Human action categorization using conditional random field

Automatic human action recognition has been a challenging issue in the field of machine vision. Some high-level features such as SIFT, although with promising performance for action recognition, are computationally complex to some extent. To deal with this problem, we construct the features based on the Distance Transform of body contours, which is relatively simple and computationally efficient, to represent human action in the video. After extracting the features from videos, we adopt the Conditional Random Field for modeling the temporal action sequences. The proposed method is tested with an available standard dataset. We also testify the robustness of our method on various realistic conditions, such as body occlusion or intersection.

Hierarchical semi-markov conditional random fields for recursive sequential data

Inspired by the hierarchical hidden Markov models (HHMM), we present the hierarchical semi-Markov conditional random field (HSCRF), a generalisation of embedded undirected Markov chains to model complex hierarchical, nested Markov processes. It is parameterised in a discriminative framework and has polynomial time algorithms for learning and inference. Importantly, we develop efficient algorithms for learning and constrained inference in a partially-supervised setting, which is important issue in practice where labels can only be obtained sparsely. We demonstrate the HSCRF in two applications: (i) recognising human activities of daily living (ADLs) from indoor surveillance cameras, and (ii) noun-phrase chunking. We show that the HSCRF is capable of learning rich hierarchical models with reasonable accuracy in both fully and partially observed data cases.

AdaBoost.MRF: boosted Markov random forests and application to multilevel activity recognition

Activity recognition is an important issue in building intelligent monitoring systems. We address the recognition of multilevel activities in this paper via a conditional Markov random field (MRF), known as the dynamic conditional random field (DCRF). Parameter estimation in general MRFs using maximum likelihood is known to be computationally challenging (except for extreme cases), and thus we propose an efficient boosting-based algorithm AdaBoost.MRF for this task. Distinct from most existing work, our algorithm can handle hidden variables (missing labels) and is particularly attractive for smarthouse domains where reliable labels are often sparsely observed. Furthermore, our method works exclusively on trees and thus is guaranteed to converge. We apply the AdaBoost.MRF algorithm to a home video surveillance application and demonstrate its efficacy.

Human activity learning and segmentation using partially hidden discriminative models

Learning and understanding the typical patterns in the daily activities and routines of people from low-level sensory data is an important problem in many application domains such as building smart environments, or providing intelligent assistance. Traditional approaches to this problem typically rely on supervised learning and generative models such as the hidden Markov models and its extensions. While activity data can be readily acquired from pervasive sensors, e.g. in smart environments, providing manual labels to support supervised training is often extremely expensive. In this paper, we propose a new approach based on semi-supervised training of partially hidden discriminative models such as the conditional random field (CRF) and the maximum entropy Markov model (MEMM). We show that these models allow us to incorporate both labeled and unlabeled data for learning, and at the same time, provide us with the flexibility and accuracy of the discriminative framework. Our experimental results in the video surveillance domain illustrate that these models can perform better than their generative counterpart, the partially hidden Markov model, even when a substantial amount of labels are unavailable.

Human action recognition based on pyramid histogram of oriented gradients

Human action recognition has been attracted lots of interest from computer vision researchers due to its various promising applications. In this paper, we employ Pyramid Histogram of Orientation Gradient (PHOG) to characterize human figures for action recognition. Comparing to silhouette-based features, the PHOG descriptor does not require extraction of human silhouettes or contours. Two state-space models, i.e.; Hidden Markov Model (HMM) and Conditional Random Field (CRF), are adopted to model the dynamic human movement. The proposed PHOG descriptor and the state-space models with respect to different parameters are tested using a standard dataset. We also testify the robustness of the method with respect to various unconstrained conditions and viewpoints. Promising experimental result demonstrates the effectiveness and robustness of our proposed method.

Learning discriminative sequence models from partially labelled data for activity recognition

Recognising daily activity patterns of people from low-level sensory data is an important problem. Traditional approaches typically rely on generative models such as the hidden Markov models and training on fully labelled data. While activity data can be readily acquired from pervasive sensors, e.g. in smart environments, providing manual labels to support fully supervised learning is often expensive. In this paper, we propose a new approach based on partially-supervised training of discriminative sequence models such as the conditional random field (CRF) and the maximum entropy Markov model (MEMM). We show that the approach can reduce labelling effort, and at the same time, provides us with the flexibility and accuracy of the discriminative framework. Our experimental results in the video surveillance domain illustrate that these models can perform better than their generative counterpart (i.e. the partially hidden Markov model), even when a substantial amount of labels are unavailable.

Classification ensemble to improve medical named entity recognition

An accurate Named Entity Recognition (NER) is important for knowledge discovery in text mining. This paper proposes an ensemble machine learning approach to recognise Named Entities (NEs) from unstructured and informal medical text. Specifically, Conditional Random Field (CRF) and Maximum Entropy (ME) classifiers are applied individually to the test data set from the i2b2 2010 medication challenge. Each classifier is trained using a different set of features. The first set focuses on the contextual features of the data, while the second concentrates on the linguistic features of each word. The results of the two classifiers are then combined. The proposed approach achieves an f-score of 81.8%, showing a considerable improvement over the results from CRF and ME classifiers individually which achieve f-scores of 76% and 66.3% for the same data set, respectively.

Enhancement of medical named entity recognition using graph-based features

Named Entity Recognition (NER) is a crucial step in text mining. This paper proposes a new graph-based technique for representing unstructured medical text. The new representation is used to extract discriminative features that are able to enhance the NER performance. To evaluate the usefulness of the proposed graph-based technique, the i2b2 medication challenge data set is used. Specifically, the 'treatment' named entities are extracted for evaluation using six different classifiers. The F-measure results of five classifiers are enhanced, with an average improvement of up to 26% in performance.

Discovering semantics from multiple correlated time series stream

In this paper, we study a challenging problem of mining data generating rules and state transforming rules (i.e., semantics) underneath multiple correlated time series streams. A novel Correlation field-based Semantics Learning Framework (CfSLF) is proposed to learn the semantic. In the framework, we use Hidden Markov Random Field (HMRF) method to model relationship between latent states and observations in multiple correlated time series to learn data generating rules. The transforming rules are learned from corresponding latent state sequence of multiple time series based on Markov chain character. The reusable semantics learned by CfSLF can be fed into various analysis tools, such as prediction or anomaly detection. Moreover, we present two algorithms based on the semantics, which can later be applied to next-n step prediction and anomaly detection. Experiments on real world data sets demonstrate the efficiency and effectiveness of the proposed method. © Springer-Verlag 2013.

A general collaborative framework for modeling and perceiving distributed network behavior

Collaborative Anomaly Detection (CAD) is an emerging field of network security in both academia and industry. It has attracted a lot of attention, due to the limitations of traditional fortress-style defense modes. Even though a number of pioneer studies have been conducted in this area, few of them concern about the universality issue. This work focuses on two aspects of it. First, a unified collaborative detection framework is developed based on network virtualization technology. Its purpose is to provide a generic approach that can be applied to designing specific schemes for various application scenarios and objectives. Second, a general behavior perception model is proposed for the unified framework based on hidden Markov random field. Spatial Markovianity is introduced to model the spatial context of distributed network behavior and stochastic interaction among interconnected nodes. Algorithms are derived for parameter estimation, forward prediction, backward smooth, and the normality evaluation of both global network situation and local behavior. Numerical experiments using extensive simulations and several real datasets are presented to validate the proposed solution. Performance-related issues and comparison with related works are discussed.

Preference networks : probabilistic models for recommendation systems

Recommender systems are important to help users select relevant and personalised information over massive amounts of data available. We propose an unified framework called Preference Network (PN) that jointly models various types of domain knowledge for the task of recommendation. The PN is a probabilistic model that systematically combines both content-based filtering and collaborative filtering into a single conditional Markov random field. Once estimated, it serves as a probabilistic database that supports various useful queries such as rating prediction and top-N recommendation. To handle the challenging problem of learning large networks of users and items, we employ a simple but effective pseudo-likelihood with regularisation. Experiments on the movie rating data demonstrate the merits of the PN.

Preference Networks: probabilistic models for recommendation systems

Abstract
Recommender systems are important to help users select relevant and personalised information over massive amounts of data available. We propose an unified framework called Preference Network (PN) that jointly models various types of domain knowledge for the task of recommendation. The PN is a probabilistic model that systematically combines both content-based filtering and collaborative filtering into a single conditional
Markov random field. Once estimated, it serves as a probabilistic database that supports various useful queries such as rating prediction and top-N recommendation. To handle the challenging problem of learning large networks of users and items, we employ a simple but effective pseudo-likelihood with regularisation. Experiments on the movie rating data demonstrate the merits of the PN.

Collaborative filtering via sparse Markov random fields

Recommender systems play a central role in providing individualized access to information and services. This paper focuses on collaborative filtering, an approach that exploits the shared structure among mind-liked users and similar items. In particular, we focus on a formal probabilistic framework known as Markov random fields (MRF). We address the open problem of structure learning and introduce a sparsity-inducing algorithm to automatically estimate the interaction structures between users and between items. Item-item and user-user correlation networks are obtained as a by-product. Large-scale experiments on movie recommendation and date matching datasets demonstrate the power of the proposed method.

A field model for repairing 3D shapes

This paper proposes a field model for repairing 3D shapes constructed from multi-view RGB data. Specifically, we represent a 3D shape in a Markov random field (MRF) in which the geometric information is encoded by random binary variables and the appearance information is retrieved from a set of RGB images captured at multiple viewpoints. The local priors in the MRF model capture the local structures of object shapes and are learnt from 3D shape templates using a convolutional deep belief network. Repairing a 3D shape is formulated as the maximum a posteriori (MAP) estimation in the corresponding MRF. Variational mean field approximation technique is adopted for the MAP estimation. The proposed method was evaluated on both artificial data and real data obtained from reconstruction of practical scenes. Experimental results have shown the robustness and efficiency of the proposed method in repairing noisy and incomplete 3D shapes.

Hyperstereopsis in helmet-mounted NVDs : time to contact estimation

The side mounting of the night-vision sensors on some helmet-mounted systems creates a situation of hyperstereopsis in which the binocular cues available to the operator are exaggerated such that distances around the point of fixation are increased. For a moving surface approaching the observer, the increased apparent distance created by hyperstereopsis should result in greater apparent speed of approach towards the surface and so an operator will have the impression they have reached the surface before contact actually occurs. We simulated motion towards a surface with hyperstereopsis and compared judgements of time to contact with that under normal stereopsis as well as under binocular viewing without stereopsis. We simulated approach of a large, random-field textured and found that time to contact estimates were shorter under the hyperstereoscopic condition than those under normal stereo and no stereo, indicating that hyperstereopsis may cause observers to underestimate time to contact leading operators to undershoot the ground plane when landing.