Theses and Dissertations
Permanent URI for this collectionhttp://ir.daiict.ac.in/handle/123456789/1
Browse
1153 results
Search Results
Item Open Access Performance Assessment of Edge Traffic Distribution Routing Algorithm for Graphene Based Network-on-Chip(2021) Gupta, Yatin Kumar; Gohel, Bakul; Agrawal, YashNetwork-on-chip (NoC) has evolved as new paradigm for high-dense interconnect configurations in advanced integrated circuit designs. The increasing numbers of transistor cores with decrease in chip area is the leading motivation behind employment of NoC over SoC architectures. NoC can be addressed a move ahead from computation-centric to communication-centric design and the implementation of scalable communication structures. NoC provides re-configurable interconnections between the different cores in SoC design. It maximizes data transfer speed and reduction in wiring congestion. For further effective enhancing performance of NoCs, it is investigated that incorporation of graphene material can be good for realizing interconnects. As the graphene has the remarkable physical properties it is one of the most important emerging research material for not only the front-end but also for the back-end devices. In this work, edge traffic distribution (ETD) algorithm is explored along with magnificent graphene based interconnects for NoC design. Performance parameters considered are delay, power, energy, and throughput. It is investigated that the ETD routing algorithm leads to reduced delay, higher throughput, and smaller packet loss. Further, it is also analyzed that if the copper based router-to-router link of a mesh based NoC is replaced by a grahene based link then it leads to smaller energy consumption whenever there is a flit transfer from one router to the other. The assessment of NoC structures has been performed using Noxim and SPICE electronic design automation tools.Item Open Access Adversarial Defense Using Partial Pseudorandom Encryption(2021) Kalgutkar, Amruta; Joshi, M. V.Machine Learning models like Deep neural networks are vulnerable to adversarial attacks. Carefully crafted adversarial examples force a learned classifier to misclassify the input which can be correctly classified by a human observer. In this thesis, we present a novel approach for defense against such Adversarial attacks. We train and test the model on transformed images in black-box and gray-box scenarios. Here, we propose a transformation technique that partially encrypts every image before training and testing using the Rivest–Shamir–Adleman (RSA) , an asymmetric-key encryption algorithm for visual encryption. The internal structure of the system and the keys generated by RSA are secret. We encrypt only those pixels which are generated by a pseudorandom number generator with a pre-decided secret seed. The images encrypted with such transformation are extremely difficult to decrypt and to launch adaptive adversarial attacks or transferability attacks which makes this visual defense technique against adversarial attack robust. As the field of Adversarial machine learning (AML) is still under study, researchers have not attempted such an approach of training the model on encrypted images for robust learning. State-of-the-art defense techniques are effective but they are computationally expensive and still will not guarantee total security. This idea of partial encryption maintains features and asymmetric key encryption makes it difficult for adversary to guess encryption parameters. This makes the technique novel and hence out-performs state-of-the-art defense techniques.Item Open Access Computational and Data Driven Approaches for Investigation of Microwave-Plasma Interaction(Dhirubhai Ambani Institute of Information and Communication Technology, 2023) Ghosh, Pratik; Chaudhury, BhaskarMicrowave-plasma interaction and High power microwave (HPM)breakdown involving plasma formation have been studied theoreticallyas well as experimentally since the 1950s for a wide variety of applications.Microwave plasma interaction can be classified into two broad categories,firstly involving low power non-ionizing waves and secondly high powerionizing waves leading toHPMbreakdown. Early studies onHPMbreakdownprimarily focused on the determination of the breakdown field as a functionof pressure, frequency and pulse duration. However, only recently, detailedexperimental investigations of the plasma dynamics during breakdownhave been possible with the use of sophisticated high-speed ICCD cameras.Particularly, in the past few years, several experiments and numericalsimulations using millimeter and sub-millimeter wave irradiation ( 100 GHz)at high pressures (ten to hundreds of Torr) have been carried out. The renewedinterest in this area is primarily because of two reasons. Firstly, the potentialapplications of such discharges to aerodynamic flow control, combustionignition, flame stabilization and to propulsion have been investigated veryrecently. Secondly, the dynamics of high frequency wave breakdown at highpressures leading to formation of complex plasma structures (spatio-temporalpropagation of plasma) such as self-organized plasma arrays is a subject ofgreat interest from scientific point of view.To completely understand the physics and properties of different types ofdischarges associated with microwave breakdown, it is crucial to furtherimprove our current understanding of the microwave-plasma interactionand plasma formation at high pressures. To fully utilize the potential of thispromising area of research, it is crucial to understand microwave-plasmainteractions, both in the context of low-power non-ionizing microwaves andwhen the power is sufficient to ionize the gaseous species and form plasma.Modeling and simulation of the strong coupling between the high frequencyEM waves and the plasma is still a challenging research problem due tothe different time and space scales involved in the process. Particularlyaccurate 2D/3D simulations are computationally very expensive and werequire new efficient computational approaches to investigate this problemfor real life applications. Most of the computational studies reported in theliterature till now (particularly recent 2D simulations) have focused only onthe wave scattering by the plasma and ionization-diffusion mechanism forplasma evolution (time scale of 100s of nanoseconds) due to computationalconstraints. Researchers have primarily studied this problem using asimple model wherein Maxwell�s equations have been coupled with plasmacontinuity equations and these models have been used to investigate theplasma dynamics in nanosecond timescales.As a first step, we have developed a comprehensive computational modelfor investigating microwave-plasma interaction and different kinds ofmillimeter wave breakdown at high pressures. An in-house 2D simulatorhas been implemented in C language and the validity of the code has beenestablished by directly comparing the simulation results with the experimentalobservations available in the literature. The computational tool consist ofthree computational solvers (EM wave solver, Plasma solver and Fluid solver)coupled with each other. The inputs to this computational tool are the fieldstrength of the EM wave, frequency of the wave, pressure and gas details. Theimportant output required for investigating the physics of plasma dynamicsare: plasma density, electric field distribution, electron temperature, gasdensity distribution etc.As a second step, to address the computational challenges associated withsuch simulations, a self-aware mesh refinement algorithm has been presentedthat uses a coarse mesh and a fine mesh that dynamically expands based on theplasma profile topology to resolve the sharp gradients in E-fields and plasmadensity in the breakdown region. The dynamic mesh refinement (DMR)technique is explained in detail, and its performance has been evaluatedusing two metrics, the accuracy and efficiency, on a standard benchmarkmicrowave breakdown problem. Different 2-D simulations are performed tocapture the front velocity and the filamentary pattern formation, and, resultsare compared for DMR (different refinement factors (r = 2, 4)) with the resultsobtained from uniform fine mesh. From the efficiency analysis, we observea speedup of 8 (of the order of O(r3), when the refinement factor (r) is 2)compared to a traditional single uniform fine mesh-based simulation. Thetechnique is scalable and performs better when the problem size increases.Two applications related to HPM breakdown have been explored usingour in-house 2D simulator, one associated with the protection of electroniccomponents and the second on HPM swtching. Breakdown thresholds, thefield strength and the initial plasma density that determines breakdowntime for such applications are reported. The dependence of cutoff time oninitial plasma as well as strength of microwave E-field are investigated. Thetransmission and rejection capability of plasma for certain frequencies areinvestigated. Additionally, effect of gas heating on the HPM breakdowninduced plasma and the cutoff time is studied for switching and limiter action.We propose a completely new machine learning based data driven approachfor investigation of microwave-plasma interaction. Complete deep learning(DL) based pipeline to train, validate and evaluate the model has beendiscussed in this thesis. A convolutional neural network (CNN)-based deeplearning model, inspired from UNet with series of encoder and decoder unitswith skip connections, for the simulation of microwave-plasma interactionhas been discussed. The microwave propagation characteristics in complexplasma medium pertaining to transmission, absorption and reflectionprimarily depends on the ratio of electromagnetic (EM) wave frequency andelectron plasma frequency, and the plasma density profile. The scattering ofa plane EM wave with fixed frequency (1 GHz) and amplitude incident ona plasma medium with different Gaussian density profiles (in the range of1 � 1017 ? 1 � 1022m?3) have been considered. The training data associatedwith microwave-plasma interaction has been generated using 2D-FDTD(Finite Difference Time Domain) based simulations. The trained deep learningmodel is then used to reproduce the scattered electric field values for the1GHz incident microwave on different plasma profiles with error margin ofless than 2%. We compare the results of the network, using various metricslike SSIM index, average percent error and mean square error, with thephysical data obtained from well-established FDTD based EM solvers. Theproposed deep learning technique is significantly fast as compared to theexisting computational techniques, and can be used as a new, prospectiveand alternative computational approach for investigating microwave-plasmainteraction in a real time scenario.Most of the plasma applications and research in the area of low-temperatureplasmas (LTPs) is based on accurate estimation of plasma density and plasmatemperature. The conventional methods for electron density measurementshave major disadvantages of operational range (not very wide), cumbersomeinstrumentation, and complicated data analysis procedures. To address suchpractical concerns, the thesis further proposes a novel machine learning(ML) assisted microwave-plasma interaction based strategy which is capableenough to determine the electron density profile within the plasma. Theelectric field pattern due to microwave scattering is measured to estimate thedensity profile. The proof of concept is tested for a simulated training data setcomprising a low-temperature, unmagnetized, collisional plasma. Differenttypes of Gaussian-shaped density profiles, in the range 1016 ? 1019m?3,addressing a range of experimental configurations have been considered inour study. The results obtained show promising performance in estimatingthe 2D radial profile of the density for the given linear plasma device.The performance of the proposed deep learning based approach has beenevaluated using three metrics- SSIM, RMSLE and MAPE. The favourableperformance affirms the potential of the proposed ML based approach inplasma diagnostics and in future to replace existing plasma diagnostics.In conclusion, the thesis presents new approaches for investigation ofmicrowave-plasma interaction and HPM breakdown, which are significantlyefficient compared to existing simulation techniques. To the best of ourknowledge, this is the first effort towards exploring a data-driven DL basedapproach for the simulation of complex microwave plasma interaction. Thesimulations presented in the thesis provide a better understanding of bothionizing and non-ionizing applications of microwave-plasma interaction.They contribute to the study of complex plasma dynamics associated withhigh-frequency HPM breakdown-induced plasma, with potential applicationssuch as switching/limiters, and plasma diagnostics.Item Open Access Desertification characterization using predictive soil modelling and pattern recognition(Dhirubhai Ambani Institute of Information and Communication Technology, 2023) Dave, Viral A.; Ghosh, RanenduThis thesis presents a hierarchical methodology for land degradation mapping,land use land cover classification, degradation process identification and map-ping using multispectral LISS-3 images. The study aims to demonstrate the im-portance of remote-sensing images for various applications, both social and en-vironmental. The study compares the results of different algorithms for differentterrains, demonstrating that Simple Linear Iterative Clustering (SLIC) segmenta-tion with the random forest(RF) method outperforms CNN and pixel-based Sup-port Vector Machine (SVM) with an accuracy of 85% for level 1 land cover clas-sification. Vegetation degradation in forest areas is assessed in central parts ofGujarat, India, and land degradation in agricultural areas due to soil salinity isstudied, particularly in southeastern parts of Gujarat, India. ML algorithms likesupport vector machine(SVM) and RF was applied to different features to identifythe degradation process. Temporal data were used to find the severity of deserti-fication using the change in degraded areas.Further, it discusses soil degradation causing desertification and severely re-ducing potential soil productivity. The study uses machine learning algorithmsand an ANN-based model to predict soil properties like EC, pH, and OC, whichare important indicators of soil degradation. Environmental parameters are takenas covariates in prediction models, including vegetation indices, terrain indices,soil parameters, spatial attributes, and meteorological parameters of the study re-gion. Field soil sampling data of the study region obtained from Soil Health Card(SHC) for the year 2014 is incorporated in training the model. The SHC data isdivided into different ratios for training and testing the model. The SCORPANmodel is considered the base approach for the development of the ANN-basedprediction model. Moreover, the thesis also discusses the mapping of vulnera-ble areas to desertification. The study combines remote sensing and geographicinformation system (GIS) to map sensitive areas. Two different approaches wereused for vulnerability assessment: Mediterranean Desertification and Land Use(MEDALUS) approach and the fuzzy logic (FL) method. Soil, climate, land uti-lization, geography, and vegetation contribute to the land degradation of anyarea. However, man�s intervention leads to significant changes in the environ-ment, making socio-economic factors a considerable input to assess desertificationvulnerability. Indices related to these factors are generated, and both methods areused to find the severity level of the desertification vulnerability in the Panchma-hal district.Lastly, the role of climate in the process of desertification is discussed. Thestudy uses the aridity index (AI), which incorporates most of the weather datalike temperature, rainfall, humidity, wind, and solar radiation, to identify the de-sertification hot-spot using AI over the Gujarat state. The study uses weatherdata from more than 18 locations all over Gujarat for the past 20 years to calcu-late AI, and the FAO Penman-Monteith method was used to calculate PET. Thestudy generates an annual AI map for the whole of Gujarat using these valuesand compares it with a globally published AI map. It also compares the changein climate with the change in vegetation over the years using the vegetation in-dex for Gujarat. In summary, this thesis provides a comprehensive approach toland degradation mapping using degradation process identification, soil predic-tion, and climate variable using geospatial technology and machine learning. Thestudy demonstrates the importance of remote sensing images in various applica-tions, including social and environmental. The study employs different machinelearning algorithms and approaches to achieve high accuracy and identify vul-nerable areas to desertification. The study also highlights the importance of soilproperties and climate in the process of desertification.Item Open Access Water Footprint in the Context of Urban Water Management: Challenges and Opportunities(Dhirubhai Ambani Institute of Information and Communication Technology, 2023) Banerjee, Alik; Parikh, Alka; Tiwari, MukeshWater, a crucial resource in preserving the ecology in good shape, has becomescarce. Water footprint (WF) measure has been proposed in the literature to understandthis prevailing water crisis. The WF, which consists of green, blue, andgrey, can be defined as the green water footprint (WFgreen) that shows how muchwater is used by forests and non-irrigated agriculture; the blue water footprint(WFblue), shows the amount of water used by irrigated agriculture, industry, andresidences, and grey water footprint (WFgrey) shows how much water would berequired to neutralize the pollution in the water and bring it back to the acceptabledischarge water quality. This study conducted a comprehensive WF calculationin Purulia, Dhanbad, and Ranchi municipalities of West Bengal and Jharkhand,India. The primary reasons for choosing these municipalities were that they arewater-scarce and have an inadequate municipal water supply system.The researcher used published data to estimate WF. The results show WFgreenvalues depict that Purulia reports the highest mean values (182.6 to 296.3 (M3*103)per square kilometer (sq km)), followed by Dhanbad (170.3 to 241.2 (M3*103) persq km), and then Ranchi (131 to 219.2 (M3*103) per sq km) for four consecutiveyears (2016-19). These figures imply that Purulia overuses its water resourcesin agriculture, and hence its high WF green needs to be corrected by increasingwater productivity. Dhanbad�s high WF is because of the water consumption byits forests. The high WF is not of concern given that the forests help hold up thesoil and water. Ranchi�s WF is low because it has less land under forests andagriculture.Moreover, WFblue values of 2019 illustrate that Ranchi reports the highest (108M3 per capita), followed by Purulia (81.5M3 per capita), and Dhanbad reports theleast (68.8M3 per capita). The primary factor for getting such results is high runofffollowed by evaporation, and then the municipality supplies water. Therefore,Steps should be taken to retain the rainwater in some form in the soil and manmadechannels.In addition, this study examines the per capita per-day water availability among272 sample households of different income classes to understand the ground-levelsituation. The result reports that slum dwellers are the worst sufferers since theydo not get even the bare minimum amount of water � 70 lpcd, while affluent peopleliving in apartments or bunglows suffer no shortage. The study finds that thisinequality prevails because the primary water source is groundwater, accessibilityto which depends on wealth ownership. As the residences change from poorto non-poor, people depend less on centralized water supply and more on tubewells/bore wells. This is because the water supplied through the municipality isnot enough. Also, the correlation between sources of water and seasonal dearthshows significantly less value, which signifies that seasonal dearth does not relateto which water sources households are fetching the water from. Water quality isterrible for all income classes, but the rich can purify it through R-O.Furthermore, the study also found that in the case of municipality water balance,all three municipalities are going through a deficit water balance. For Purulia,it is 14 (M3*103) per day; for Dhanbad, it is 490.4 (M3*103) per day; andfor Ranchi, it is 439.6 (M3*103) per day, respectively. This means that water withdrawalis far more than the recharge rate. Water availability is expected to be evenmore compromised as we move forward.In such a situation, check dams, ponds, wells, reservoirs, etc., seem to be helpingin water conservation. In addition, water recycling, as tried out by Surat MunicipalCorporation, can also reduce WF. Based on these practical solutions, in theend, some policy recommendations are proposed for water conservation.Item Open Access Feature for Live and Spoofed Speech Detection(Dhirubhai Ambani Institute of Information and Communication Technology, 2023) Gupta, Priyanka; Patil, Hemant A.The authorization to access specific information is given by a biometric system.Biometric systems are used for security purposes in a way that they prevent unauthorized access to important information or data (information privacy). The accessgranted by the biometric is done by capturing traits of humans, which make allhuman beings unique w.r.t. that particular trait. This thesis focuses on voicebased biometric systems, also known as Automatic Speaker Verification (ASV)systems, given that speech is the most natural and powerful form of communication used by humans to communicate with the outside world. It is the most intuitive, simple, and easy-to-produce characteristic. Since ASV systems have beenused for applications, such as in banking transactions and access to buildings associated with classified information, only authorized legitimate or genuine usersare granted access.ASV systems suffer from vulnerabilities to attacks and can be compromisedat various stages. The attacks may be categorized as direct and indirect attacks,depending on the extent of the attacker�s accessibility to the ASV framework. Besides, due to the recent commercial success of several Intelligent Personal Assistants (IPAs), also known as voice assistants, such as Speech Interpretation andRecognition Interface (SIRI), Amazon Alexa, Google Home, and so on, manyvoice-enabled devices in Internet of Things (IoT) have been commonly prone tospoofing attacks. To that effect, there is active research in the direction of designing countermeasure systems for ASV systems, particularly for spoofing attacks,namely, Speech Synthesis (SS), Voice Conversion (VC), and replay.This thesis is a humble attempt to alleviate some of the research gaps in designing features for countermeasure systems. In particular, this thesis proposesQuadrature Energy Separation Algorithm (QESA) in the light of incorporating thequadrature-phase component with the in-phase component of the signal. To thateffect, an existing feature set for replay Spoofed Speech Detection (SSD), namely,CFCCIF-ESA is extended to the CFCCIF-QESA feature set for enhanced performance of the countermeasure system. The performance of the proposed CFCCIFQESA feature set is evaluated on various datasets for various spoofing attacksgiven in the literature. Furthermore, the existing Linear Frequency Residual Cepstral Coefficients (LFRCC) feature set is optimized w.r.t. to its Linear Prediction(LP) order for the replay SSD task. In particular, it is found that the LP orderneeded for a good prediction of speech is not the same as that needed for thereplay SSD task. The resulting optimized LFRCC feature set is evaluated on theASVSpoof 2019 PA dataset. In addition to this, another feature, known as the uncertainty vector (u-vector), is developed from the Heisenberg�s uncertainty principle in the signal processing framework. The proposed u-vector is evaluated usingthe ASVSpoof 2017 dataset for replay attacks.Furthermore, in the direction to make countermeasure systems independent ofthe type of spoofing attack, features have been proposed for the Voice LivenessDetection (VLD) task. VLD is performed by the detection of pop noise which is thediscriminating acoustic cue present in live speech, produced due to the breathingeffect captured by the microphone when the speaker�s mouth is close to the microphone. The work on VLD in this thesis is based on two key hypotheses, namely,Parseval�s energy equivalence for STFT, CWT, and analytic CWT, whereas the second hypothesis is that the energy of pop noise decreases with the distance of a microphone from the speaker that is used to capture genuine speech. The proposedfeatures for VLD in this thesis are wavelet-based, wherein three wavelets are used,namely, Bump, Morlet, and Morse wavelet, where Morse wavelet is presented as asuperfamily of analytic wavelets, called as Generalized Morse Wavelets (GMWs).Detailed experimental analysis such as speaker-microphone proximity, the effectof phoneme type, and the effect of frequency range is studied.Apart from this, the security of speech data is also taken into account and thisthesis proposes an improved Voice Privacy (VP) system, which is based on Linear Prediction (LP) of speech. Furthermore, the VP system is studied along withthe attacker�s perspective using the target selection approach, and particularly,target selection w.r.t. twins is studied, wherein the most vulnerable twin-pair(i.e., target) is selected. Lastly, some of the proposed feature sets in this thesis arealso evaluated for tasks related to other Assistive Speech Technologies (AST) applications, such as the classification of healthy vs. pathological infant cries, anddysarthric severity-level classification.Item Open Access Design of Quasi-periodic and Aperiodic Array Lattices to Improve Array Antenna Performance(Dhirubhai Ambani Institute of Information and Communication Technology, 2023) Mevada, Pratik; Gupta, SanjeevThe thesis addresses one of the possible solutions to the grating lobe occurrence in the beam steerable periodic arrays for large angular beam scans. The controlled aperiodicity has been introduced to the periodic array to achieve the same and the object is set to design the beam steerable array antenna offering improvement in the peak SLL performance with beam scan and reducing the number of array elements. The designs of such aperiodic and quasi-periodic array antennas have been carried out using the innovative strip projection (SP) based method. The strip projection method uses the area of rotated higher dimensional lattice and projects it to lower dimensions to generate an aperiodic array. The designs of aperiodic linear and planar arrays have been carried out to achieve �30� conical beam scan range with peak SLL <-10dB over -90� to 90o angular range. The novelty of the proposed SP method is that the number of optimization variables is fixed and independent of the size of the aperiodic array. The reported techniques to generate the aperiodic arrays lack in this aspect. The proposed method facilitates a significant reduction of the design efforts, especially in the case of the larger beam-steerable arrays. The proposed method is relatively straightforward to implement compared to the reported algorithms.The performance of the aperiodic linear array antenna has been compared with the aperiodic arrays designed using evolutionary optimization algorithms, namely, genetic algorithm (GA), particle swarm optimization (PSO) and Jaya algorithms and it is found that the proposed design method is comparatively more efficient and faster. The aperiodic array lattice is also populated with X-band electromagnetically coupled patch antenna integrated with a phase shifter and simulated. The aperiodic patch array antenna has been fabricated and characterized in the anechoic chamber. The comparison of the measured and simulated results is presented. In measurement, a significant improvement of 5.72dB in peak SLL is achieved at �30� beam scan angle.The design of an aperiodic planar array antenna has been carried out for a 15 x 15? aperture size. The optimized array has 21.9% less elements than the conventional periodic rectangular lattice. The Pinwheel based aperiodic array lattice has also been designed for the same beam scan requirement and presented for comparison. It is observed that the peak SLL performance is maintained at <-11.63dB and <-12.70dB over the 0�-30� beam scan range by the proposed aperiodic array and Pinwheel based array, respectively. Moreover, both types of lattice have been populated with S-band cubic-shaped dielectric resonator (DR) antenna element and their simulations have been carried out using a 3D electromagnetic solver. For the quantification of the aperiodicity in the structure, position standard deviation (O) is also defined and computed.The projection concept is generalized and implemented to design a quasi-periodic beam steering array antenna by projecting the vertices of co-centric polyhedrons on the 2D aperture plane. The modelling and design of quasi-periodic array lattice are carried out by projecting the vertices of co-centric polyhedrons, namely dodecahedrons and icosahedrons, on the aperture plane. The angular orientation of the polyhedrons is optimized to achieve a 4.2dB peak SLL improvement for a �30� beam scan. The optimized array lattice is populated with cubic shaped DR based elements and integrated with a Voronoi based metallic fence and decoupling network (DCN) for mutual coupling improvement between the elements. The polyhedron projection based concept has been extended to design interferometric arrays for radio astronomy. The stereographic projection has been used for the projection of vertices of the rotated polyhedron and forms the aperiodic array, whose performance is subsequently evaluated for radio interferometric imaging. The necessary test framework for imaging of 1" sample image using the designed array lattice has been developed in Matlab and the array lattice has been optimized to achieve the maximum fidelity index (FI). The various cases of the aperiodic array with various combinations of polyhedrons have been evaluated and compared with the Giant Metrewave Radio Telescope (GMRT) array. The aperiodic array generated by the three co-centric polyhedrons, i.c., dodecahedron, octahedron and tetrahedron, is proved to have a better fidelity index (FI) over the various declinations (8). In addition, the projection based aperiodic array antenna has also been evaluated for minimum variance distortionless response (MVDR) type of beamformer, which is a widely used technique in various fields like communication, radar, acoustics, and sonar. Matlab codes are developed to implement DoA estimation using the MVDR technique and applied to the conventional periodic and proposed aperiodic linear arrays. It is shown that aperiodicity in the element position has eliminated the unwanted lobes in the detection range.Item Open Access Handcrafted Features for Anti-Spoofing(Dhirubhai Ambani Institute of Information and Communication Technology, 2023) Patil, Ankur T.; Patil, Hemant A.Amongst various biometrics, voice is the most natural and convenient way of the communication for human-machine interaction. To that effect, the use of AutomaticSpeaker Verification (ASV) for authentication is increasing in various sensitiveapplications, which create a chance for fraudulent attack as attackers canbreach the authentication by using various spoofing attacks. To alleviate this issue,we can either develop an ASV system, which is inherently protected fromthe spoofing attacks or develop a separate countermeasure (CM) system that canassist the ASV system in tandem against the spoofing attacks. The earlier approacheshave trade-off between performance of the ASV system and robustnessagainst spoofing attacks. Hence, it would be advantageous to implementthe separate Spoof Speech Detection (SSD) system, and hence majority researchattempts are focusing upon the later approach. To that effect, various internationalchallenge campaigns were organized during INTERSPEECH conferences,such as ASVSpoof 2015, ASVSpoof 2017, and ASVSpoof 2019, which providesstandard datasets, protocol, and evaluation metrics. This thesis focuses on developingthe handcrafted feature sets for CM systems against the spoofing attacks,namely, Speech Synthesis (SS), Voice Conversion (VC), and replay. These featuresets are either developed by applying the subband filtering on the speech signalsor derived from the spectrogram representations.In this thesis work, various subband filtering-based feature sets are developed,namely, Enhanced Teager Energy-Based Cepstral Coefficients (ETECC), Cross-Teager Energy Cepstral Coefficients (CTECC), and Energy Separation AlgorithmbasedInstantaneous Frequency estimation for Cochlear Cepstral Features (CFCCIFESA).These feature sets are either modification in Teager Energy Operator (TEO)-based representations or utilization of Energy Separation Algorithm (ESA) for InstantaneousFrequency (IF) estimation. The ETECC feature set is developed byaccurately estimating the energies in high frequency regions using compensationof the signal mass. In Teager Energy-Based Cepstral Coefficients (TECC), TEO isutilized to estimate the energy, which considers the approximation sin(?) ? ?,which is applicable for low frequencies. However, the discriminative information or the replay detection is prominently present in the mid and high frequency regions.Hence, ETECC feature set is proposed to obtain the efficient representationfor SSD task by accurately estimating the energies at high frequency regions. Furthermore,signal processing-based approach is presented for replay SSD in VoiceAssistants (VAs). It utilizes the Cross-Teager Energy Operator (CTEO) for extractingthe acoustic cues from replay speech. CTEO gives the interactions amongthe multi-channel signal by estimating the cross-Teager energies between signals.To that effect, it is necessary to efficiently represent the acoustic cues for replayspoofs and hence, maximum cross-Teager energies among the subband filteredmulti-channel signal is utilized for feature representation. Thus, the rationale behindoptimal channel selection is to find the most noisy (distorted) transmissionchannel. The cepstral features extracted using CTEO are referred as Cross-TeagerEnergy Cepstral Coefficients (CTECCmax). The experiments are performed usingRealistic Replay Attack Microphone Array Speech Corpus (ReMASC), which is speciallydesigned for the replay SSD in VAs. The proposed CTECCmax feature setperforms better than other state-of-the-art feature sets. The proposed CFCCIFESAfeature set combines the magnitude and phase (in the form of IFs) informationto develop the efficient feature representation for SS, VC, and replay spoofingattacks. The proposed CFCCIF-ESA utilizes ESA to accurately estimate themodulation patterns due to their relatively low computational complexity, hightime resolution, and instantaneously adapting nature. In previously proposedCochlear Filter Cepstral Coefficient Instantaneous Frequency (CFCCIF) featureset, IFs were estimated using Hilbert transform-based approach, whose time resolutionis relatively low (as it requires a segment of speech) as compared to theESA-based approach.Furthermore, Constant-Q Transform (CQT)-based feature representation andSpectral Root Cepstral Coefficients (SRCC) are developed using spectrogram representationsand effectively utilized for anti-spoofing. According to Heisenberg�suncertainty principle in signal processing framework, the CQT has variable spectrotemporalresolution, in particular, better frequency resolution for low frequencyregion and better temporal resolution for high frequency region. This property ofthe CQT representation is effectively utilized to identify the low frequency characteristicsof pop noise. Here, pop noise is attributed to the live speaker and hence, itis exploited for Voice Liveness Detection (VLD) task. SRCC feature set is derivedfrom the theory of homomorphic filtering, which obeys the generalized superpositiontheory. In spectral root homomorphic deconvolution system, convolutionallycombined vectors are mapped to another convolutionally combined vector space, where signal components are more easily separable by liftering operation.Logarithm operation in Mel Frequency Cepstral Coefficients (MFCC) extractionis replaced by power-law nonlinearity (i.e., (�)?) to derive SRCC feature set. Theproper choice of the ? depends upon the pole-zero arrangements in the transferfunction obtained from the speech signal and it helps to capture the system informationof the speech signal, with a minimum number of cepstral coefficients. Inthis thesis, optimum ?-value is chosen by estimating the energy concentration incepstral coefficients and by visualizing the spectrogram w.r.t. ?-value.To validate performance of our proposed feature sets, the experiments are performedusing various datasets, state-of-the-art feature sets, classifiers, and evaluationmetrics. The development and performance analysis of each proposedfeature set is provided in the corresponding chapters. Furthermore, other contributionsin the thesis, namely, feature normalization for anti-spoofing, analysis onDelay and Sum (DAS) vs. Minimum Variance Distortionless Response (MVDR)beamforming techniques for anti-spoofing in VAs, severity-level classification ofdysarthric speech, and classification for normal vs. pathological cries, are alsodiscussed. Thesis concludes with potential future research directions and openresearch problems.Item Metadata only QoS-aware Task Allocation and Scheduling in Cloud-Fog-Edge Architecture with Proactive Migration Strategy(Dhirubhai Ambani Institute of Information and Communication Technology, 2023) Joshi, Nikita; Srivastava, SanjayThe vision of the Internet of Things (loT) is made possible by advancements in sensor technologies, smart devices, wearable gadgets, and communication paradigms. IoT Service Providers (loSPs) provide loT services such as smart cities, virtual and augmented reality and pervasive healthcare. These services produce a significant amount of time-sensitive data to be processed. IoT devices can not process these data due to resource and energy constraints. Centralized cloud computing services can provide on-demand computationaland storage capabilities through the Internet. Users of cloud computing benefit from minimal startup costs for their IT needs, unlimited resources, and easy ac cess to them. Therefore, IoSPs send task requests to process these data to Cloud Service Providers (CSPs). However, the emerging trend of delay-sensitive appli cations demands low latency, privacy, and context awareness, which can barely be satisfied by applications processed at distant cloud centers as it takes a signif icant amount of time to send and receive a massive volume of data. Several newa distributed computing models have emerged that provide computing and storage services close to service consumers to satisfy the latency and context awareness needs of the applications such as Fog computing, Cloudlets and Micro data cen ters. They extend the cloud services to the network edge and give services to the end-users with low latency and decrease data traffic.Fog Service Providers (FSPs) are available in the market who have deployed computation servers that can act as fog devices. CSPs can offload the task request received from IoSP to FSP to decrease the service delay and pay to FSP from the payment received from IoSPs. The IoT tasks have Quality of Service (QoS) re quirements such as deadlines and priorities. In addition, IoT tasks are online in nature. As a result, task requirements are not known in advance. The allocation of resources to these tasks in a commercial three-tier architecture is a complex prob lem since QoS requirements of IoT applications and competition among loSPs and FSPs for the price of the resources must be considered. The perishable nature of cloud-fog resources makes allocation more challenging. Resources not allocated at a specific time cannot be reused later. Considering this property when deter mining the price of resources and allocations is essential.In such a competitive market where each service provider, such as loSPs, CSPs, and FSPs, wishes to maximize its profit, auction mechanisms are the best tools for finding prices of resources and allocating them in a manner that allows service providers to benefit from market demand and IoSPs to benefit from competition among providers. In our model, CSP acts as a broker between IoSPs and FSPs. FSPs have resources to sell with price demand, and IoSPs have task requirements with the cost they are willing to pay. FSPs want to sell at a maximum price, and IoSPs want to purchase resources at a minimum price. Moreover, due to the per ishable nature of the resources, if FSPs bid too high, resources will not be allo cated, resulting in their loss. The double auction can be used to find equilibrium in this situation. A multi-attribute double auction mechanism is employed in oura model to account for both QoS requirements of tasks and monetary competition between IoSPs and FSPs.We perform task allocation in batch mode and online mode. Batch mode al location is done at every fixed interval. It is assumed that task requests can be generated at any time. Depending on the criticality of the task, it can be executed in batch mode or online mode. In batch mode, eligible FSP is found for eacha task, and then (TASC) is used for price determination. After that, considering the perishable nature of the resources, the remaining resources are also allocated at a lower price with some constraints on IoSPs. For doing the auction for online tasks, there are no other competitors to compare and make efficient decisions. We perform a virtual auction at the beginningof the system to handle this situation, giving us the critical price for differentsupply-demand scenarios and QoS levels. This critical price decides whether an online task request should be accepted or rejected. Also, the dynamic nature of the Internet, the scarcity of resources in the cloud fog, and the variability in service rates in the cloud fog may delay the execution of IoT tasks after allocation, low ering the task value. Therefore, service delay between IoSP and allocated FSP is monitored after allocating the tasks. If, for any reason, it is observed that the task may not be completed before the deadline, then more resources are allocated, or the task is migrated to another FSP. We prove that our algorithm is a polynomial time algorithm of time complexity O(N2K2M) where N is the number of IoSPs, a M is the number of FSPs, and K is the number of tasks.The remote patient monitoring system is used as a case study to verify the pro posed QoS-aware Task Allocation and Scheduling (QoTAS). We consider the tasks performed in remote patient monitoring systems with their resource and QoS re quirements. The Internet behavior is simulated in Netsim to get network delay between FSP-IoSP and FSP-FSP. We perform experiments to compare the batch allocation algorithm (QoTAS-B) with Multi attribute-based double auction mech anism (MADA) [81] and batch with online allocation algorithm (QoTAS-BO) with Reverse auction-based online allocation (RAOA) [18]. Which shows proposed algorithm outperforms the existing work [81] and [18]. Also, results show that migration increases the task completion ratio.Item Open Access Heavy Metal Detection in Crops and Soil Clay Mineral Abundance Mapping using Hyperspectral Data(Dhirubhai Ambani Institute of Information and Communication Technology, 2023) Priya, Swati; Ghosh, Ranendu; Mandal, SrimantaPresence of heavy metal in crops is an indicator of environmental pollution. Theheavy metals found in the plant indicate that the specific metal exists in the terrestrialenvironment. These metals affect leaves� spectral characteristics and interferewith plants� biochemical features, such as chlorophyll concentration and photosynthesis.Accurate detection of heavy metals in plants is necessary for agriculturalmanagement and to preserve ecological balance. Field spectroscopy techniquesare used to measure the spectral changes triggered due to contaminationwith heavy metals. The advantage of these remote sensing approaches to assessheavy metal contamination is that they can frequently collect data across a widegeographic area.The study mainly focuses on detecting different levels of heavy metal pollutionfrom airborne hyperspectral data using reference data from in situ controlledpot experiments. We constructed a training set spectrum from a controlled experimenton cotton and tobacco for two important heavy metals, Lead (Pb) andCadmium (Cd). Cotton and tobacco crops were grown in pots after artificiallycontaminating the soil with four Pb and Cd heavy metal treatments. The hyperspectraland biochemical data generated spectra of heavy metal concentrations atdifferent crop growth stages. Standard reflectance spectra at different contaminationlevels do not show significant changes at different wavelengths due to thepresence of heavy metal. These spectra were further decomposed using wavelettransform at different levels to capture the subtle changes in spectra using thedetailed component of wavelets. The reconstructed detailed wavelet reflectanceat the third level of decomposition was found to be significant with heavy metalstress. The correlation analysis established that the wavelength range of 651-742nano meter (nm) in cotton was sensitive to Pb stress, and 631-802 nm was sensitiveto Cd stress in tobacco. The reconstructed detail reflectance at a particularwavelength was then further used as reference spectra with different heavy metallevels to map heavy metal pollution.The AVIRIS-NG data obtained for the study area was first classified to identifythe tobacco crop in the Anand region and the cotton crop in the Surendranagarregion using a combination of Autoencoder (AE) for feature extraction followedby an artificial neural network for classification. The training data obtained fromthe pot experiment were utilized to map Pb and Cd pollution from classified airbornehyperspectral data from Airborne Visible InfraRed Imaging Spectrometer- Next Generation (AVIRIS-NG) using a spectral matching algorithm known asDynamic Spectral Warping (DSW). The results confirm the efficiency of the developedalgorithm in estimating Cd content in tobacco and Pb content in cottoncrops. The model was validated by collecting the exact field points and heavymetal concentration, which shows a promising result for this algorithm.Diverse soil minerals may be easily identified through modern hyperspectraltechnology for remote sensing. The aerial hyperspectral sensor�s enhanced spatialand spectral resolution can identify the abundance of several clay minerals, suchas Kaolinite, Montmorillonite, and Illite. This study maps the clay mineral distributionin the Udaipur area of Rajasthan and the Ambaji region of Gujarat usinghyperspectral data acquired by the AVIRIS-NG sensor on an airborne platform.The representative soil sampling sites were selected from hyperspectral datausing the Spectral Feature Fitting (SFF) algorithm. X-ray Diffraction (XRD) analysiswas carried out to find different clay minerals in the samples. Then the regressionanalysis was carried out to find the relation between Absorption PeakDepth (APD) extracted from hyperspectral data corresponding to the actual locationof sampling sites and the corresponding clay percentage obtained from XRDanalysis. Regression analysis between absorption peak depth values estimatedfrom hyperspectral data at 2205 nm � 2214 nm spectral region of soil samplingsites and corresponding clay content value showed a significant relationship. Theregression line obtained for the known pixel is used to prepare the mineral abundance map over the study area. The study over the Udaipur region shows thedominance of montmorillonite clay minerals, and the Ambaji region showed anabundance of kaolinite.