Shyamanta M. Hazarika

Towards an architecture for cognitive vision using qualitative spatio-temporal representations and abduction

In recent years there has been increasing interest in constructing cognitive vision systems capable of interpreting the high level semantics of dynamic scenes. Purely quantitative approaches to the task of constructing such systems have met with some success. However, qualitative analysis of dynamic scenes has the advantage of allowing easier generalisation of classes of different behaviours and guarding against the propagation of errors caused by uncertainty and noise in the quantitative data. Our aim is to integrate quantitative and qualitative modes of representation and reasoning for the analysis of dynamic scenes. In particular, in this paper we outline an approach for constructing cognitive vision systems using qualitative spatial-temporal representations including prototypical spatial relations and spatio-temporal event descriptors automatically inferred from input data. The overall architecture relies on abduction: the system searches for explanations, phrased in terms of the learned spatio-temporal event descriptors, to account for the video data.

Qualitative Spatial and Temporal Reasoning: Emerging Applications, Trends, and Directions

The field of Qualitative Spatial and Temporal Representation and Reasoning (QSTR) has evolved as a specialised discipline within Artificial Intelligence (Allen, 1983; Freksa, 1991; van Beek, 1992; Ladkin & Maddux, 1994; Cohn & Renz, 2007; Renz & Nebel, 2007). Recent years have witnessed remarkable advances in some of the long-standing problems of the field, primarily pertaining to spatial calculi and model construction issues emanating from the founding premises and early work in the community (Ligozat, 1990; Guesgen & Hertzberg, 1993, 1988). Subsequently, major developments have accrued with new results about tractability of spatial calculi and characterisation of important subclasses of relations (e.g., Nebel & Bürckert, 1994; Bessière et al., 1996; Renz, 1999, 2007; Li et al., 2009) and explicit construction of models of one or more aspects of space (e.g., Freksa, 1992; Randell et al., 1992; Cohn et al., 1997; Bennett, 2001; de Weghe et al., 2005; Moratz, 2006). Similar to these works, which are situated within an Artificial Intelligence/Knowledge Representation (KR) context, many crucial advances have accrued from other communities concerned with the development of formalisms and algorithms for modelling and reasoning about spatial information, a prime example here being the domain of spatial information theory for Geography (and Geographic Information Systems (GIS)) (Egenhofer & Franzosa, 1991; Egenhofer & Mark, 1995).

Qualitative Spatio-Temporal Continuity

We explore different intuitive notions of spatio-temporal continuity and give a formal characterization of continuity for space-time histories.We investigate the types of transitions possible for the RCC-8 topological relations under each distinct notion of spatio-temporal continuity and provide a hierarchy of conceptual neighbourhood diagrams.

Qualitative Spatio-Temporal Representation and Reasoning: Trends and Future Directions

This chapter focuses on the topological and mereological relations, contact, and parthood, between spatiotemporal regions as axiomatized in so-called mereotopologies. Despite, or because of, their simplicity, a variety of different first-order axiomatizations have been proposed. This chapter discusses their underlying ontological choices and different ways of systematically looking at them. The chapter further gives an overview of the algebraic, topological, and graph-theoretic representations of mereotopological models which help to better understand the model-theoretic consequences of the various ontological choices. While much work on mereotopologies has been primarily theoretical, the focus started shifting towards applications and domain-specific extensions of mereotopology. These aspects will most likely guide the future direction of the field: How can mereotopologies be extended or otherwise adjusted to better suit practical needs? Moreover, the integration of mereotopology into more comprehensive and maybe more pragmatic ontologies of space and time remains another challenge in the field of region-based space.

Security Pattern Lattice: A Formal Model to Organize Security Patterns

Except for some work in classifying security patterns (SP) based on taxonomy and linguistic metaphors not much has been done in organizing SP. No suitable formal model for organization of security patterns is yet available. In this paper, exploiting results from formal concept analysis (FCA) a formal model to organize SP is introduced. Using a trust-based security model, trust elements (TE) and SP are enumerated. These elements and patterns are treated as formal concepts leading to a security pattern lattice SPL and organized as a concept lattice to generate the security pattern lattice. Within such a model the patterns are categorized using the FCA technique of scaling.

Formal concept analysis: current trends and directions

Formalization of human thinking helps in fostering the process of learning by giving an explicit representation to human thoughts. Formal Concept Analysis (FCA) finds it’s core here. It considers a “concept” as a formal unit of human thought. A concept is represented as a set of inter related objects called the extent and the set of the properties of these objects, called the intent. Making use of the mathematical principles of Lattice Theory and Map Theory of Abstract Algebra, a set of tools and algorithms have been developed in FCA. These helps us to analyze and represent any context as a relation between it’s extent and intent. Concepts drawn from the subsets of the extent and intent can be organized in the form of a lattice giving a subsumption hierarchy. Such concept lattices could be maintained by different operations on the lattice like scaling, pruning, navigating etc. A host of applications and software have been developed over the years which serves the usage of FCA tools and processes for specific purposes in various fields. This paper reviews the theoretical foundation, research and applications of FCA in different areas. The paper projects current trends in FCA and concludes with a discussion on open issues and limitations of FCA.

Motor imagery based BCI for a maze game

Electroencephalogram (EEG) signals generated out of motor imagery (MI) can be used for Brain Computer Interfacing (BCI). In order to accomplish this goal, we have classified four different MI tasks using two hybrid features of bispectrum of EEG through a RBF kernel support vector machine. As a demonstration of its applicability for a non-invasive BCI, we design and develop a BCI maze game, where a player plays the game in real time using his brain signals.

Bispectral Analysis of EEG for Emotion Recognition

Abstract Emotion recognition from electroencephalogram (EEG) signals is one of the most challenging tasks. Bispectral analysis offers a way of gaining phase information by detecting phase relationships between frequency components and characterizing the non- Gaussian information contained in the EEG signals. In this paper, we explore derived features of bispectrum for quantification of emotions using a Valence-Arousal emotion model; and arrive at a feature vector through backward sequential search. Cross- validated accuracies of 64.84% for Low/High Arousal classification and 61.17% for Low/High Valence were obtained on the DEAP data set based on the proposed features; comparable to classification accuracies reported in the literature.

Enhanced Shape Context for Object Recognition

This paper presents an enhanced approach to recognize objects based on a similarity measure obtained from shape context. Typically, shape context computation samples at regular interval on the contour of an object without regard to landmarks. Corner points of an object being landmarks on the contour; set of corner points is a good descriptor of shape. The paper explores the possibility of computing shape context by sampling the corner points using arch height. Sampling on this boundary feature of the object considerably reduces the number of points. Landmark based shape context provides a stricter algorithm on similarity. Shape context based object recognition being an iterative process involving comparisons and transformations, the reduction in the number of sample points provides a basis for faster recognition of similar objects.

Bispectrum analysis of EEG during observation and imagination of hand movement

This paper explores the use of bispectrum analysis of electroencephalogram (EEG) signals in estimation of imagination and observation of hand movements. Five different tasks were observed and subsequently performed in separate runs of 10 trials. Subjects were asked to observe and subsequently perform task on the presentation of an audio cue (both at the start and end of session). The acquired EEG signals after averaging in time domain were analyzed using bispectrum. The result shows bispectrum analysis of EEG signals provide a way to discriminate mental representation during observation and imagination of hand movement. Our preliminary investigation, have shown the ability of bispectrum in detecting non-linear phase coupling between alpha and beta rhythms during observation and imagination of hand movement. Further, different bispectral peaks (for tasks with and without prior visual representation) have reinforced our belief that visual representation of motor acts make difference during motor imagination.