Abstract

A hand-drawn sketch is a convenient way to search for an image or a video from a database where examples are unavailable or textual queries are too difficult to articulate. In this thesis, we have tried to propose solutions for some problems in sketch-based multimedia retrieval. In case of image search, the queries could be approximate binary outlines of the actual objects. In case of videos, we consider the case where the user can specify the motion trajectory using a sketch, which is provided as a query. However there are multiple problems associated with this paradigm. Firstly, different users sketch the same query differently according to their own perception of reality. Secondly, sketches are sparse and abstract representations of images and the two modalities can not be compared directly. Thirdly, compared to images, datasets of sketches are rare. It is very difficult, if not impossible to train a system with sketches of every possible category. The features should be robust enough to retrieve classes that were not a part of training. In this thesis, the work can be broadly divided into three parts. First, we develop a motion-trajectory based video retrieval strategy and propose a representation for sketches that aims to reduce the perceptual variability among different users. We also propose a novel retrieval strategy, which combines multiple feature representations for a final result using a cumulative scoring mechanism. In order to tackle the problem of multiple modalities, we propose a sketch-based image retrieval strategy by mapping the two modalities into a lower dimensional sub-space where they are maximally correlated. We use Cluster Canonical Correlation Analysis (c-CCA), a modified version of standard CCA, for the mapping. Finally, we investigate the use of semantic features derived from a Convolutional Neural Network, and extend the idea of sketch-based image retrieval to the task of zero-shot learning or unknown class retrieval. We define an objective function for the network such that, while training, a close miss is penalized less than a distant miss. Our training encodes semantic similarity among the different classes. We perform experiments to evaluate our algorithms on well known datasets and our results show that our features perform reasonably well in challenging scenarios.

Abstract

Software Quality has been an important economic factor for every successful Software Product. It is quite challenging for large-scale software products, especially integrated software products to maintain its quality after every version release. Integrated Software Product is a software that combines two or more sub-products to achieve a similar business goal. These sub-products are interconnected to each other and share common packages of programs to form one application. Unlike non-integrated software products, integrated software products are complex in design. They require detailed exploration on the spread of a defect across entire integrated product suite along with causal analysis so as to improve overall product quality. Either due to security constraints or due to complicated workflow, quality engineers may not be able to access and analyze the source code of all sub-products together in a large scale integrated software product. Thus investigating defect dependency in large scale integrated software product is difficult. In this thesis, we present our efforts towards exploring various ways to estimate defect widespread or defect dependency of a defect in a large scale integrated software product. Current approaches are mostly non-generic and are developed for a specific product design with domain constraints and with predefined conditions. Very few approaches were found to be adaptable for large software products. However, they were neither evaluated nor validated using a real-time defect dataset of a large scale integrated software product. This motivated us to present a few approaches to estimate the defect widespread or defect dependency of a defect in a large scale integrated software product. Our approaches are heuristic by design, this is to achieve a generalized way to study the defect dependency in a software product of any domain and size. These approaches help product owners of an integrated software product to prioritize defects based on its widespread across the large software product and also help quality teams to evaluate the defect dependency so as to address dependent defects based on artifacts accessible to them. As part of our initial research, we examined Defect Dependency calculation using the concept of Generalized Dependency Degree method which is measure a of studying the dependency of an element over another. This method was primarily inspired by dependency degree approach from Rough Set Theory. We formulated this method into an approach by representing all recorded defects per sub product in a large scale integrated software product as onevset. We calculated Generalized Dependency Degree between defect sets of each sub-product so as to estimate the defect dependency over another. We implemented this approach on a real-time industrial defect dataset of a large scale integrated Human vi vii Resource Management product. Also captured significant results across various version releases. This is a simple and a generic approach to implement on a large software products. However, the level of abstraction on defect dependency calculation for a large software was limited to defects recorded at sub-product only. This prompted us to explore new ways to study defect dependency at a deeper level of abstraction in a large software product. Correlation analysis in statistics is one of the efficient way to understand the independence or dependency of elements in a given relation. We explored various ways on implementing correlation analysis on a defect dataset and formulated an approach based on correlation to study defect dependency between defects recorded in a large software product. For a better level of generalization, we have grouped defects recorded per feature in a large software product and performed defect correlation analysis to understand the defect dependency among defect sets of available software features. We evaluated this approach against Open Source (NetBeans IDE) defect dataset and captured noteworthy observations. We were able to record list of most defective features which are unstable when deployed together in a large software product.These results help product managers to prioritize defects recorded in defective features and focus towards building a stable software. However, product managers still may not be able to isolate the exact cause-effect of a single defect over entire large software product i.e. with the above proposed methods to study defect dependency, it is difficult to estimate the dependency of a single defect over entire large software product. This stimulated us towards exploring defect dependency at the lowest level of abstraction in a large software product. We composed another approach with a model framework as basis to evaluate defect dependency so as to fill this gap. We have implemented our approach using an i* modelling framework on a few real time software projects and have recorded notable observations. The approaches composed as part of th

Abstract

With the rapid growth of camera-based mobile devices, applications that answer questions such as, “What does this sign say?" are becoming increasingly popular. This is related to the problem of optical character recognition (OCR) where the task is to recognize text occurring in images. The OCR problem has a long history in the computer vision community. However, the success of OCR systems is largely restricted to text from scanned documents. Scene text, such as text occurring in images captured with a mobile device, exhibits a large variability in appearance. Recognizing scene text has been challenging, even for the state-of-the-art OCR methods. Many scene understanding methods recognize objects and regions like roads, trees, sky in the image successfully, but tend to ignore the text on the sign board. Towards filling this gap, we devise robust techniques for scene text recognition and retrieval in this thesis. This thesis presents three approaches to address scene text recognition problems. First, we propose a robust text segmentation (binarization) technique, and use it to improve the recognition performance. We pose the binarization problem as a pixel labeling problem and define a corresponding novel energy function which is minimized to obtain a binary segmentation image. This method makes it possible to use standard OCR systems for recognizing scene text. Second, we present an energy minimization framework that exploits both bottom-up and top-down cues for recognizing words extracted from street images. The bottom-up cues are derived from detections of individual text characters in an image. We build a conditional random field model on these detections to jointly model the strength of the detections and the interactions between them. These interactions are top-down cues obtained from a lexicon-based prior, i.e., language statistics. The optimal word represented by the text image is obtained by minimizing the energy function corresponding to the random field model. The proposed method significantly improves the scene text recognition performance. Thirdly, we present a holistic word recognition framework, which leverages scene text image and synthetic images generated from lexicon words. We then recognize the text in an image by matching the scene and synthetic image featureswith our novel weighted dynamic time warping approach. This approach does not require any language statistics or language specific character-level annotations. Finally, we address the problem of image retrieval using textual cues, and demonstrate large-scale text-to-image retrieval. Given the recent developments in understanding text in images, an appealing approach to address this problem is to localize and recognize the text, and then query the database, as in a text retrieval problem. We show that this approach, despite being based on state-of-the art methods, is insufficient, and propose an approach without relaying on an exact localization and recognition pipeline. We take a query-driven search approach, where we find approximate locations of characters in the text query, and then impose spatial constraints to generate a ranked list of images in the database. We evaluate our proposed methods extensively on a number of scene text benchmark datasets, namely, street view text, ICDAR 2003, 2011 and 2013, and a new dataset IIIT 5K-word, we introduced, and show better performance than all the comparable methods. The retrieval performance is evaluated on public scene text datasets as well as three large datasets, namely, IIIT scene text retrieval, Sports-10K and TV series-1M, we introduced.

Abstract

Digitization can provide a means of preserving the content of the materials by creating an accessible facsimile of the object in order to put less strain on already fragile originals such as out of print books. The document analysis community formed to address this by digitizing the content thus, making it easily shareable over Internet, making it searchable and, enabling language translation on them. In this thesis, we have tried to see optical character recognition as a mapping problem. We proposed extensions to two method and reduced its limitations. We proposed an application for sequence to sequence learning architecture which removed two limitations of previous state of art method based of connectionist temporal classification output layer. This method also gives representations which can be used for efficient retrieval. We also proposed an extension to profile features which enabled us to use same idea but by learning features from data. In first work, we propose an application of sequence to sequence learning approach for printed text Optical Character Recognition. In contrast to present day existing state-of-art OCR solution which uses Connectionist Temporal Classification (CTC) output layer our approach makes minimalistic assumptions on the structure and length of the sequence. We use a two step encoder-decoder approach – (a) A recurrent encoder reads a variable length printed text word image and encodes it to a fixed dimensional embedding. (b) This fixed dimensional embedding is subsequently comprehended by decoder structure which converts it into a variable length text output. The learnt deep word image embedding from encoder can be used for printed text based retrieval systems. The expressive fixed dimensional embedding for any variable length input expedites the task of retrieval and makes it more efficient which is not possible with other recurrent neural network architectures. Thus single model can do predictions and features learnt with supervision can be used for efficient retrieval. In the second work, we investigate the possibility of learning an appropriate set of features for designing OCR for a specific language. We learn the language specific features from the data with no supervision. In this work, we learn features using a unsupervised feature learning and use it with the RNN based recognition solution. We learn features using a stacked Restricted Boltzman Machines (RBM). These features can be interpreted as deep extension of projection profiles. These features can be used as a plug and play features to get improvements where profile feature are used. We validate these features on five different languages. In addition, these novel features also resulted in better convergence rate of the RNNs.

Abstract

After Bell’s celebrated work in 1964, demonstrating the incompatibility of Quantum Mechanics with the notion of local realism, there have been numerous conceptual and technical developments for studying nonlocality in quantum systems. The Bell inequality and Bell-type inequalities, set an upper bound on the correlations between measurement statistics of many particle systems that cannot be explained by local hidden variable models. Quantum entanglement has proven to be a key resource for various information theoretic protocols such as dense coding, teleportation, etc. Their numerous applications makes it consequential to pinpoint their behaviour. In the first chapter, a cursory delineation of the historical events in the development of Quantum theory, is followed by a concise introduction to the theory and mathematical tools needed for this thesis. The landmark papers of EPR and John Bell are discussed, so as to provide an understanding of the significance of Bell’s theorem and thus the Bell’s inequalities. Quantum Entanglement is discussed, and in particular, measures that are used later to quantify tripartite correlations are described in greater detail. The second chapter discusses Bell-CHSH inequality for two qubit systems and Bell-type inequalities for multi-qubit systems; and the theory independent derivation of Bell-CHSH inequality, known as the Device Independent scenario, where the characteristics of the probability distributions shared between two or more parties are sufficient to define different kinds of nonlocal correlations, without recourse to any underlying theory. The third chapter briefly goes over the disjoint sets of genuinely entangled three qubit pure states that constitute the GHZ and W classes of states. It also introduces a new class of states called the Maximally Bell-Inequality Violating (MBV) class of states. This is followed by the description of the analytical calculations for the maximum bipartite Bell Inequality violations of the three classes. For each of the three classes a complementary relation is then established between the maximum bipartite Bell Inequality Violation and tripartite correlation measures - Generalized Geometric Measure (GGM), Tangle and Discord Monogamy Score (DMS). Numerical evidence for this is also provided. The fourth chapter discusses first the convex roof extension of the complementary relations described above and then it goes on to discuss the results of the obvious next step, comparing tripartite nonlocality with tripartite correlation measures. The analysis is put forth as a future problem which possibly has quite a few fascinating insights in store.

Abstract

Scientific method of finding a new law pertains to three sequential steps. First we guess a law, second we compute its consequences and third we compare the consequences with experiments. On top of these lies the pillar of science, if these consequences disagree with observation the guessed law is wrong. Science can never assert that any theory, or law explains nature. It can only tell which ones cannot. The twentieth century saw Humankind asking one of the most fundamental question regarding the nature of reality and determinism. EPR in 1935 questioned the completeness of Quantum Mechanics (a proba- bilistic theory as a fundamental theory). Bell’s framework developed to answer the question, introduced first notions of theory independent tests. While we cannot say whether there exists a deterministic ex- planation to every prediction of QM, but we can reject a subclass of deterministic explanation, local deterministic explanation. This experimentally verified departure of predictions of QM or any theory from any local deterministic explanation is referred to as non-locality. In first chapter origins and mathematical formalism of quantum mechanics is presented. Followed by presentation of EPR argument, Bell’s theorem, implications and applications in the second chapter. Fol- lowing chapters contain the main contributions from the author. In third chapter, we study the consequences of both answers set in a device (theory) independent frame- work, based only on observed statistics. We start with taking up post-selection as an assumption and model the same using independent devices governed by Boolean functions. We establish analogy be- tween the post selection functions and the general probabilistic games in a two party binary input-output scenario. As an observation, we categorize all possible post-selection functions based on the effect on a uniform input probability distribution. We find that post-selection can transform simple no signaling probability distributions to signaling. Similarly, solving NP complete problems is easy independent of classical or quantum computation (in particular we prove that Post RP = NP). Finally, we demonstrate an instance of the violation of the pigeon hole principle independent of underlying theory. As result of our theory independent modeling we conclude that post-selection as an assumption adds power to the underlying theory. In particular, quantum mechanics benefits more with the post-selection assumption, only because it admits a more general set of allowed probabilities as compared to the local hidden vari- able model. Without the assumption we associate a device independent efficiency factor to quantify the cost of post selection. Our study shows that in the real world post-selection is not efficient enough to be of any advantage. But from an adversarial perspective it is still of significance. As an application, we obtain robust bounds on faking the bell violation in terms of minimum efficiency required using post selection. In fourth chapter we study a semi device-independent protocol wherein the knowledge of success prob- ability of the associated dimension witness and the observed efficiency of the detectors (η a vg) are suf- ficient to determine the security. We consider practically possible, individual attacks with and without eavesdropper having access to quantum memory. We find critical detection efficiencies required for security in both cases. In fifth chapter we introduce generalizations to PR and (2 → 1) RAC and study their inter-convertibility. We introduce generalizations based on the number of inputs provided to Alice, B n -BOX and (n → 1) RAC. We show that a B n -BOX is equivalent to a no-signaling (n → 1) RACBOX (RB). Further we introduce a signaling (n → 1) RB which cannot simulate a B n -BOX. Finally to quantify the same we provide a resource inequality between (n → 1) RB and B n -BOX, and show that it is saturated. As an application we prove that one requires atleast (n − 1) PRs supplemented with a bit of communication to win a (n → 1) RAC. We further introduce generalizations based on the dimension of inputs provided to Alice and the mes- sage she sends, B n d (+)-BOX, B n d (−)-BOX and (n → 1, d) RAC (d > 2). We show that no-signaling condition is not enough to enforce strict equivalence in the case of d > 2. We introduce classes of no-signaling (n → 1, d) RB, one which can simulate B n d (+)-BOX, second which can simulate B n d (−)- BOX and third which cannot simulate either. Finally to quantify the same we provide a resource in- equality between (n → 1, d) RB and B n d (+)-BOX, and show that it is saturated. In sixth chapter we introduce the measurement-device-independent randomness certification task where one has trusted quantum state preparation device but the mesurement devices are completely unspec- ified. Interestingly we show that there exist entangled states, w

Abstract

Background Determining gene and transcript expression patterns is crucial for understanding cellular and molecular biology concepts. Currently, a large amount of gene expression data is available, and many tools and databases are developed to predict or extract gene expression patterns. But, these resources have many limitations. So, there is a need to develop a new resource that can help in establishing expression patterns. Gene expression patterns can also assist in the identification of biomarkers, which in turn is an important step for identification of better diagnostics and prognostics. Microarray experiments have been extensively used to identify genes associated with various physiological conditions. But, there is a scope for making a better use of such available data. For example, the gene-level expression data can be used to identify transcript isoforms associated with a disease or a physiological condition. Establishing expression of the transcript isoforms in various physiological and disease conditions is important, as most of the human genes undergo alternative splicing and produce an enormous number of transcript isoforms, and these isoforms are also associated with diseases. In fact, establishing alternatively spliced isoforms in different tissues and conditions is essential for understanding many fundamental molecular events, such as mechanisms of alternative splicing and its regulation. Moreover, the mechanisms of alternative splicing are far from being understood clearly. Non-obstructive azoospermia is a common male infertility disorder that can be considered as a case for reanalyzing the gene expression data corresponding to it. Microarray studies have enabled identification of genes differentially regulated in this disease. But, no meta-analysis attempts have been made to identify genes strongly associated with the disease, which can be further used as potential biomarkers. It should be possible to identify transcript isoforms as potential biomarkers for non-obstructive azoospermia. Objectives The current study aimed at: Compiling existing gene expression data for one tissue, the mammalian testis Comparing the utilities of the existing gene expression databases with the newly developed testis-specific database (developed using the compiled data) Establishing expression profiles at the transcript isoform level via a new software and validating the results Using the new software and RNA-sequencing experiments to establish the transcriptome for non-obstructive azoospermia and identifying potential biomarkers at transcript isoform level and Analyzing the data to explore the factors associated with alternative splicing, in the context of mammalian testis Results The mass scale gene expression data, corresponding to various testicular physiological conditions and normal physiological condition for other tissues, were curated, and gene expression datasets were accumulated/derived. These datasets were later used to derive genes associated with different physiological conditions. The database developed using such data was found to be better and more reliable than most other gene expression databases. This could be due to both wealth of data and the new features used - including a meta-analysis method employed in the development of the new database. A web server, TIPMaP, was developed to predict expression profiles of transcript-isoforms using the original/existing microarray gene expression data. These data were reanalyzed using good quality probes and the transcript expression profiles were determined. The expression pattern predicted by the tool was validated for a few transcripts using RT-PCR and RNA-sequencing experiments. There was a high percentage agreement between the experimental observations and the web server results, and the extent of such agreement was dependent on the consistency in the results across the microarray hybridizations. Thus, the tool aids in establishing expression patterns of specific transcripts and provides a quantitative means of reproducibility of these patterns. Transcripts differentially regulated/expressed in non-obstructive azoospermia were identified by re-analyzing an already published gene expression dataset using TIPMaP and by RNA-sequencing. The results helped to identify the potential biomarkers for non-obstructive azoospermia. An in-depth analysis of alternatively spliced transcript isoforms helped to compile the type of events that occur in the mammalian testis-associated genes. Additionally, a set of factors that might promote alternative splicing in the testis tissue were identified. These factors include a few testis-specific splice factors, over/under-represented splicing regulatory motifs, and RNA secondary structural stabilities at the splice site junctions. Conclusions The majority of the existing gene expression data was compiled for the mammalian testis tissue via a laborious biocuration process. The new gene expressi

Abstract

The past decade has witnessed a significant improvement in neuroimaging technologies and analysis methods in order to observe and measure brain activity. The advances in the understanding of how the brain is organized is the key to progress in fighting neurological disorders. The efforts to find ways to detect and diagnose these impairments depend largely on the data that we are able to obtain about cerebral hemodynamics, fluctuations, structural changes and the associated cortical activities in the brain. The aim of this thesis is to develop innovative methodologies for examining cortical excitability regulation particularly for stroke patients during transcranial direct current stimulation (tDCS) with the help of Near Infrared Spectroscopy (NIRS) and Electroencephalography (EEG) combined brain imaging. EEG is an electrophysiological method that measures the electrical neural activity directly while NIRS is a method that measures the brain activity indirectly through hemodynamic responses associated with neuron behavior. Transcranial direct current stimulation (tDCS) is a non-invasive technique of modulating brain activity that delivers low intensity direct current to cerebral cortex. It is widely used for stimulating or inhibiting spontaneous neuronal activity. The alterations in cerebrovascular hemodynamics caused by tDCS can be studied through functional NIRS. A low cost tDCS device was developed and integrated with the custom built NIRS data acquisition device to study the effects of electrical stimulation on brain in patients for prediction of neurological disorders such as ischemic stroke. The NIRS – tDCS device was taken for clinical validation at Institute of Neurosciences, Kolkata and it proved to be a good predictor of cerebral vascular status. It can be used for possible identification of acute stroke. The tDCS –NIRS protocol was then further extended to tDCS-NIRS-EEG protocol. It was further investigated that the application of tDCS on cerebral site causes systemic interference in the superficial layers of the head. Due to this, there were some fluctuations found in NIRS measurements. The anterior temporal artery tap technique is discussed in the thesis to remove these fluctuations, which may also be able to classify carotid stenosis, external carotid artery stenosis, and internal carotid artery stenosis patients using the laterality in the hemodynamic response evoked by anodal tDCS both at the brain as well as at the superficial layers. These findings may have important implications for both prognosis and rehabilitation of patients with intracranial stenosis. The EEG and NIRS have different spatial and temporal resolutions, the online tracking of the relation between EEG and NIRS data acquired simultaneously during tDCS has not been investigated earlier. This could provide a sensitive means to monitor the tDCS neuro-modulatory effect. A Kalman filter based approach for online parameter estimation of an autoregressive model to track the relation between tDCS induced cortical neural activity leading to changes in EEG power spectrum and oxy Hemoglobin signals. This new online NIRS EEG tracking method may allow quantitative assessment of the existence of a coupling relationship between electrophysiological and hemodynamic response to tDCS, which could be used to monitor tDCS neuro-modulatory effect in health and disease.

Abstract

Over the past three decades, there has been a continuous demand for an increase in the transmission data rate in the aeronautical electronics and communication technologies. This has resulted in the update or creation of new defense and aerospace standards. MIL-STD-1553B is one such bus standard, extensively used in the aerospace and aircraft industry which currently falls short of the modern-day payload requirements as its links support a bandwidth capacity of 1-Mb/s. Present day solutions resort to using an aggregated use of numerous serial buses along with multiple MIL-STD-1553B buses to accommodate high bandwidth inter-aircraft communication requirement. This often leads to less deterministic and complicated systems which require a high cost to develop, integrate, maintain and test. This thesis focuses on improving the bandwidth performance of legacy MIL-STD-1553 data bus by retrofitting redesigned 1553 controllers and analog transceiver upgrades to drive the existing buses at higher data-rates. Therefore, this work explores the enhancement at both digital and analog levels in order to minimize wiring complexity, cable weight and integration cost/time while retaining the valuable features of legacy 1553 standard. Firstly, this research proposes a hardware realization of a 100-Mb/s enhanced data rate MIL-STD-1553B digital controller chip: MIL-STD-1553+, for this purpose. This controller features an integrated remote terminal, bus monitor and bus controller, thus offering a complete 1553 solution at this increased throughput. This work first carried out a comparative study of the reported power and critical path timing aspects of possible architectural techniques with which the existing 1-Mb/s controller design can be improved. Based on this study, a finalization of a synchronous back-end and host processor interface to a true dual port memory to facilitate faster read-and-write memory accesses during controller run-time. A suitable technique of employing high speed majority-based free-running decoders has been chosen at the protocol controller unit‟s front-end interface to receive 100-Mb/s 1553 data. These dual-redundant decoders incorporate both inverted and non-inverted clocks in order to bring the overall clocking rate down to 800-MHz, which conventionally would need a clock of at least 1.6-GHz, thus reducing the power dissipation of the suggested controller. This thesis also suggested partitioning the existing protocol control architecture into two concurrently running state machines which enable the scaled-up scheduling, storage and retrieval of 1553 messages, at 100-Mb/s bit-rate over channel. Finally, the throughput and area-power performance analysis done for a gate-level synthesis and post-layout simulation on a 65-nm, 1.2 V CMOS technology node respectively. This enhanced integrated controller is also implemented on FPGA and checked for block and system level functionality using software logic analyzer IP‟s. Secondly, low power hardware realization of a legacy 1553 analog transceiver upgrade in a 1.8 V, 0:18 m CMOS technology has been made to be able to receive upto 100-Mb/s. Moreover, this transceiver upgrade will support existing legacy 1553 hardware and devices not dedicated to this boosted data rate as well. For this design, we propose a power-saving technique which equalizes the channel inter-symbol interference (ISI) at 100-Mb/s, based on a reconstructed discretized eye diagram accumu-lated after averaging out a large set of comparator decisions, clocked at only 1-MHz. This work presents the working principle, analog design components and digital histogram processing algorithm associated to this transceiver upgrade in detail. This thesis post-layout simulated each design component in this upgrade to verify functionality and followed this with a throughput performance analysis of this system in a communication environment, to observe an optimum error-rate performance.

Abstract

Egocentric cameras are wearable cameras mounted on a person’s head or shoulder. With their ability to capture what the wearer is seeing, such cameras are spawning new set of exciting applications in computer vision. Recognising activity of the wearer from an egocentric video is an important but challenging problem. This problem is more challenging than third person activity recognition due to unavailability of wearer’s pose. Unstructured movement of the camera due to natural head motion of the wearer causes sharp changes in the visual field of the egocentric camera making problem even more challenging. This causes many standard third person action recognition techniques to perform poorly on such videos. On the other hand, objects present in the scene and hand gestures of the wearer are the most important cues for first person action recognition. However, such cues are difficult to segment and recognize in an egocentric video. Carefully crafted features based on hands and objects cues for the problem have been shown to be successful for limited targeted datasets. In the first part of our work, we propose a novel representation of the first person actions derived from feature trajectories. The features are simple to compute using standard feature tracking and does not assume segmentation of hand/objects or recognizing object or hand pose unlike in many previous approaches. We train a bag of words classifier with the proposed features and report a significant performance improvement on publicly available datasets. In the second part of the thesis, we propose convolutional neural networks ( CNN s) for end to end learning and classification of wearers actions. The proposed network makes use of egocentric cues by capturing hand pose, head motion and saliency map. The proposed network model is compact and therefore can be trained from relatively small number of labeled egocentric videos that are available in egocentric settings. We show that the proposed network can generalize and give state of the art performance on various egocentric action datasets widely different from each other visually as well as dynamically.