Rahman Ali

Human Facial Expression Recognition Using Stepwise Linear Discriminant Analysis and Hidden Conditional Random Fields

This paper introduces an accurate and robust facial expression recognition (FER) system. For feature extraction, the proposed FER system employs stepwise linear discriminant analysis (SWLDA). SWLDA focuses on selecting the localized features from the expression frames using the partial F-test values, thereby reducing the within class variance and increasing the low between variance among different expression classes. For recognition, the hidden conditional random fields (HCRFs) model is utilized. HCRF is capable of approximating a complex distribution using a mixture of Gaussian density functions. To achieve optimum results, the system employs a hierarchical recognition strategy. Under these settings, expressions are divided into three categories based on parts of the face that contribute most toward an expression. During recognition, at the first level, SWLDA and HCRF are employed to recognize the expression category; whereas, at the second level, the label for the expression within the recognized category is determined using a separate set of SWLDA and HCRF, trained just for that category. In order to validate the system, four publicly available data sets were used, and a total of four experiments were performed. The weighted average recognition rate for the proposed FER approach was 96.37% across the four different data sets, which is a significant improvement in contrast to the existing FER methods.

Facial expression recognition using active contour-based face detection, facial movement-based feature extraction, and non-linear feature selection

Knowledge about people’s emotions can serve as an important context for automatic service delivery in context-aware systems. Hence, human facial expression recognition (FER) has emerged as an important research area over the last two decades. To accurately recognize expressions, FER systems require automatic face detection followed by the extraction of robust features from important facial parts. Furthermore, the process should be less susceptible to the presence of noise, such as different lighting conditions and variations in facial characteristics of subjects. Accordingly, this work implements a robust FER system, capable of providing high recognition accuracy even in the presence of aforementioned variations. The system uses an unsupervised technique based on active contour model for automatic face detection and extraction. In this model, a combination of two energy functions: Chan–Vese energy and Bhattacharyya distance functions are employed to minimize the dissimilarities within a face and maximize the distance between the face and the background. Next, noise reduction is achieved by means of wavelet decomposition, followed by the extraction of facial movement features using optical flow. These features reflect facial muscle movements which signify static, dynamic, geometric, and appearance characteristics of facial expressions. Post-feature extraction, feature selection, is performed using Stepwise Linear Discriminant Analysis, which is more robust in contrast to previously employed feature selection methods for FER. Finally, expressions are recognized using trained HMM(s). To show the robustness of the proposed system, unlike most of the previous works, which were evaluated using a single dataset, performance of the proposed system is assessed in a large-scale experimentation using five publicly available different datasets. The weighted average recognition rate across these datasets indicates the success of employing the proposed system for FER.

Semi-supervised learning using frequent itemset and ensemble learning for SMS classification

We have used semi-supervised learning with the help of frequent itemset and ensemble learning to classify SMS data into ham and spam.We have used UCI publicly available SMS spam collection, SMS spam collection corpus v.0.1 small and big data set for experimenting our result.We have compared our result with existing semi-supervised learning methods PEBL and SpyEM.We have obtained good results on very low amount of positive dataset and different amount of unlabeled dataset. Short Message Service (SMS) has become one of the most important media of communications due to the rapid increase of mobile users and its easy to use operating mechanism. This flood of SMS goes with the problem of spam SMS that are generated by spurious users. The detection of spam SMS has gotten more attention of researchers in recent times and is treated with a number of different machine learning approaches. Supervised machine learning approaches, used so far, demands a large amount of labeled data which is not always available in real applications. The traditional semi-supervised methods can alleviate this problem but may not produce good results if they are provided with only positive and unlabeled data. In this paper, we have proposed a novel semi-supervised learning method which makes use of frequent itemset and ensemble learning ( FIEL ) to overcome this limitation. In this approach, Apriori algorithm has been used for finding the frequent itemset while Multinomial Naive Bayes, Random Forest and LibSVM are used as base learners for ensemble learning which uses majority voting scheme. Our proposed approach works well with small number of positive data and different amounts of unlabeled dataset with higher accuracy. Extensive experiments have been conducted over UCI SMS spam collection data set, SMS spam collection Corpus v.0.1 Small and Big which show significant improvements in accuracy with very small amount of positive data. We have compared our proposed FIEL approach with the existing SPY-EM and PEBL approaches and the results show that our approach is more stable than the compared approaches with minimum support.

ATHENA: A Personalized Platform to Promote an Active Lifestyle and Wellbeing Based on Physical, Mental and Social Health Primitives

Technology provides ample opportunities for the acquisition and processing of physical, mental and social health primitives. However, several challenges remain for researchers as how to define the relationship between reported physical activities, mood and social interaction to define an active lifestyle. We are conducting a project, ATHENA(activity-awareness for human-engaged wellness applications) to design and integrate the relationship between these basic health primitives to approximate the human lifestyle and real-time recommendations for wellbeing services. Our goal is to develop a system to promote an active lifestyle for individuals and to recommend to them valuable interventions by making comparisons to their past habits. The proposed system processes sensory data through our developed machine learning algorithms inside smart devices and utilizes cloud infrastructure to reduce the cost. We exploit big data infrastructure for massive sensory data storage and fast retrieval for recommendations. Our contributions include the development of a prototype system to promote an active lifestyle and a visual design capable of engaging users in the goal of increasing self-motivation. We believe that our study will impact the design of future ubiquitous wellness applications.

Multimodal hybrid reasoning methodology for personalized wellbeing services

A wellness system provides wellbeing recommendations to support experts in promoting a healthier lifestyle and inducing individuals to adopt healthy habits. Adopting physical activity effectively promotes a healthier lifestyle. A physical activity recommendation system assists users to adopt daily routines to form a best practice of life by involving themselves in healthy physical activities. Traditional physical activity recommendation systems focus on general recommendations applicable to a community of users rather than specific individuals. These recommendations are general in nature and are fit for the community at a certain level, but they are not relevant to every individual based on specific requirements and personal interests. To cover this aspect, we propose a multimodal hybrid reasoning methodology (HRM) that generates personalized physical activity recommendations according to the user׳s specific needs and personal interests. The methodology integrates the rule-based reasoning (RBR), case-based reasoning (CBR), and preference-based reasoning (PBR) approaches in a linear combination that enables personalization of recommendations. RBR uses explicit knowledge rules from physical activity guidelines, CBR uses implicit knowledge from experts׳ past experiences, and PBR uses users׳ personal interests and preferences. To validate the methodology, a weight management scenario is considered and experimented with. The RBR part of the methodology generates goal, weight status, and plan recommendations, the CBR part suggests the top three relevant physical activities for executing the recommended plan, and the PBR part filters out irrelevant recommendations from the suggested ones using the user׳s personal preferences and interests. To evaluate the methodology, a baseline-RBR system is developed, which is improved first using ranged rules and ultimately using a hybrid-CBR. A comparison of the results of these systems shows that hybrid-CBR outperforms the modified-RBR and baseline-RBR systems. Hybrid-CBR yields a 0.94% recall, a 0.97% precision, a 0.95% f-score, and low Type I and Type II errors.

H2RM: A Hybrid Rough Set Reasoning Model for Prediction and Management of Diabetes Mellitus.

Diabetes is a chronic disease characterized by high blood glucose level that results either from a deficiency of insulin produced by the body, or the body’s resistance to the effects of insulin. Accurate and precise reasoning and prediction models greatly help physicians to improve diagnosis, prognosis and treatment procedures of different diseases. Though numerous models have been proposed to solve issues of diagnosis and management of diabetes, they have the following drawbacks: (1) restricted one type of diabetes; (2) lack understandability and explanatory power of the techniques and decision; (3) limited either to prediction purpose or management over the structured contents; and (4) lack competence for dimensionality and vagueness of patient’s data. To overcome these issues, this paper proposes a novel hybrid rough set reasoning model (H2RM) that resolves problems of inaccurate prediction and management of type-1 diabetes mellitus (T1DM) and type-2 diabetes mellitus (T2DM). For verification of the proposed model, experimental data from fifty patients, acquired from a local hospital in semi-structured format, is used. First, the data is transformed into structured format and then used for mining prediction rules. Rough set theory (RST) based techniques and algorithms are used to mine the prediction rules. During the online execution phase of the model, these rules are used to predict T1DM and T2DM for new patients. Furthermore, the proposed model assists physicians to manage diabetes using knowledge extracted from online diabetes guidelines. Correlation-based trend analysis techniques are used to manage diabetic observations. Experimental results demonstrate that the proposed model outperforms the existing methods with 95.9% average and balanced accuracies.

Video-Based Human Activity Recognition Using Multilevel Wavelet Decomposition and Stepwise Linear Discriminant Analysis

Video-based human activity recognition (HAR) means the analysis of motions and behaviors of human from the low level sensors. Over the last decade, automatic HAR is an exigent research area and is considered a significant concern in the field of computer vision and pattern recognition. In this paper, we have presented a robust and an accurate activity recognition system called WS-HAR that consists of wavelet transform coupled with stepwise linear discriminant analysis (SWLDA) followed by hidden Markov model (HMM). Symlet wavelet has been employed in order to extract the features from the activity frames. The most prominent features were selected by proposing a robust technique called stepwise linear discriminant analysis (SWLDA) that focuses on selecting the localized features from the activity frames and discriminating their class based on regression values (i.e., partial F-test values). Finally, we applied a well-known sequential classifier called hidden Markov model (HMM) to give the appropriate labels to the activities. In order to validate the performance of the WS-HAR, we utilized two publicly available standard datasets under two different experimental settings, n–fold cross validation scheme based on subjects; and a set of experiments was performed in order to show the effectiveness of each approach. The weighted average recognition rate for the WS-HAR was 97% across the two different datasets that is a significant improvement in classication accuracy compared to the existing well-known statistical and state-of-the-art methods.

An innovative platform for person-centric health and wellness support

Modern digital technologies are paving the path to a revolutionary new concept of health and wellness care. Nowadays, many new solutions are being released and put at the reach of most consumers for promoting their health and wellness self-management. However, most of these applications are of very limited use, arguable accuracy and scarce interoperability with other similar systems. Accordingly, frameworks that may orchestrate, and intelligently leverage, all the data, information and knowledge generated through these systems are particularly required. This work introduces Mining Minds, an innovative framework that builds on some of the most prominent modern digital technologies, such as Big Data, Cloud Computing, and Internet of Things, to enable the provision of personalized healthcare and wellness support. This paper aims at describing the efficient and rational combination and interoperation of these technologies, as well as their integration with current and future personalized health and wellness services and business.

Human facial expression recognition using curvelet feature extraction and normalized mutual information feature selection

To recognize expressions accurately, facial expression systems require robust feature extraction and feature selection methods. In this paper, a normalized mutual information based feature selection technique is proposed for FER systems. The technique is derived from an existing method, that is, the max-relevance and min-redundancy (mRMR) method. We, however, propose to normalize the mutual information used in this method so that the domination of the relevance or of the redundancy can be eliminated. For feature extraction, curvelet transform is used. After the feature extraction and selection the feature space is reduced by employing linear discriminant analysis (LDA). Finally, hidden Markov model (HMM) is used to recognize the expressions. The proposed FER system (CNF-FER) is validated using four publicly available standard datasets. For each dataset, 10-fold cross validation scheme is utilized. CNF-FER outperformed the existing well-known statistical and state-of-the-art methods by achieving a weighted average recognition rate of 99 % across all the datasets.

Rough set-based approaches for discretization: a compact review

The extraction of knowledge from a huge volume of data using rough set methods requires the transformation of continuous value attributes to discrete intervals. This paper presents a systematic study of the rough set-based discretization (RSBD) techniques found in the literature and categorizes them into a taxonomy. In the literature, no review is solely based on RSBD. Only a few rough set discretizers have been studied, while many new developments have been overlooked and need to be highlighted. Therefore, this study presents a formal taxonomy that provides a useful roadmap for new researchers in the area of RSBD. The review also elaborates the process of RSBD with the help of a case study. The study of the existing literature focuses on the techniques adapted in each article, the comparison of these with other similar approaches, the number of discrete intervals they produce as output, their effects on classification and the application of these techniques in a domain. The techniques adopted in each article have been considered as the foundation for the taxonomy. Moreover, a detailed analysis of the existing discretization techniques has been conducted while keeping the concept of RSBD applications in mind. The findings are summarized and presented in this paper.