further research. The paper provides a review of the literature on small area population forecasting methods published over the period 2001–2020. The key themes covered by the review are extrapolative and comparative methods, simplified cohort-component methods, model averaging and combining, incorporating socioeconomic variables and spatial relationships, ‘downscaling’ and disaggregation approaches, linking population with housing, estimating and projecting small area component input data, microsimulation, machine learning, and forecast uncertainty. Several avenues for further research are then suggested, including more work on model averaging and combining, developing new forecasting methods for situations which current models cannot handle, quantifying uncertainty, exploring methodologies such as machine learning and spatial statistics, creating user-friendly tools for practitioners, and understanding more about how forecasts are used.Supplementary InformationThe online version contains supplementary material available at 10.1007/s11113-021-09671-6.

Face recognition using facial features

Face recognition is in demand today due to its utility in person identification, where a small input image or a glimpse can help recognise the person from a database. In India this method of identification is not that popular but it is the need of the hour, biometric identification systems should also have facial recognition, so as to make the system contactless, and more robust in terms of security. Therefore, in this paper the authors have tried to analyse the aptness of the two face recognition algorithms namely facenet and facial landmarks for (i) Crime detection, (ii) face verification and (iii) human tracking. The features are extracted and pre-processed, converted into an array and the region of interest is matched to identify the person. The features selected are Euclidean distance between eyes, structure of nose, lip to lip distance. This project will help in securing industrial area where there is possibility of trespassing, person identification from the crowd, tracking missing personal and help keep a track on some anti-social elements of the society. It has a great application in places with high security like the parliament or bank vault etc.

Can local nutrient-circularity and erosion control increase yields of resource-constraint smallholder farmers? A case study in Kenya and Uganda

As many regions in sub-Saharan Africa, the border region of Kenya (KE) and Uganda (UG) has faced a declining soil fertility for decades, resulting from soil erosion, intensely managed agricultural soils due to population pressure and small inputs of mineral and organic fertilizers. With limited financial means, farmers need measures and/or technologies that effectively reduce nutrient losses or increase inputs at a low cost. In this study, four such measures are in focus, namely erosion reduction practices, vermicomposting of animal manure, collection of human urine in jerry cans and, collection of human excreta in urine-diverting dry toilets. Current soil nutrient balances in five districts in the Sio-Malaba-Malakisi River Basin and the potential of these measures to reduce the soil nutrient deficit are studied using the method of material flow analysis and the software STAN. Furthermore, crop-nutrient-response functions are used to determine their potential impact on maize harvests. Overall, results reveal that there exists a non-negligible and exploitable potential of local resources to reduce the soil nutrient deficit, improve harvests and in turn food security of the smallholder farmers in the region. Soil nutrient deficits could be reduced by 20–30%, 23–42% and 9–15% for nitrogen (N), phosphorus (P) and potassium (K), respectively. Subsequently, maize harvests could be increased by 8–40%, depending on the applied technology and area. This research provides useful insights for agricultural extension workers, politicians and researchers alike, highlighting that simple and easily available technologies can harness similar amounts of nutrients as more complex and expensive ones if all specific technology-constraints are adequately incorporated in the analyses.

Multi-Scale Input Strategies for Medulloblastoma Tumor Classification using Deep Transfer Learning

Abstract Medulloblastoma (MB) is a primary central nervous system tumor and the most common malignant brain cancer among children. Neuropathologists perform microscopic inspection of histopathological tissue slides under a microscope to assess the severity of the tumor. This is a timeconsuming task and often infused with observer variability. Recently, pre-trained convolutional neural networks (CNN) have shown promising results for MB subtype classification. Typically, high-resolution images are divided into smaller tiles for classification, while the size of the tiles has not been systematically evaluated. We study the impact of tile size and input strategy and classify the two major histopathological subtypes-Classic and Desmoplastic/Nodular. To this end, we use recently proposed EfficientNets and evaluate tiles with increasing size combined with various downsampling scales. Our results demonstrate using large input tiles pixels followed by intermediate downsampling and patch cropping significantly improves MB classification performance. Our top-performing method achieves the AUC-ROC value of 90.90% compared to 84.53% using the previous approach with smaller input tiles.

Bending response optimization of an ionic polymer-metal composite actuator using orthogonal array method

In the recent past decades, Ionic polymer-metal composites (IPMCs) have been researched in detail due to their flexible characteristics and light in weight. IPMC is most well-known electro-active polymers (EAPs), which can be used in the medical applications, biomimetic, micro robotic actuators, artificial muscles, dynamic sensors and underwater propulsions. Due to their large bending actuation with low voltage sources and biocompatibility, IPMCs are receiving more and more attention to many researchers in recent days. To reduce the experimental costs, the principal factors which are responsible for more bending response are designed in this work using Taguchi orthogonal array method. Multiple process parameters include input voltage and thickness of the fabricated Ag-IPMC actuator was optimized to produce a maximum bending response. Experiments were carried out using L9 orthogonal array to recognize the optimal process parameters. In the optimized conditions, the IPMC actuator exhibited a noteworthy improvement in bending deflection compared to the normal conditions. Under the application of 1.0 V, the bending response increases around 30% in the optimised condition compared to normal IPMC actuator. This study lays a solid foundation for developing high performance IPMC actuator where higher bending displacement is required i.e., in the robotic and biomimetic application with a very small input voltage.

Evaluation of multi-level Bang–Bang phase detector with metastability effect using Markov chain

Bang–Bang phase detector (BBPD) as a bistable system has the metastability problem. In addition, BBPD is a non-linear block in the phase locked-loop (PLL) and clock and data recovery (CDR) that makes their analysis complicate. To simplify the analysis of the non-linear BBPD, the linear expression for the gain of multi-level BBPD (ML-BBPD) is derived in this paper by considering the metastability failure. The ML-BBPD gain is developed based on Markov chain to provide more realistic model. Simulation results show that the metastability failure not only attenuates the gain of ML-BBPD for the small input jitter, but also increases the output jitter and consequently the BER of the CDR. It is shown that by increasing the number of sampling clocks the output jitter and BER is reduced. Also, the gain of ML-BBPD is decreased for small values of the input jitter, but for the large values of input jitter ML-BBPD gain is the same for every over sampling ratio.

Widely flatness gain bandwidth with double pass parallel hybrid fiber amplifier

A Widely flatness gainbandwidth with double pass parallel hybrid fiber amplifier is experimentally demonstrated in this study. The proposed design combines serial erbium–Raman fiber amplifier in one branch and Raman fiber amplifier in the second branch. Multiple Raman pump units with a maximum power of 800 mW (250 mW of 1410 nm, 225 mW of 1480 nm, and 325 mW of 1495 nm) are utilized. Pump recycling technique is applied to achieve acceptable pumping efficiency. A maximum flatness gain bandwidth of 80 nm (1525–1605 nm) and average gain level of 22.5 dB are obtained at a small input signal power of \(-25\) dBm and optimum pump power values. While, a wider flatness gain of 90 nm (1520–1610 nm) and average gain level of 11.5 dB are achieved at a large input signal power of \(-5\) dBm.

A Note on the Cardinalities of Sets of Scalar Alternatives

AbstractFormal theories of scalar implicature appeal crucially to a set of alternatives. These are the alternative statements that a speaker could have made but chose not to in pragmatic accounts, and the alternative statements that figure in the computation of exhaustivity operators in grammatical approaches. I show that the three sufficiently explicit theories of alternatives in the literature generate sets of alternatives that grow at least exponentially as a function of the input, and that these theories generate very large sets even for relatively small inputs. For pragmatic accounts of scalar implicature, I argue these results are hard or impossible to square with what we know independently about manipulating alternatives from the psychology of human reasoning. I propose that they pose a weaker but more general challenge for grammatical approaches, since alternatives as required by exhaustivity operators occur elsewhere in grammar, for example as part of the semantics of operators like “only” and “even.”

Measuring Current in a Power Converter Using Fuzzy Automatic Gain Control

The accuracy of current measurements can be increased by appropriate amplification of the signal to within the measurement range. Accurate current measurement is important for energy monitoring and in power converter control systems. Resistance and inductive current transducers are used to measure the major current in AC/DC power converters. The output value of the current transducer depends on the load motor, and changes across the whole measurement range. Modern current measurement circuits are equipped with operational amplifiers with constant or programmable gain. These circuits are not able to measure small input currents with high resolution. This article proposes a precise loop gain system that can be implemented with various algorithms. Computer analysis of various automatic gain control (AGC) systems proved the effectiveness of the Mamdani controller, which was implemented in an MCU (microprocessor). The proposed fuzzy controller continuously determines the value of the conversion factor. The system also enables high resolution measurements of the current emitted from small electric loads (≥1 A) when the electric motor is stationary.

Are Climate Change Policies Politically Costly?

Are policies designed to avert climate change (Climate Change Policies, or CCPs) politically costly? Using data on governmental popular support and the OECD’s Environmental Stringency Index, we find that CCPs are not necessarily politically costly: policy design matters. First, only market-based CCPs (such as emission taxes) generate negative effects on popular support. Second, the effects are muted in countries where non-green (dirty) energy is a relatively small input into production. Third, political costs are not significant when CCPs are implemented during periods of low oil prices, generous social insurance and low inequality.

Data-driven inference of high-accuracy isostable-based dynamical models in response to external inputs.

Isostable reduction is a powerful technique that can be used to characterize behaviors of nonlinear dynamical systems using a basis of slowly decaying eigenfunctions of the Koopman operator. When the underlying dynamical equations are known, previously developed numerical techniques allow for high-order accuracy computation of isostable reduced models. However, in situations where the dynamical equations are unknown, few general techniques are available that provide reliable estimates of the isostable reduced equations, especially in applications where large magnitude inputs are considered. In this work, a purely data-driven inference strategy yielding high-accuracy isostable reduced models is developed for dynamical systems with a fixed point attractor. By analyzing steady-state outputs of nonlinear systems in response to sinusoidal forcing, both isostable response functions and isostable-to-output relationships can be estimated to arbitrary accuracy in an expansion performed in the isostable coordinates. Detailed examples are considered for a population of synaptically coupled neurons and for the one-dimensional Burgers' equation. While linear estimates of the isostable response functions are sufficient to characterize the dynamical behavior when small magnitude inputs are considered, the high-accuracy reduced order model inference strategy proposed here is essential when considering large magnitude inputs.

Two Sides of the Same Coin: Boons and Banes of Machine Learning in Hardware Security

The last decade has witnessed remarkable research advances at the intersection of machine learning (ML) and hardware security. The confluence of the two technologies has created many interesting and unique opportunities, but also left some issues in their wake. ML schemes have been extensively used to enhance the security and trust of embedded systems like hardware Trojans and malware detection. On the other hand, ML-based approaches have also been adopted by adversaries to assist side-channel attacks, reverse engineer integrated circuits and break hardware security primitives like Physically Unclonable Functions (PUFs). Deep learning is a subfield of ML. It can continuously learn from a large amount of labeled data with a layered structure. Despite the impressive outcomes demonstrated by deep learning in many application scenarios, the dark side of it has not been fully exposed yet. The inability to fully understand and explain what has been done within the super-intelligence can turn an inherently benevolent system into malevolent. Recent research has revealed that the outputs of Deep Neural Networks (DNNs) can be easily corrupted by imperceptibly small input perturbations. As computations are brought nearer to the source of data creation, the attack surface of DNN has also been extended from the input data to the edge devices. Accordingly, due to the opportunities of ML-assisted security and the vulnerabilities of ML implementation, in this paper, we will survey the applications, vulnerabilities and fortification of ML from the perspective of hardware security. We will discuss the possible future research directions, and thereby, sharing a roadmap for the hardware security community in general.

Vessel segmentation from volumetric images: a multi-scale double-pathway network with class-balanced loss at the voxel level.

Vessel segmentation from volumetric medical images is becoming an essential pre-step in aiding the diagnosis, guiding the therapy, and patient management for vascular-related diseases. Deep learning-based methods have drawn many attentions, but most of them did not fully utilize the multi-scale spatial information of vessels. To address this shortcoming, we propose a multi-scale network similar to the well-known multi-scale DeepMedic. It also includes a double-pathway architecture and a class-balanced loss at the voxel level (MDNet-Vb) to achieve both the computation efficiency and segmentation accuracy. The proposed network consists two parallel pathways to learn the multi-scale vessel morphology. Specifically, the pathway with a normal resolution uses three-dimensional (3D) U-Net fed with small inputs to learn the local details with relatively small storage and time consumption. The pathway with a low-resolution employs 3D fully convolutional network (FCN) fed with downsampled large inputs to learn the overall spatial relationships between vessels and adjacent tissues, and the morphological information of large vessels. To cope with the class-imbalanced issue in vessel segmentation, we propose a class-balanced loss at the voxel level with uniform sampling strategy. The class-balanced loss at the voxel level re-balances the loss function with a coefficient that is inversely proportional to the normalized effective number at the voxel level of each class. The uniform sampling strategy extracts training data by sampling uniformly from two classes in every epoch. Our MDNet-Vb outperforms several state-of-the-art methods including ResNet, DenseNet, 3D U-Net, V-Net, and DeepMedic with the highest dice coefficients of 72.91% and 69.32% on cardiac computed tomography angiography (CTA) dataset and cerebral magnetic resonance angiography (MRA) dataset, respectively. Among four different double-pathway networks, our network (3D U-Net+3D FCN) not only has the fewest training parameters and shortest training time, but also gets competitive dice coefficients on both the CTA and MRA datasets. Compared with classical losses, our class-balanced focal loss (FL-Vb) and dice coefficient loss at the voxel level (Dsc-Vb) alleviates class imbalanced issue by improving both the sensitivity and dice coefficient on the CTA and MRA datasets. Moreover, simultaneously training on two datasets shows that our method has the highest dice coefficient of 73.06% and 65.40% on CTA and MRA datasets, respectively, outperforming the commonly used methods, such as U-Net and DeepMedic, which demonstrates the generalization potential of our network for segmenting different blood vessels. Our MDNet-Vb method demonstrates its superiority over other state-of-the-art methods, on both cardiac CTA and cerebral MRA datasets. For the network architecture, the MDNet-Vb combined the 3D U-Net and 3D FCN, which dramatically reduces the network parameters yet maintains the segmentation accuracy. The class-balanced loss at the voxel level further improves accuracy by properly alleviating the class-imbalanced issue between different classes. In summary, MDNet-Vb is promising for vessel segmentation from various volumetric medical images.

Strain-Mitigated Two Dimensional Based Electronics for Integrated Wearable Sensor-Systems

The ability to fabricate incredibly sophisticated integrated circuits at low cost rendered modern electronics more affordable, efficient, and functional. This monumental advance in technology would usher in the age of the Internet-of-things and transform wearable electronics from a largely academic pursuit into an everyday reality around which a robust and rapidly growing industry has formed. However, even the highly connected wearable electronics of today leave much to be desired with respect to their compatibility with soft tissues such as skin. More specifically, the poor compliance of the rigid, silicon-based integrated circuits conventionally used in the construction of these devices with the soft, curvilinear surfaces of the human body can result in high impedances at the skin-sensor interface which in turn contribute to low signal-to-noise ratios as well as poor sensitivity and accuracy. Motion artifacts originating from the poor conformability of such devices only exacerbate these problems. This mechanical mismatch represents a significant hurdle in the effort to move away from the conventional reactive approach to healthcare towards a more proactive approach that leverages ‘truly’ wearable sensors to improve health outcomes.Transistors are the building blocks upon which the vast majority of modern electronics are built. Although single transistors can be used to amplify or otherwise control electronic signals using only a small input signal, most of their functionality comes in the form of integrated circuits (ICs) consisting of many connected transistors. Conventional techniques for fabricating ICs exploit the rigidity of the semiconducting silicon substrate to ensure the device performance is consistent throughout the circuit. In contrast, integrated circuits for highly compliant wearable electronics that can conform even to microscopic features of the skin[1,2] face the inherent challenge of maintaining their electrical performance under mechanical deformations. As is the case with most sensors, the front-end circuitry for these devices (responsible for conveying the information contained in the signal – either wirelessly or without wires – to the end user) requires reliable and performant integrated circuits. Of course, integrating soft, ultra-thin sensors with the rigid silicon-based integrated circuits results in a mechanical mismatch with mechanical failure at the interface where the two components meet being the most likely scenario.Instead, here we present an approach which uses two-dimensional (2D) transistors in a specially designed strain-neutralizing configuration which circumvents much of the difficulties associated with the aforementioned mismatch. The extraordinary properties of 2D materials—particularly their excellent mechanical flexibility, ultimate thinness, optical transparency, and favorable transport properties for realizing electronics—make them prime candidates for replacing conventional silicon-based transistor circuits in next-generation wearable applications. Using finite-element analysis we have designed and simulated a compliant, strain-insensitive substrate which can resist uniaxial applied strains upwards of 35%. Moreover, simulations of the electronic band structure of the transistors under applied strain is used to optimize the circuit design and further minimize the effect of strain in the transistors. These electronic and mechanical simulations are used together in the configuration and subsequent fabrication of fully integrated strain-neutralized 2D transistors compatible with state-of-the-art soft stretchable wearable sensors.[1]S. Kabiri Ameri et al., “Graphene electronic tattoo sensors,” ACS Nano, 11, 7634–7641, 2017.[2] S. Kabiri Ameri et al., “Imperceptible electrooculography graphene sensor system for human–robot interface”, npj 2D Materials and Applications, 2, 1-7, 2018.

OP14 Extracellular RNAs as liquid biopsy non-invasive biomarker in IBD

Abstract Background Extracellular RNAs (exRNAs) are RNA species present outside of the cells in which they were transcribed. They are found in human serum, though the exact role of circulating exRNAs remains to be established, especially in inflammatory bowel diseases (IBD). Besides their potential help in our pathophysiological understanding of disease, they might serve as liquid biopsies or non-invasive biomarkers. We characterised exRNAs in serum from IBD patients, and questioned their potential in separating ulcerative colitis (UC) from Crohn’s disease (CD). Methods We carried out SILVER-seq (Small Input Liquid Volume Extracellular RNA-sequencing) on serum droplets (5-7ml) from a cross-sectional cohort of 26 IBD patients (15 UC, 11 CD) with active endoscopic disease (Mayo endoscopic sub score or Simple Endoscopic Score for Crohn’s disease) (Table 1). Normalization and differential expression were done using DESeq2 R package, co-expression network analyses performed using WGCNA (FDR adjusted p ≤0.05). Using randomized generalized linear modelling (RGLM), a diagnostic exRNA marker was designed to separate UC from CD samples (15 UC, 11 CD). Results We detected 60,675 exRNAs in serum from IBD patients, capturing 76.1% of all genes expressed in intestinal tissue, and including highly abundant intestinal genes (e.g MUC2) and intestinal barrier genes (e.g claudin 8, occludin and RETNLB). Co-expression network analysis identified 69 clusters of which 1 significantly correlated with the distinction between CD and UC (FDR p=0.003, r=-0.70). One of the hub genes within this module (consisting of 148 genes, upregulated in UC) was GNA12 (p=2.3E-4, r=0.66 for correlation with the module eigengene), encoding for a membrane bound GTPase that plays a key role in tight junction assembly and has previously been identified as UC-specific SNP in GWAS (Figure 1). Serum GNA12 expression was not associated with faecal calprotectin (p=0.55, r=0.12), disease duration (p=0.24, r=0.25), age (p=0.43, r=0.16) or gender (p=1.0), but did correlate with other UC-specific genes including TNFRSF14 (p=0.04, r=0.4), HNF4A (p=0.04, r=-0.4) and CAMK2A (p=0.004, r=0.54). Through machine learning within the UC-specific module (containing 148 genes), we identified an 8-gene exRNA panel, including GNA12, that could accurately discriminate between UC and CD patients (accuracy 96.2%). Figure 1: Visualisation of the identified exRNA network including GNA12 Conclusion Liquid biopsies are a novel non-invasive tool in IBD biomarker development. Although larger in-depth studies are required to further validate, explore and characterise the potential of serum exRNAs in the field of IBD, the current pilot project identified a new non-invasive tool to accurately distinguish CD from UC patients.

Input aperture restriction of the spatial spectral compressive spectral imager and a comprehensive solution for it.

Compressive spectral imaging (CSI) is an attractive spectral imaging technique since it could acquire a spectral image data cube in a single snapshot. One notable CSI scheme is the spatial spectral compressive spectral imager (SSCSI), which has low complexity and high quality of the recovering spectral image. However, the SSCSI suffers from a small input aperture, which reduces the optical efficiency and signal-to-noise ratio of the system. In this paper, the effect of the input aperture size on the SSCSI system is analyzed. It shows that with the increase of input aperture, the incident light from different spectral bands will overlap with each other on the mask, and the encoding pattern of each spectral band will be ambiguous. Thus, the reconstruction quality of the data cube will highly deteriorate. A new scheme is proposed to deal with this problem. First, the observed image is resampled and recombined into new sub-observed images to improve the frequency response of the encoding pattern. Then each sub-observed image is divided into multiple sub-sets to reduce the coherence of the sensing matrix. Compared to the original reconstruction algorithm for the SSCSI system, the peak signal-to-noise ratio (PSNR) is promoted by more than 3dB, and the spectral reconstruction accuracy and noise suppression capability are also improved.

Role of riverine inputs, low saline plume advection and mesoscale physical processes in structuring the Chlorophyll a distribution in the western Bay of Bengal during Fall Inter Monsoon

This study delineates the role of small and medium river inputs, Low Saline Plume Advection (LSPA) and eddies in hydrography alteration and Chlorophyll a (Chl. a) in the Western Bay of Bengal. Samples were collected across five transects viz: Hooghly (HO), Mahanadi (MN), Rushikulya (RK), Visakhapatnam (VSKP) and Godavari (GD) during Fall Intermonsoon. Each transect consists of 7 or 8 locations from inshore to offshore. LSPA propagates southward concordance with the East India Coastal Current (EICC) and its southward flow strengthened by a cold-core eddy. LSPA results in the intermittent low salinity in the cross-shore section of HO, MN and RK. Upper layer Chl. a is 2–3 folds higher in inshore and in LSP-influenced locations than in its adjacent stations. The present study identified Double Chlorophyll a Maxima (DoCM) in LSPA-influenced slope regions of MN and RK. DoCM is less known in the BoB. DoCM has both the Surface Chl. a Maxima (SCM) and Subsurface Chl. a Maxima (SSCM). SSCM layer is relatively shallow and intense in slope and offshore regions of MN and RK due to their closeness with cold-core eddy. The present study highlights that freshwater discharge from small and medium rivers impacts hydrobiology around 10–50 km from the shore depends on the magnitude of river influx. LSPA is away from the local inputs and impacts hydrobiology (>500 km) along the path. EICC and eddies together regulated the direction of LSPA. Existing eddies nature alters vertical hydrobiology in slope and offshore regions.