Emerging Trends and Advanced Technologies for Computational Intelligence

In this paper, we report a threshold-based method of fall detection using plantar inclinometer sensor, which provides us the information of angle variations during walking, and of angle status after a fall. The angle variations and status are collected in three-dimensional space. We analyzed the normal range of angle variations during walking, and selected the thresholds by testing the distribution of plantar angles of falls. In the experiments, thresholds were selected from plantar angles of fall status in four directions: forward, backward, left and right. Using the selected thresholds, we detected falls of five subjects in different situations for five hundred times and obtained the average detection rate of 85.4 %.

This paper proposes an approach for an ICT-supported medical assessment, by merging measures of signs and symptoms from heterogeneous sources. The disease status estimate of patients that suffer from spondylarthropathy is evaluated with different types of uncertainties using a fuzzy rule-based evidential reasoning (FURBER) approach. The approach treats measures of signs and symptoms in order to define the disease status. We take in consideration the Bath indices and the ASDAS index, described by using fuzzy linguistic variables. A fuzzy rule-base designed on the basis of a belief structure is exploited to capture uncertainty and non-linear relationships between these parameters and the disease status. The inference of the rule-based system is implemented using an evidential reasoning algorithm. An expected utility-based health score is used to assess disease activity over time and to measure the response to treatment. Our tool may be particularly helpful in monitoring the response of treatments and in interpreting the response to therapeutic interventions in clinical trials. A case study is used to illustrate the application of the proposed approach.

Rescue system with sensor network for vital sign (Body Temperature, Heart Rate, Blood Pressure, Bless Rate and Consciousness) monitoring is proposed together with rescue simulations with consideration of triage by using the measured vital sign. Triage is a key for evacuation from disaster areas. Triage can be done with the gathered physical and psychological data with the sensor network for vital sign monitoring. Through a comparison between with and without consideration of triage, it is found that the time required for evacuation from disaster areas with consideration triage is 30 % less than that without triage.

Nowadays almost everyone has access to mobile devices that offer better processing capabilities and access to new information and services, the Web is undoubtedly the best tool for sharing content, especially through social networks. Web content enhanced by mobile capabilities, enable the gathering and aggregation of information that can be useful for our everyday lives as, for example, in urban mobility where personalized real-time traffic information, can heavily influence users’ travel habits, thus contributing for a better way of living. Current navigation systems fall short in several ways in order to satisfy the need to process and reason upon such volumes of data, namely, to accurately provide information about urban traffic in real-time and the possibility to personalize the information presented to users. The work presented here describes an approach to integrate, fuse and process tweet messages from traffic agencies, with the objective of detecting the geographical span of traffic events, such as accidents or road works. Tweet messages are considered in this work given their uniqueness, their real time nature, which may be used to quickly detect a traffic event, and their simplicity. We also address some imprecisions ranging from lack of geographical information, imprecise and ambiguous toponyms, overlaps and repetitions as well as visualization to our data set in the UK, and a qualitative study on the use of the approach using tweets in other languages, such as Greek. Finally, we present an application scenario, where traffic information is processed from tweets massages, triggering personalized notifications to users through Google Cloud Messaging on Android smartphones. The work presented here is still part of on-going work. Results achieved so far do not address the final conclusions but form the basis for the formalization of a domain knowledge along with the urban mobility services.

The research presents a process for applying data mining techniques on dental medical records comprised of oral conditions and different dental procedures that are performed on various patients. The dental expert decides to pursue a set of procedures based on the examination and diagnostics. Digital dentistry is becoming more and more active now, hence this research addresses the issues in exploiting the digital data at its potential like heterogeneous data gathering, access restrictions or inadequate patient data and lack of expert systems to utilize the data. It proposes a way to deal with the dental medical records and apply data mining. Having gathered the dental data and prepared it through pre-processing techniques, unsupervised learning techniques were applied to perform clustering in order to discover interesting patterns and assigning these a label class. Mostly the patients lie in the mild and moderate dental patient’s class. The most common problem that is being noticed in patients is tooth cavity with a treatment named “resin-based composite—one surface, posterior”. Using this labelled data set, supervised learning algorithms were applied to train and test the data for predicting the targeted class accurately. A comparison between classification algorithms based on their accuracy was made to filter out the best outcome. An expert system has also been developed to support the idea, ease up the decision making process and automate the manual practices that are being used. It provides quick recommendations to the medical expert in examining the patient depending upon the diagnosis. Research reveals that decision tree runs better than others on our data set with highest accuracy in predicting the Patients’ targeted classes.

This paper evaluates, using the Random Entry Protocol technique, a high-frequency trading strategy based on a Dynamic Bayesian Network (DBN) that can identify predictive trend patterns in foreign exchange orden-driven markets. The proposed DNB allows simultaneously to represent expert knowledge of skilled traders in a model structure and to learn computationally from data information that reflects relevant market sentiment dynamics. The DBN is derived from a Hierarchical Hidden Markov Model (HHMM) that incorporates expert knowledge in its design and learns the trend patterns present in the market data. The wavelet representation is used to produce compact representations of the LOB liquidity dynamics that simultaneously reduces the time complexity of the computational learning and improves its precision. In previous works, this trading strategy has been shown to be competitive when compared with conventional techniques. However, these works failed to control for unwanted dependencies in the return series used for training and testing that may have skewed performance results to the positive side. This paper constructs key trading strategy estimators based on the Random Entry Protocol over the USD-COP data. This technique eliminates unwanted dependencies on returns and order flow while keeps the natural autocorrelation structure of the Limit Order Book (LOB). It is still concluded that the HHMM-based model results are competitive with a positive, statistically significant P/L and a well-understood risk profile. Buy-and-Hold results calculated over the testing period are provided for comparison reasons.

Cloud computing is evolving into a popular platform that enables on-demand provisioning of computing resources to a growing population of clients. Core to the provisioning of service in the cloud is the discovery of these services in an efficient and timely manner. Centralized and hierarchical approaches to service discovery have exhibited bottlenecks as network load increases and limitation in scalability. Efforts have been made in combining cloud systems and Peer to peer P2P systems to address the problem encountered in the conventional service discovery approaches but not without a new set of challenges ranging from network flooding to poor performance in dynamic networks. This paper presents an efficient and scalable approach for semantic cloud service discovery in a P2P cloud environment. The approach is based on Learning Automata LA and Ant Colony Optimization ACO. The ability of ACO to adapt to changes in real time makes it a better choice in dynamic environments such as cloud. We evaluate this approach against the some existing P2P service discovery approaches, the proposed mechanism showed an improved performance.

Recently, it has been noted that user behaviour can have a large impact on the final energy consumption in buildings. Through the combination of mathematical modelling and data from wireless ambient sensors, we can model human behaviour patterns and use the information to regulate building management systems (BMS) in order to achieve the best trade-off between user comfort and energy efficiency. Furthermore, streaming sensor data can be used to perform real-time classification. In this work, we have modelled user activity patterns using both offline and online learning approaches based on non-linear multi-class Support Vector Machines. We have conducted a comparison study with other machine learning approaches (i.e. Linear SVM, Hidden-Markov and K-nearest models). Experimental results show that our proposed approach outperforms the other methods for the scenarios evaluated in terms of accuracy and processing speed.

The sharing of communication networks, especially with multimedia services such as IPTV, video conferencing and VoIP has increased in recent years. These services require more resources and generate a great demand on the network infrastructure, requiring the guarantee quality of services. For this, scheduling mechanisms, call admission control and traffic policing should be present to guarantee quality of service. The networks of communication for wireless broadband, based on the IEEE 802.16 standard, called WiMAX are used in this work, because this standard only specify the mechanisms of how these policies should be implemented. Based on these factors, a new architecture was developed in order guarantee the quality of service, using the meta-heuristic Particle Swarm Optimization for fixed WiMAX networks, presenting a method for calculating the duration of the time frame, which allows a control of queues in the scheduler in order to uplink traffic from the base station.

Conventional contrast enhancement techniques often fail to produce satisfactory results for low-contrast images, and cannot be automatically applied to different images because their processing parameters must be specified manually to produce a satisfactory result for a given image. This work presents a colour-preserving contrast enhancement (CPCE) algorithm for images. Modification to images was performed in the HSV colour-space. The Hue component is preserved (unchanged), luminance modified using Contrast Limited Adaptive Histogram Equalization (CLAHE), while Saturation components were up-scaled using a derived mapping function on the approximate components of its discrete wavelet transform. Implementation was done in MATLAB and compared with CLAHE and Histogram Equalization (HE) algorithms in the RGB colour space. Subjective (visual quality inspection) and objective parameters (Peak-signal-to-noise ratio (PSNR), Absolute Mean Brightness Error (AMBE) and Mean squared error (MSE)) were used for performance evaluation. The method produced images with the lowest MSE, AMBE, and highest PSNR when tested, yet preserved the visual quality of the image.

This study proposes the Sequential Pattern Discovery algorithms to solve weather prediction problem. A novel weather pattern discovery framework is presented to highlight the important processes in this work. Two algorithms are employed; namely episodes and sequential pattern mining algorithms. The episodes mining algorithm is introduced to find frequent episodes in rainfall sequences and sequential pattern mining algorithm to find relationship of patterns between weather stations. Real data are collected from ten rainfall stations of Selangor State, Malaysia. The sequential pattern algorithm is applied to extract the relationship between ten rainfall stations in 33 years periods of time. The patterns are evaluated experimentally by support and confidence values while some specific rules are mapped to the location of stations and analysed for more verification. The proposed study produces valuable patterns of weather and preserves important knowledge for weather prediction.

A recommender system is a tool employed to filter the huge amounts of data that companies have to deal with, and produce effective suggestions to the users. The estimation of the interest of a user toward an item, however, is usually performed at the level of a single item, i.e., for each item not evaluated by a user, canonical approaches look for the rating given by similar users for that item, or for an item with similar content. Such approach leads toward the so-called overspecialization/serendipity problem, in which the recommended items are trivial and users do not come across surprising items. This work first shows that user preferences are actually distributed over a small set of classes of items, leading the recommended items to be too similar to the ones already evaluated, then we propose a novel model, named Class Path Information (CPI), able to represent the current and future preferences of the users in terms of a ranked set of classes of items. The proposed approach is based on a semantic analysis of the items evaluated by the users, in order to extend their ground truth and infer the future preferences. The performed experiments show that our approach, by including in the CPI model the same classes predicted by a state-of-the-art recommender system, is able to accurately model the user preferences in terms of classes, instead of in terms of single items, allowing to recommend non trivial items.

Closed-circuit television cameras are used extensively to monitor streets for the security of the public. Whether passively recording day-to-day life, or actively monitoring a developing situation such as public disorder, the videos recorded have proven invaluable to police forces world wide to trace suspects and victims alike. The volume of video produced from the array of camera covering even a small area is large, and growing in modern society, and post-event analysis of collected video is a time consuming problem for police forces that is increasing. Automated computer vision analysis is desirable, but current systems are unable to reliably process videos from CCTV cameras. The video quality is low, and computer vision algorithms are unable to perform sufficiently to achieve usable results. In this chapter, we describe some of the reasons for the failure of contemporary algorithms and focus on the fundamental task of feature correspondence between frames of video—a well-studied and often considered solved problem in high quality videos, but still a challenge in low quality imagery. We present solutions to some of the problems that we acknowledge, and provide a comprehensive analysis where we demonstrate feature matching using a 138-dimensional descriptor that improves the matching performance of a state-of-the-art 384-dimension colour descriptor with just \(36\,\%\) of the storage requirements.

Disease diagnosis and severity assessment are necessary and critical for predicting the likely crop yield losses, evaluating the economic impact of the disease, and determining whether preventive treatments are worthwhile or particular control strategies could be taken. In this work, we propose to make advances in the field of automatic detection and diagnosis and severity assessment of crop diseases using image pattern recognition. We have developed a two-stage crop disease pattern recognition system which can automatically identify crop diseases and assess sevrity based on combination of marker-controlled watershed segmentation, superpixel based feature analysis and classification. We have conducted experimental evaluation using different feature selection and classification methods. The experimental result shows that the proposed approach can accurately detect crop diseases (i.e. Septoria and Yellow rust, which are the two most important and major types of wheat diseases in UK and across the world) and assess the disease severity with efficient processing speed.

This chapter presents a discussion about the relationship between the image complexity and the visual working memory capacity. In advertisement and web site design, the mismatch between the target objects and the real salient objects can represent the degree of image complexity which is an important reason of low efficiency and unpleasant reading. Many psychological experiments have also shown the effect of image complexity on short term memory. In this chapter, a method was introduced to measure this mismatch and the image complexity. The present algorithm used in this method combines the mathematic algorithm like SIFT (Scale Invariant Feature Transformation) and K-means with the cognitive science theory of visual working memory capacity. Results of the measurement method were validated by the visual working memory practical experiments. Besides, the results from EEG study of visual working memory on the same group of test images are also consistent with the value from our algorithms.

Virtual Reality with associated hardware and software advances is becoming a viable tool in neuroscience and similar fields. Technology has been harnessed to modify a user’s state of mind for some time through different approaches. Combining this background with merged reality systems, it is possible to develop intelligent tools which can manipulate brain states and enhance training mechanisms.

This work aims at defining a new approach for real-time obstacle detection based on sparse disparity map. To achieve fast time execution, the image processing is performed on a subset of points. The proposed method begins by extracting the pixel primitives from a pair images. Then, a detection of the ground is made to limit obstacles area using color information. Subsequently, an extraction of subset of points is performed using the features extracted in the previous step. The use of the extracted features as well as the subset of points will enable us to calculate a sparse disparity map which is used to calculate the v-disparity image. The use of v-disparity image allows us to extract the obstacle profile in order to eliminate the ground pixels. To eliminate the obstacles belonging to the background, we use the ground limits to remove all object points outside these limits. Finally, a segmentation based on disparity value of the remaining points is applied to extract obstacles. To end, validation step is used to remove false positives. The experimental results obtained are interesting with real time execution. This allows early detection to prevent collisions.

In this paper we present a vision-based approach to detect multiple persons with occlusion handling from a mobile robot in real-world scenarios under two lighting conditions, good illumination (lighted) and poor illumination (dark). We use depth and thermal information that are fused for occlusion handling. First, a classifier is trained using thermal images of the human upper-body. This classifier is used to obtain the bounding box coordinates of human. The depth image is later fused with the region of interest obtained from the thermal image. Using the initial bounding box, occlusion handling is performed to determine the final position of human in the image. The proposed method significantly improves human detection even in crowded scene and poor illumination.

Many studies have been conducted, investigating the effects that diabetes has to the retinal vasculature. Identifying and quantifying the retinal vascular changes remains a very challenging task, due to the heterogeneity of the retina. Monitoring the progression requires follow-up studies of progressed patients, since human retina naturally adapts to many different stimuli, making it hard to associate any changes with a disease. In this novel study, data from twenty five diabetic patients, who progressed to diabetic retinopathy, were used. The progression was evaluated using multiple geometric features, like vessels widths and angles, tortuosity, central retinal artery and vein equivalent, fractal dimension, lacunarity, in addition to the corresponding descriptive statistics of them. A statistical mixed model design was used to evaluate the significance of the changes between two periods: 3 years before the onset of diabetic retinopathy and the first year of diabetic retinopathy. Moreover, the discriminative power of these features was evaluated using a random forests classifier and also a penalized logistic regression. The area under the ROC curve after running a ten-fold cross validation was 0.7925 and 0.785 respectively.

A RADAR system deals with many different and diverse problems for the last few decades. The detection capability of radar is one of the most important factors. The main objective of radar target detection is to improve probability of detection while reducing probability of a false alarm at the same time. To improve the probability of detection of moving target, a new approach is proposed in this paper using wavelet and Hough transforms. The wavelet de-noising technique is used to remove noise from received signal. Then the image processing technique of the Hough transform is used to detect moving target. To reduce the noises form received signal, we propose a new wavelet threshold function that reduces constant error of soft thresholding and improves the discontinuity of hard thresholding. We present performances of our method on a basis of the new thresholding technique and compare with traditional method. It is shown that detection performance of proposed method is superior to that obtained through traditional method.