Intelligent Interactive Technologies and Multimedia

The future of design computing is in the making of intelligent and interactive technologies for a smart living environment. This chapter explores the concepts of technology as magic and provides examples of creative design computing projects to illustrate the opportunities and challenges for making smart objects, interactive furniture and responsive environments.

Initially commercial crowdsourcing services (such as Amazon’s Mechanical Turk) were focused largely on providing micro-labor services for tasks such as image labeling and text processing. However it is becoming increasingly apparent that these services can also be regarded as providing parallel, on-demand, networks of (so-called) ‘Human Processing Units’ (HPUs). Such services are able to provide specialist computational facilities in a manner analogous to the way Graphics Processing Units (GPUs) support the specialist process of high speed rendering. This paper describes how this new technology could extend the functionality of mechanical CAD/CAM or PLM systems. Crucial to the commercial feasibility of such systems is the ability to access networks of HPUs where engineering data can be processed securely (unlike open crowdsourcing sites such as mTurk). The paper reports the initial results of work done to establish the feasibility of a proposed architecture for integrating HPUs into desktop CAD that uses established BPO centers in rural India to provide a secure source of geometric intelligence.

In this paper, we explore the use of machine learning for multimedia indexing and retrieval involving single/multiple features. Indexing of large image collection has been well researched problem. However, machine learning for combination of features in image indexing and retrieval framework is not explored. In this context, the paper presents novel formulation of multiple kernel learning in hashing for multimedia indexing. The framework learns combination of multiple features/ modalities for defining composite document indices in genetic algorithm based framework. We have demonstrated the evaluation of framework on dataset of handwritten digit images. Subsequently, the utility of the framework is explored for development for multi-modal retrieval of document images.

We give an overview of our general framework for registering 2D and 3D objects without correspondences. Classical solutions consist in extracting landmarks, establishing correspondences and then the aligning transformation is obtained via a complex optimization procedure. In contrast, our framework works without landmark correspondences, is independent of the magnitude of transformation, easy to implement, and has a linear time complexity. The efficiency and robustness of the method has been demonstarted using various deformations models. Herein, we will focus on medical applications.

With the increasing role of computing devices facilitating natural human computer interaction (HCI) will have a positive impact on their usage and acceptance as a whole. Techniques such as vision, sound, speech recognition allow for a much richer form of interaction between the user and machine. The emphasis is to provide a natural form of interface for interaction. As gesture commands are found to be natural for humans, the development of the gesture based system interface have become an important research area. One of the drawbacks of present gesture recognition systems is application dependent which makes it difficult to transfer one gesture control interface into multiple applications. This paper focuses on designing a hand gesture recognition system which is adaptable to multiple applications thus making the gesture recognition systems to be application adaptive. The designed system is comprised of the different processing steps like detection, segmentation, tracking, recognition etc. For making system application-adaptive different quantitative and qualitative parameters have been taken into consideration. The quantitative parameters include gesture recognition rate, features extracted and root mean square error of the system and the qualitative parameters include intuitiveness, accuracy, stress/comfort, computational efficiency, the user’s tolerance, and real-time performance related to the proposed system. These parameters have a vital impact on the performance of the proposed application adaptive hand gesture recognition system.

Among the various existing techniques for the enhancement of dark images, it has been observed that if the images have certain bright area, then it becomes more bright after application of those techniques. The idea has emerged from this existing drawback. In this paper we have changed the parameter used earlier. In aspect of the visual performance the proposed algorithm emerges as quite easy and efficient with keeping concentration on the bright area that should not to be degraded. The decomposition of areas has been done on the basis of entropy of the image. Dynamic SR has been applied in iterative manner and the optimum output response is ensured by the various parameter metrics of the image. Parameter ensures the optimality via statistical and quantitative analysis of the result obtained. The proposed approach is compared with other existing techniques and it has been found that since number of iterations has been reduced drastically, hence the total consumed time. The rate of accuracy is highly increased and no information loss in the output. Also the color quality is maintained and sharpness has been enhanced as well.

In the today’s era of growing information it is very difficult to go through the actual contents of a text-document. Title of a text-document is a collection of meaningful words that signify the content at a glance. This paper presents the method for identifying a title for a text-document using elements of language specific grammar structure i.e. Subject, verb and objects. Ordered this grammar structure is called as RDF (Resource Description Framework). Thismethod firstly, selects certain sentences in the current document. Then it parses them into RDF. Using search engine it retrieves matched documents for the RDF’s. Finally it adopts the title of the best retrieved document as the new title of the current document. This approach works well for any domain related text-documents. The use of Anaphoric reference resolution of discourse processing has further enhanced the result of title identification.

Social capital is about the value of social networks, bonding similar people and bridging between diverse people, with norms of reciprocity. Social capital is fundamentally about how people interact with each other[24]. Public Spaces can provide an ideal location for building social capital. We outline our design approach, which involved conducting a user perception study of interaction in private public spaces. Based on the study, a framework of relationship - patterns was developed. We have further identified key components of social interaction installations and common interaction design modalities. The study reported in this paper will be used to develop a design that aims to foster community bonding in a city environment through socialization and play. The possibility of adopting digital technologies in giving a new meaning to public space has been explored.

Named entities (NEs) play an important role in Cross Lingual Information Retrieval (CLIR). To verify whether documents in two different languages share information about same things, we may check if those two documents have fair number of NEs in common. Comparable documents generally share many named entities. In the present work, we test the effectiveness of named entities in aligning English-Bengali comparable document pairs. We develop an aligned corpus of English-Bengali document pairs using Wikipedia. We crawl English-Bengali document pairs by visiting the cross-lingual links found in the documents on Wikipedia. These document pairs are assumed to be comparable. To find the effectiveness of NE in aligning English-Bengali document pair, each English document is compared with all the other Bengali documents and the most similar Bengali document in terms of NE similarity is found. And then it is verified whether it is aligned successfully (since we already know the correct alignment). Rule based transliteration module is used to transliterate English named entities into Bengali named entities. Since, transliteration modules may not always produce exact transliterations; textual properties like longest common subsequence and minimum edit distance are adopted to check whether two Bengali words can be considered as alignment of each other. Our system achieved an accuracy of 45% for 100 English-Bengali document pairs.

This paper presents an effective self-embedding pixel wise fragile watermarking scheme for image content authentication along with tampered region localization capability. In this approach the watermark is generated from the five most significant bits (MSBs) of each pixel using three different algorithms and embedded into the three least significant bits (LSBs) of the corresponding pixel in the host image. At the receiver side by comparing the recalculated and extracted LSBs, one can easily identify the tampered pixels of the watermarked image. Results of experiments demonstrate that the proposed scheme has very high fragility and fidelity.

Improving the image quality and the rendering speed have always been a challenge to the programmers involved in large scale volume rendering especially in the field of medical image processing. The paper aims to perform volume rendering using the GPU, in which, with its massively parallel capability has the potential to revolutionize this field. The final results would allow the doctors to diagnose and analyze the 2D CT-scan data using three dimensional visualization techniques. The system is used in two types of data, one is human abdomen (45 MB) and colon_phantom8 (300MB) volume data. Further, the use of CUDA framework, a low learning curve technology, for such purpose would greatly reduce the cost involved in CT scan analysis; hence bring it to the common masses. The volume rendering has been done on Nvidia Tesla C1060 card and its performance has also been benchmarked.

Breast cancer is one of the major causes of death among women. If a cancer can be detected early, the options of treatment and the chances of total recovery will increase. From a woman’s point of view, the procedure practiced (compression of breasts to record an image) to obtain a digital mammogram (DM) is exactly the same that is used to obtain a screen film mammogram (SFM). The quality of DM is undoubtedly better than SFM.

However, obtaining DM is costlier and very few institutions can afford DM machines. According to the National Cancer Institute 92% of breast imaging centers in India do not have digital mammography machines [14] and they depend on the conventional SFM.

Hence in this context, one should answer ‘Can SFM be enhanced up to a level of DM?’ In this paper we discuss, our experimental analysis in this regard. We applied elementary image enhancement techniques to obtain enhanced SFM. We performed the quality analysis of digital mammogram and enhanced SFM using standard metrics like PSNR and RMSE on more than 350 mammograms. The results showed that the clarity of processed SFM is as good as digital mammogram.

In many modern image processing applications determining quality of the image is one of the most challenging tasks. Researchers working in the field of image quality assessment design algorithms for measuring and quantifying image quality. The human eye can identify the difference between a good quality image and a noisy image by simply looking at the image, but designing a computer algorithm to automatically determine the quality of an image is a very challenging task. In this paper we propose an image quality measure using the concept of object separability. We define object separability using variance. Two objects are very well separated if variance of individual object is less and mean pixel values of neighboring objects are very different.

A way of generating the future frames of an image sequence is presented. In this paper first we present a way of predicting the future positions of rigid moving objects in a given sequence of images from a static camera.The moving object is first extracted from the images and its centroid is found as a measure of its position. These positions are used to find the future positions of the object using Artificial Neural Network models. This approach is found to predict the positions with very good accuracy. Next we give an algorithm for generating complete future image frames. The optical flow of the images is calculated to find the velocity of each pixel. Time series of the velocities are constructed for each pixel for both dimensions. A separate neural network model is used to predict the future velocities of each pixel and the pixels are then mapped to their new positions. Two different types of neural network models(sigmoidal function networks and radial basis function networks) have been used.

The work described in this paper stems from the Cognitive Wheelchair Project - an effort to build a cognitively enhanced collaborative control architecture for an intelligent wheelchair. A number of challenges arises when developing such a system including ensuring indiscernibility of assistance provided by the system i.e., user unable to realize so easily that he is getting help. In this paper, our focus is primarily on design of such a reactive navigator for collaborative control of an intelligent wheelchair. Under conditions attuned to replicate the scenarios available to the wheelchair, we conducted a series of maze solving experiments. A set of design elements were extracted from the wayfinding experiment leading to the finite state machine (FSM) characterizing the reactive navigator. The FSM arrived at through such an exercise is expected to emulate the cognitive processes of human wayfinding under environment conditions as perceivable to an intelligent wheelchair and ensure indiscernibility of assistance.

Searching the vast and distributed structure of the web requires the efficient search schemes. Semantic annotation is used to associate the meaningful tags with a document to perform semantic search. This paper puts forward an automatic approach for annotating web documents for efficient information retrieval. The proposed algorithm for semantic annotation constitutes five rules based on ontology and provides the semantic tags along with the degree of correlation between a tag and the consequent web document. As the annotation would be done automatically, the results obtained for a query would always be relevant and thus the improved precision and recall.

Rapid advancement in technology calls for efficient applications for empowering the rural population. Education is one of the fields holding innumerable developmental opportunities in rural India and with growing educational research, the challenges faced in the rural education system can be met. Concerns surrounding the Learner, Teacher and Infrastructure can be catered by introducing an intelligent interactive tutor in rural areas. This paper looks at the potential applications supported by the recent developments in Learning Technologies, which can be implemented in the context of Rural India. We also propose a model which uses the Problem Based Learning (PBL) approach to develop conceptual, practical and strategic knowledge of the learners and allow better transferability. Cognitive Load Theory and Learner Models provide the Instruction Design guidelines for the proposed tutoring system. The testing for effectiveness of the conceptual model of this tutor is under progress.

Precise classification of agricultural crops provides vital information on the type and extent of crops cultivated in a particular area. This information plays an important role in planning further cultivation activities. Image classification forms the core of the solution to the crop coverage identification problem. In this paper we present the experimental results obtained by using Gene Expression Programming (GEP) to classify the crop data obtained from satellite images. We have adopted supervised one-against-all learning technique to perform the classification of data. Gene Expression Programming provides an efficient method for obtaining classification rules in the form of a mathematical expression for a given data set containing input and output variables. We have also compared the classification efficiencies obtained with those of other classifiers namely Support vector machines and Artificial neural networks. Sensitivity Analysis has also been carried out to determine the significance of each input variable.

Early identification of defects and prevention of defects migration are key goals of the software security testing process. Early integration of security testing activities into the development lifecycle leads to secure software development. The prescribed key activities of security testing are closely interconnected with security development life cycle to deliver secure software. Software test process elaborates various testing activities and describes which activity is to be carried out when. Given the need and significance of phased approach of security testing, this paper proposes different testing activities to be carried out while integrating it within the security development life cycle.

Software project success or failure depends on the ineffective software project management. Success or failure of any project can be attributed by incorrect handling of one or more project variables such as people, proper technology, proper project scheduling and selection. Among these attributes proper project selection is one of the most vital part of software project management. There exist many uncertainties in project management and current software engineering techniques are unable to eliminate them. So there is huge scope for developing. The current researchers have developed a unique model which is capable to take decision on the field of software project selection. This model has two embedded sub models namely fuzzy AHP (Analytic Hierarchy Process) and strategic game model. Here in the first case experts opinions are considered under fuzzy environment and in the second case, different decisions makers act as players in the game module. Different criteria are taken into consideration for choosing optimal strategy of the players. An elaborated case study is also analyzed for testing the output of the system.

Most work on multimodal interaction in the human computer interaction (HCI) space has focused on enabling a user to use one or more modalities in combination to interact with a system. However, there is still a long way to go towards making human-to-machine communication as rich and intuitive as human-to-human communication. In human-to-human communication, modalities are used individually, simultaneously, interchangeably or in combination. The choice of modalities is dependent on a variety of factors including the context of conversation, social distance, physical proximity, duration, etc. We believe such intuitive multimodal communication is the direction in which human-to-machine interaction is headed in the future. In this paper, we present the insights we have from studying current human-machine interaction methods. We carried out an ethnographic study to observe and study users in their homes as they interacted with media and media devices, by themselves and in small groups. One of the key learning we have from this study is the understanding of the impact of the user’s context on the choice of interaction modalities. The user context factors that influence the choice of interaction modalities include, but are not limited to: the distance of the user from the device/media, the user’s body posture during the media interaction, the user’s involvement level with the media, seating patterns (cluster) of the co-located participants, the roles that each participant plays, the notion of control among the participants, duration of the activity and so on. We believe that the insights from this study can inform the design of the next generation multimodal interfaces that are sensitive to user context, perform a robust interpretation of the interaction inputs and support more human-like multimodal interaction.

Discovering the hierarchy of organizational structure in a dynamic social network can unveil significant patterns which can help in network analysis. In this paper, we formulated a methodology to establish the most influential person in a temporal communication network from the perspective of frequency of interactions which works on hierarchal structure. With the help of frequency of interactions, we have calculated the individual score of each person from Page Rank algorithm. Subsequently, a graph is generated that showed the influence of each individual in the network. Rigorous experiments we performed using Enron data set to establish a fact that our proposed methodology correctly identifies the influential persons over the temporal network. We have used Enron Company’s email data set that describes how employees of company interacted with each other. We could analyze from our methodology and verify from the facts in the Company’s dataset since after bankruptcy, the result of interactions and behaviors of the individual of the network are absolutely known. Our result shows that the proposed methodology is generic and can be applied to other data sets of communication to identify influential at particular instances.

Knowledge base can be defined as the database of knowledge. It comprises of several factors including information, intelligence, skill set and experience. Experience is the most important factor amongst these. This paper talks about the model to calculate the knowledge of an entity in terms of mathematics. Knowledge will be calculated by taking the above mentioned four elements. This will be beneficial for any organizations in the sense that they will be able to calculate their current knowledge and the knowledge required at a particular instance. This will in turn save the wastage of resources that the organization holds knowledge can be created through various factors and utilization of resources of the organization. The second benefit will be reduction in the amount of time taken to create knowledge. These two benefits shall be discussed in detail in the later sections.

This paper presents an integration technique for extraction of Information from high resolution satellite image and also demonstrates the accuracy achieved by the final extracted information. The integration technique comprises of an improved mathematical morphology based watershed transform and a non-linear derivative method. It overcomes all the disadvantages of existing region based and edge based methods by incorporating aforesaid integration methods. It preserves the advantages of multi-resolution and multi-scale gradient approaches. Using these approaches, it avoids excessive fragmentation into regions. The watershed segmentation is proved to be a powerful and fast technique for both contour detection and region-based segmentation. In principle, watershed segmentation depends on ridges to perform a proper segmentation, a property that is often fulfilled for contour detection where the boundaries of the objects have been expressed as ridges. On the other hand, the non-linear derivative method is used for resolving the discrete edge detection problem. Since it automatically selects the best edge localization, which is very much useful for estimation of gradient selection. The main benefit of univocal edge localization is to provide a better direction estimation of the gradient, which helps in producing a confident edge reference map for synthetic images. This nonlinearity will effectively improve global filtration process and regarded to be an effective technique for regularization in order to provide information extraction in a valid manner. The practical merit of this proposed method is to derive an impervious surface from emerging urban areas. Hence this proposed method gives a major contribution in the field of satellite image for information extraction.

Classical biometric system are prone to compromise at several points. Two of the vulnerable points are : 1. biometric database 2. biometric feature matcher subsystem. We propose a two level watermarking scheme to secure these vulnerable points. Watermark W

1

is used for database authentication and made resistive to lossy compression. It is derived using block based singular values (SV’s) of a fingerprint image. W

1

establish linkages between watermark and fingerprint image. Watermark W

2

is used to secure feature matcher subsystem. It is computed using second and third order moments of the fingerprint image. W

2

is made resistive to mild affine transformation and lossy compression to incorporate practical aspects of biometric fingerprint system. The proposed watermarking method not only provides protection to database and matcher subsystem, it also gives security against copy attack.

In present paper, a non-Euclidean approach for finding high dimensional objects has been proposed. The approach is based on the fact that fractal dimension represents the roughness of 2D objects in digital images which can be measured and used to infer about the structure of objects. Since fractal dimension varies in the range 2.0 to 3.0, where the objects having higher value of fractal dimension represent more dense objects in terms of their space filling property, the measurement of fractal dimension leads to discriminate various objects. The image texture obtained from fractal map has been used for this discrimination. The texture map is segmented on the basis of fractal dimension values and segmentation evaluation has been done. The results obtained for the test images are promising and show that the image texture can be segmented using fractal dimension values. The possible future scope of the work has also been highlighted with the applications in real life, e.g., computer vision.

In this paper an efficient preprocessing module has been described which focuses on building a trademark database that can be used for developing a trademark retrieval system. The preprocessing module focuses on noise removal from the trademark images using an adaptive filtering technique using Wiener filters, followed by Karhunen-Loève Transform that makes the trademark search process rotation invariant by rotating the object along positive y direction. Since the registered trademarks are huge in number and will increase invariantly in the future it will be strenuous for the search system to search for similarity in such huge database. Intention is to reduce the search space hence Fuzzy Clustering has been applied.

Diabetic Retinopathy (DR) is one of the chronic diseases which has caused stir in the medical world, since initial symptoms are hard to detect or predict and if it goes unnoticed then it may lead to permanent blindness. So the need arises for fast and efficient systems which can detect whether the patient is suffering from DR or not. In automatic detection of lesion in eye due to diabetic retinopathy the lesions are detected based upon the lesion’s characteristic for e.g. exudates are bright spots and hemorrhages are dark lesions. The detection of lesions facilitates in initial screening step of the disease, with this we can perform automatic screening of images whether they are DR infected or not. In present system with the help of morphological image processing techniques, we are trying to detect lesions in two categories i.e. dark and bright lesions. The present system is able to detect 90 % exudates in image and 85% dark lesions.

Cloud computing is an emerging technology providing software and hardware resources to the users’ as pay-per-use basis, on the other hand in surge demand of current needs of anywhere and anytime the concept of mobile computing came into view. Both aims to provide service to the users as per their requirements and cloud computing provide better flexibility in terms of PaaS, SaaS and IaaS. Database handling is the important consideration of above type of computing environments. Many researchers are proposed issues of database using SQL and NoSQL space suitably for cloud computing scenario. Again available classification for the cloud database sited as a structured and unstructured schema-based along with small and big databases concepts. In this paper we have conducted a study of cloud databases as well as its classification in terms of ACID and NoACID. Again our work focuses on the architectural issues on Database as a Service (DaaS) based on the live migration from ACID based Database to NoACID based Database and vice-versa.

Human gait provides an important and useful behavioral biometric signature which characterizes the nature of an individual’s walking pattern. This inherent knowledge of gait feature confirms the correct identification of a person in a video surveillance footage scenario. In this paper, we attempt to use computer vision based technique to derive the gait signature of a person which is a major criterion for the gait based recognition system. The gait signature has been obtained from the sequence of silhouette images at various gait speeds varying from 2km/hr. to 7km/hr. The OU- ISIR Treadmill walking speed databases have been used in our research work. The joint angles of knee and ankle are computed from the stick figure of corresponding human silhouettes which lead to construct our feature template together with the other gait attributes such as width, height, area and diagonal angle of human silhouette. The combined gait features will make the system robust in different gait speeds. The major concept behind making the gait recognition speed invariant is that the human can walk in finite speed so instead of training the classifier for a single speed the classifier is to be trained for multiple speeds. A minimum distance classifier is used to separate out different cluster of subject with combined feature vectors at different gait speeds.

This study has examined the problem of accurate optimization for fully automatic registration of brain images. Though the proposed global optimization techniques produce encouraging results, their speed of convergence is slow in compare to other local optimization techniques. To speed up the optimization techniques, we introduce multiresolution framework and gain a hierarchical knowledge of transformation parameters. This approach has tried to avoid the stuck in problem of local optimization technique and enhances the speed of convergence of high-dimensional searching algorithms.