In this paper, we leverage gait to potentially detect some of the important neurological disorders, namely Parkinson’s disease, Diplegia, Hemiplegia, and Huntington’s Chorea. Persons with these neurological disorders often have a very abnormal gait, which motivates us to target gait for their potential detection. Some of the abnormalities involve the circumduction of legs, forward-bending, involuntary movements, etc. To detect such abnormalities in gait, we develop gait features from the key-points of the human pose, namely shoulders, elbows, hips, knees, ankles, etc. To evaluate the effectiveness of our gait features in detecting the abnormalities related to these diseases, we build a synthetic video dataset of persons mimicking the gait of persons with such disorders, considering the difficulty in finding a sufficient number of people with these disorders. We name it NeuroSynGait video dataset. Experiments demonstrated that our gait features were indeed successful in detecting these abnormalities.

Motion detection and segmentation of traffic vehicles in an outdoor environment, particularly under nonideal weather conditions, in the presence of camera noise and with variable or unfavorable luminance conditions is still an area of active research. Gaussian-based background modeling is commonly used to detect moving objects in computer vision systems. However, it has some limitations: It cannot effectively deal with sudden change in illumination, snowfall, fog, and repetitive motions such as swaying leaves. These nonideal outdoor conditions result in false motion detection. We propose an alternative technique to detect and segment the moving vehicles by making use of dynamically adaptive threshold using the full-search sum of absolute difference (FSSAD) algorithm. We show that motion energy that is obtained using sequence of frames can effectively be used to differentiate between moving vehicles and a dynamic background. Because FSSAD is computationally expensive, we propose a modification using the adaptive-motion threshold, which not only reduces the false motion but improves the computational efficiency as well. Performance evaluation of our proposed framework has been carried out on publicly available benchmark data sets. Qualitative comparison with other methods that are proposed in the literature shows that our approach achieves better segmentation and is suitable for real-time implementation.

MOTIVATION: Discovery of nucleotide motifs that are localized with respect to a certain biological landmark is important in several appli-cations, such as in regulatory sequences flanking the transcription start site, in the neighborhood of known transcription factor binding sites, and in transcription factor binding regions discovered by massively parallel sequencing (ChIP-Seq). RESULTS: We report an algorithm called LocalMotif to discover such localized motifs. The algorithm is based on a novel scoring function, called spatial confinement score, which can determine the exact interval of localization of a motif. This score is combined with other existing scoring measures including over-representation and relative entropy to determine the overall prominence of the motif. The approach successfully discovers biologically relevant motifs and their intervals of localization in scenarios where the motifs cannot be discovered by general motif finding tools. It is especially useful for discovering multiple co-localized motifs in a set of regulatory sequences, such as those identified by ChIP-Seq. AVAILABILITY AND IMPLEMENTATION: The LocalMotif software is available at http://www.comp.nus.edu.sg/~bioinfo/LocalMotif.

This paper deals with the potential application of National Oceanic and Atmospheric Administration (NOAA)/Advanced Very High Resolution Radiometer (AVHRR) data to detect subsurface fire (subsurface hotspots) by proposing an efficient contextual algorithm. Most of the solutions proposed to date are mainly focused on the problem of surface fires, and very few research works have been performed to develop techniques for the subsurface fire problem. Although few algorithms based on the fixed-thresholding approach have been proposed for subsurface hotspot detection, however, for each application, thresholds have to be specifically tuned to cope with unique environmental conditions. The main objective of this paper is to develop an instrument-independent adaptive method by which direct threshold or multithreshold can be avoided. The proposed contextual algorithm is very helpful to monitor subsurface hotspots with operational satellite data, such as the Jharia region of India, without making any region-specific guess in thresholding. Novelty of the proposed work lies in the fact that once the algorithmic model is developed for the particular region of interest after optimizing the model parameters, there is no need to optimize those parameters again for further satellite images. Hence, the developed model can be used for optimized automated detection and monitoring of subsurface hotspots for future images of the particular region of interest. The algorithm is adaptive in nature and uses vegetation index and different NOAA/AVHRR channel's statistics to detect hotspots in the region of interest. The performance of the algorithm is assessed in terms of sensitivity and specificity and compared with other well-known thresholding techniques such as Otsu's thresholding, entropy-based thresholding, and existing contextual algorithm proposed by Flasse and Ceccato. The proposed algorithm is found to give better hotspot detection accuracy with lesser false alarm rate.

G-protein coupled receptors (GPCRs) play a vital role in different biological processes, such as regulation of growth, death, and metabolism of cells. GPCRs are the focus of significant amount of current pharmaceutical research since they interact with more than 50% of prescription drugs. The dipeptide-based support vector machine (SVM) approach is the most accurate technique to identify and classify the GPCRs. However, this approach has two major disadvantages. First, the dimension of dipeptide-based feature vector is equal to 400. The large dimension makes the classification task computationally and memory wise inefficient. Second, it does not consider the biological properties of protein sequence for identification and classification of GPCRs. In this paper, we present a novel-feature-based SVM classification technique. The novel features are derived by applying wavelet-based time series analysis approach on protein sequences. The proposed feature space summarizes the variance information of seven important biological properties of amino acids in a protein sequence. In addition, the dimension of the feature vector for proposed technique is equal to 35. Experiments were performed on GPCRs protein sequences available at GPCRs Database. Our approach achieves an accuracy of 99.9%, 98.06%, 97.78%, and 94.08% for GPCR superfamily, families, subfamilies, and subsubfamilies (amine group), respectively, when evaluated using fivefold cross-validation. Further, an accuracy of 99.8%, 97.26%, and 97.84% was obtained when evaluated on unseen or recall datasets of GPCR superfamily, families, and subfamilies, respectively. Comparison with dipeptide-based SVM technique shows the effectiveness of our approach.

In this paper, harmonic analysis of 10-year time series (1995-2005) NOAA/AVHRR yearly composite images is performed to develop an innovative technique for hotspot detection and land-features classification based on temporal changes in the NDVI and various AVHRR band values. NOAA/AVHRR images are used due to wide coverage, high frequency and free acquisition offered by NOAA/AVHRR sensors. Proposed algorithm consists of three steps: (1) preprocessing of NOAA/AVHRR images to correct geometric distortions and calibrate the data radiometrically (2) detection of cloud and water pixels in the preprocessed image and application of harmonic analysis on 10-years time series AVHRR images to produce phase and amplitude images, and (3) application of image processing techniques on the amplitude images of different bands to detect hotspots and classify the region of interest. The obtained results indicate that the proposed method can classify the region of interest successfully with through out greater than 91% classification accuracy.

With the availability of multisensor and multiresolution image data from operational Earth Observation satellites, the fusion of digital image data has become a valuable tool in land cover classification. Digital image fusion is a relatively new research field at the leading edge of available technology. It forms a rapidly developing area of research in land cover classification. It is needed that to fuse high resolution satellite data with low resolution satellite data, to enhance the classification and interpretation in low resolution satellite data. The AVHRR and MODIS data are freely available, but resolution is poor. Therefore in this paper, it is attempted to highlight the AVHRR and MODIS utility with fusion of ASTER data. In this paper, a fusion method based on the Curvelet transform is introduced. The curvelet transform represents edges more accurately, since edges play a fundamental role in image understanding, one good way to enhance spatial resolution is to enhance the edges. Curvelet-based image fusion method provides richer information in the spatial and spectral domains simultaneously.

In silico motif finding algorithms are often used for discovering protein-DNA binding sites in a set of regulatory sequences. Current algorithms mainly address motif discovery in short sequences. Analyzing long sequences can be quite challenging not only due to increasing time and memory requirements of the algorithm, but also decreasing accuracy. However, in case the motif is localized in a short interval of the long sequences relative to an anchor point, it is tenable to detect it easily by restricting the search to that interval. But the region of localization of the motif is not known a priori. This paper reports an algorithm called LocalMotif to detect localized motifs in long regulatory sequences. A novel score function predicts the region of localization of the motif. This score is combined with other scoring measures including Z-score and relative entropy to detect the motif. The algorithm is optimized for fast processing of long regulatory sequences. Tests on simulated and real datasets confirm that LocalMotif accurately determines the region of localization of motifs and automatically discovers the biologically relevant motifs, which can be detected by other motif finding algorithms only when the search is restricted to the relevant interval

Exact determination of gene network is required to discover the higher-order structures of an organism and to interpret its behavior. Most research work in learning gene networks either assumes that there is no time delay in gene expression or that there is a constant time delay. The paper shows how Bayesian networks can be applied to represent multitime delay relationships as well as directed loops. The intractability of the network learning algorithm is handled by using an improved mutual information criteria. Also, a new structure learning algorithm, "learning by modification", is proposed to learn the sparse structure of a gene network. The experimental results on synthetic data and real data show that our method is more accurate in determining the gene structure as compared to the traditional methods. Even for transcriptional loops spanning over the whole cell, our algorithm can detect them.

Bayesian theory is of great interest in pattern classification. We present an approach to aid in the effective application of Bayesian networks in tasks like video classification, where descriptors originate from varied sources and are large in number. In order to extend the application of conventional Bayesian theory to the case of continuous and nonparametric descriptor space, dimension partitioning into attributes by minimizing the discrete Bayes error is proposed. The partitioning output goes to the dimensionality reduction module. A new algorithm for dimensionality reduction for improving the classification accuracy is proposed based on the class pair discriminative capacity of the dimensions. It is also shown how attributes can be weighed automatically in a single-label assignment based on comparing the class pairs. A computationally efficient method to assign multiple labels on the samples is also presented. Comparison with standard classification tools on video data of more than 4000 segments shows the potential of our approach in pattern classification.

The tasks of discretization and feature selection are frequently used to improve classification accuracy. We use discrete approximation of Bayes error rate to perform discretization on the features. The discretization procedure targets minimization of Bayes error rate within each partition. A class-pair discriminatory measure can be defined on discretized partitions which forms the basis of the feature selection algorithm. A small value of this measure for a class-pair indicates that the class-pair in consideration is confusing and the features which distinguish them well should be chosen first. A video classification problem on a large database is considered for showing the comparison of a classifier using our discretization and feature selection tasks with SVM, neural network classifier, decision trees and K-nearest neighbor classifier.

Many video related applications require identification of shot transitions and their type. Several algorithms for detecting both abrupt and gradual transitions exist. However, there is no integrated and robust framework for detecting transition types accurately. We show that combing the individual works of different researchers for identifying different transition types yields poor performance. An approach is presented using several algorithms working on patterns of effective average gradient, double chromatic difference, etc., such that the approach is robust to false detection or false-alarms. A comparison with standard approaches over 3 hours of experimental data on non-trivial video classes, such as commercials, MTV, sports, etc., shows the superiority of our approach.

In this paper, we develop the concept of descriptors based on perceptual-level motion features such as time-to-collision, shot transition and temporal motion and it is shown that by including them the representational level of the video classes is significantly enhanced, e.g. violence could be detected. The temporal context cues, which had been largely neglected by present content-based retrieval (CBR) systems, are integrated into the framework. A dynamic Bayesian framework for the CBR systems which can learn the temporal structure through the fusion of all the features is designed The experimental results for more than 4 hours of videos are presented for a number of key applications like sequence identifier, highlight extraction for sports, and detecting climax or violence.