Crowd Flow Segmentation based on Motion Vectors in H.264 Compressed Domain

In this work, we have explored the prospect of segmenting crowd flow in H. 264 compressed videos by merely using motion vectors. The motion vectors are extracted by partially decoding the corresponding video sequence in the H. 264 compressed domain. The region of interest ie., crowd flow region is extracted and the motion vectors that spans the region of interest is preprocessed and a collective representation of the motion vectors for the entire video is obtained. The obtained motion vectors for the corresponding video is then clustered by using EM algorithm. Finally, the clusters which converges to a single flow are merged together based on the bhattacharya distance measure between the histogram of the of the orientation of the motion vectors at the boundaries of the clusters. We had implemented our proposed approach on the complex crowd flow dataset provided by 1] and compared our results by using Jaccard measure. Since we are performing crowd flow segmentation in the compressed domain using only motion vectors, our proposed approach performs much faster compared to other pixel domain counterparts still retaining better accuracy.

Conservation Properties of the Trapezoidal Rule in Linear Time Domain Analysis of Acoustics and Structures

The trapezoidal rule, which is a special case of the Newmark family of algorithms, is one of the most widely used methods for transient hyperbolic problems. In this work, we show that this rule conserves linear and angular momenta and energy in the case of undamped linear elastodynamics problems, and an ``energy-like measure'' in the case of undamped acoustic problems. These conservation properties, thus, provide a rational basis for using this algorithm. In linear elastodynamics problems, variants of the trapezoidal rule that incorporate ``high-frequency'' dissipation are often used, since the higher frequencies, which are not approximated properly by the standard displacement-based approach, often result in unphysical behavior. Instead of modifying the trapezoidal algorithm, we propose using a hybrid finite element framework for constructing the stiffness matrix. Hybrid finite elements, which are based on a two-field variational formulation involving displacement and stresses, are known to approximate the eigenvalues much more accurately than the standard displacement-based approach, thereby either bypassing or reducing the need for high-frequency dissipation. We show this by means of several examples, where we compare the numerical solutions obtained using the displacement-based and hybrid approaches against analytical solutions.

Response to ``Comment on `Frequency-domain stimulated and spontaneous light emission signals at molecular junctions''' J. Chem. Phys. 142, 137101 (2015)]

In a recent work [U. Harbola, B. K. Agrawalla, and S. Mukamel, J. Chem. Phys. 141, 074107 (2014)], we have presented a superoperator (Liouville space) diagrammatic formulation of spontaneous and stimulated optical signals from current-carrying molecular junctions. We computed the diagrams that contribute to the spontaneous light emission SLE (fluorescence and Raman) signal using a diagrammatic method which clearly distinguishes between the Raman and the fluorescence contributions. We pointed out some discrepancies with the work of Galperin, Ratner and Nitzan (GRN) [M. Galperin, M. A. Ratner and, A. Nitzan, J. Chem. Phys. 130, 144109 (2009)]. In their response [M. Galperin, M. A. Ratner and A. Nitzan, “Comment on‘ Frequency-domain stimulated and spontaneous light emission signals at molecular junctions’” [J. Chem. Phys. 141, 074107 (2014)], J. Chem. Phys. 142, 137101 (2015)] to our work, GRN have argued that there are no differences in the choice of Raman diagrams in both works. Here we reply to their points and show where the differences exist.

A carboxy terminal domain of the L protein of rinderpest virus possesses RNA triphosphatase activity - The first enzyme in the viral mRNA capping pathway

The large protein L of negative-sense RNA viruses is a multifunctional protein involved in transcription and replication of genomic RNA. It also possesses enzymatic activities involved in capping and methylation of viral mRNAs. The pathway for mRNA capping followed by the L protein of the viruses in the Morbillivirus genus has not been established, although it has been speculated that these viruses may follow the unconventional capping pathway as has been shown for some viruses of Rhabdoviridae family. We had earlier shown that the large protein L of Rinderpest virus expressed as recombinant L-P complex in insect cells as well as the ribonucleoprotein complex from purified virus possesses RNA triphosphatase (RTPase) and guanylyltransferase activities, in addition to RNA dependent RNA polymerase activity. In the present work, we demonstrate that RTPase as well as nucleoside triphosphatase (NTPase) activities are exhibited by a subdomain of the L protein in the C terminal region (a.a. 1640 1840). The RTPase activity depends absolutely on a divalent cation, either magnesium or manganese. Both the RTPase and NTPase activities of the protein show dual metal specificity. Two mutant proteins having alanine mutations in the glutamic acid residues in motif-A of the RTPase domain did not show RTPase activity, while exhibiting reduced NTPase activity suggesting overlapping active sites for the two enzymatic functions. The RTPase and NTPase activities of the L subdomain resemble those of the Vaccinia capping enzyme D1 and the baculovirus LEF4 proteins. (C) 2015 Elsevier Inc. All rights reserved.

Compressed domain human action recognition in H.264/AVC video streams

This paper discusses a novel high-speed approach for human action recognition in H.264/AVC compressed domain. The proposed algorithm utilizes cues from quantization parameters and motion vectors extracted from the compressed video sequence for feature extraction and further classification using Support Vector Machines (SVM). The ultimate goal of the proposed work is to portray a much faster algorithm than pixel domain counterparts, with comparable accuracy, utilizing only the sparse information from compressed video. Partial decoding rules out the complexity of full decoding, and minimizes computational load and memory usage, which can result in reduced hardware utilization and faster recognition results. The proposed approach can handle illumination changes, scale, and appearance variations, and is robust to outdoor as well as indoor testing scenarios. We have evaluated the performance of the proposed method on two benchmark action datasets and achieved more than 85 % accuracy. The proposed algorithm classifies actions with speed (> 2,000 fps) approximately 100 times faster than existing state-of-the-art pixel-domain algorithms.

Study of wrist pulse signals using time domain spatial features

Blood travels throughout the body and thus its flow is modulated by changes in body condition. As a consequence, the wrist pulse signal contains important information about the status of the human body. In this work we have employed signal processing techniques to extract important information from these signals. Radial artery pulse pressure signals are acquired at wrist position noninvasively for several subjects for two cases of interest, viz. before and after exercise, and before and after lunch. Further analysis is performed by fitting a bi-modal Gaussian model to the data and extracting spatial features from the fit. The spatial features show statistically significant (p < 0.001) changes between the groups for both the cases, which indicates that they are effective in distinguishing the changes taking place due to exercise or food intake. Recursive cluster elimination based support vector machine classifier is used to classify between the groups. A high classification accuracy of 99.71% is achieved for the exercise case and 99.94% is achieved for the lunch case. This paper demonstrates the utility of certain spatial features in studying wrist pulse signals obtained under various experimental conditions. The ability of the spatial features in distinguishing changing body conditions can be potentially used for various healthcare applications. (C) 2015 Elsevier Ltd. All rights reserved.

Evidence for N-7 guanine methyl transferase activity encoded within the modular domain of RNA-dependent RNA polymerase L of a Morbillivirus

Post-transcriptional modification of viral mRNA is essential for the translation of viral proteins by cellular translation machinery. Due to the cytoplasmic replication of Paramyxoviruses, the viral-encoded RNA-dependent RNA polymerase (RdRP) is thought to possess all activities required for mRNA capping and methylation. In the present work, using partially purified recombinant RNA polymerase complex of rinderpest virus expressed in insect cells, we demonstrate the in vitro methylation of capped mRNA. Further, we show that a recombinant C-terminal fragment (1717-2183 aa) of L protein is capable of methylating capped mRNA, suggesting that the various post-transcriptional activities of the L protein are located in independently folding domains.

A survey on compressed domain video analysis techniques

Image and video analysis requires rich features that can characterize various aspects of visual information. These rich features are typically extracted from the pixel values of the images and videos, which require huge amount of computation and seldom useful for real-time analysis. On the contrary, the compressed domain analysis offers relevant information pertaining to the visual content in the form of transform coefficients, motion vectors, quantization steps, coded block patterns with minimal computational burden. The quantum of work done in compressed domain is relatively much less compared to pixel domain. This paper aims to survey various video analysis efforts published during the last decade across the spectrum of video compression standards. In this survey, we have included only the analysis part, excluding the processing aspect of compressed domain. This analysis spans through various computer vision applications such as moving object segmentation, human action recognition, indexing, retrieval, face detection, video classification and object tracking in compressed videos.

A coupled discriminative dictionary and transformation learning approach with applications to cross domain matching

Cross domain and cross-modal matching has many applications in the field of computer vision and pattern recognition. A few examples are heterogeneous face recognition, cross view action recognition, etc. This is a very challenging task since the data in two domains can differ significantly. In this work, we propose a coupled dictionary and transformation learning approach that models the relationship between the data in both domains. The approach learns a pair of transformation matrices that map the data in the two domains in such a manner that they share common sparse representations with respect to their own dictionaries in the transformed space. The dictionaries for the two domains are learnt in a coupled manner with an additional discriminative term to ensure improved recognition performance. The dictionaries and the transformation matrices are jointly updated in an iterative manner. The applicability of the proposed approach is illustrated by evaluating its performance on different challenging tasks: face recognition across pose, illumination and resolution, heterogeneous face recognition and cross view action recognition. Extensive experiments on five datasets namely, CMU-PIE, Multi-PIE, ChokePoint, HFB and IXMAS datasets and comparisons with several state-of-the-art approaches show the effectiveness of the proposed approach. (C) 2015 Elsevier B.V. All rights reserved.

A coupled discriminative dictionary and transformation learning approach with applications to cross domain matching

Cross domain and cross-modal matching has many applications in the field of computer vision and pattern recognition. A few examples are heterogeneous face recognition, cross view action recognition, etc. This is a very challenging task since the data in two domains can differ significantly. In this work, we propose a coupled dictionary and transformation learning approach that models the relationship between the data in both domains. The approach learns a pair of transformation matrices that map the data in the two domains in such a manner that they share common sparse representations with respect to their own dictionaries in the transformed space. The dictionaries for the two domains are learnt in a coupled manner with an additional discriminative term to ensure improved recognition performance. The dictionaries and the transformation matrices are jointly updated in an iterative manner. The applicability of the proposed approach is illustrated by evaluating its performance on different challenging tasks: face recognition across pose, illumination and resolution, heterogeneous face recognition and cross view action recognition. Extensive experiments on five datasets namely, CMU-PIE, Multi-PIE, ChokePoint, HFB and IXMAS datasets and comparisons with several state-of-the-art approaches show the effectiveness of the proposed approach. (C) 2015 Elsevier B.V. All rights reserved.

CROWD FLOW SEGMENTATION IN COMPRESSED DOMAIN USING CRF

Crowd flow segmentation is an important step in many video surveillance tasks. In this work, we propose an algorithm for segmenting flows in H.264 compressed videos in a completely unsupervised manner. Our algorithm works on motion vectors which can be obtained by partially decoding the compressed video without extracting any additional features. Our approach is based on modelling the motion vector field as a Conditional Random Field (CRF) and obtaining oriented motion segments by finding the optimal labelling which minimises the global energy of CRF. These oriented motion segments are recursively merged based on gradient across their boundaries to obtain the final flow segments. This work in compressed domain can be easily extended to pixel domain by substituting motion vectors with motion based features like optical flow. The proposed algorithm is experimentally evaluated on a standard crowd flow dataset and its superior performance in both accuracy and computational time are demonstrated through quantitative results.

ON THE UTILITY OF CANONICAL CORRELATION ANALYSIS FOR DOMAIN ADAPTATION IN MULTI-VIEW HEADPOSE ESTIMATION

The utility of canonical correlation analysis (CCA) for domain adaptation (DA) in the context of multi-view head pose estimation is examined in this work. We consider the three problems studied in 1], where different DA approaches are explored to transfer head pose-related knowledge from an extensively labeled source dataset to a sparsely labeled target set, whose attributes are vastly different from the source. CCA is found to benefit DA for all the three problems, and the use of a covariance profile-based diagonality score (DS) also improves classification performance with respect to a nearest neighbor (NN) classifier.

A short in-frame deletion in NTRK1 tyrosine kinase domain caused by a novel splice site mutation in a patient with congenital insensitivity to pain with anhidrosis

Background: Congenital insensitivity to pain with anhidrosis (CIPA) is a rare autosomal recessive genetic disease characterized by the lack of reaction to noxious stimuli and anhidrosis. It is caused by mutations in the NTRK1 gene, which encodes the high affinity tyrosine kinase receptor I for Neurotrophic Growth Factor (NGF). -- Case Presentation: We present the case of a female patient diagnosed with CIPA at the age of 8 months. The patient is currently 6 years old and her psychomotor development conforms to her age (RMN, SPECT and psychological study are in the range of normality). PCR amplification of DNA, followed by direct sequencing, was used to investigate the presence of NTRK1 gene mutations. Reverse transcriptase (RT)-PCR amplification of RNA, followed by cloning and sequencing of isolated RT-PCR products was used to characterize the effect of the mutations on NTRK1 mRNA splicing. The clinical diagnosis of CIPA was confirmed by the detection of two splice-site mutations in NTRK1, revealing that the patient was a compound heterozygote at this gene. One of these alterations, c.574+1G > A, is located at the splice donor site of intron 5. We also found a second mutation, c.2206-2 A > G, not previously reported in the literature, which is located at the splice acceptor site of intron 16. Each parent was confirmed to be a carrier for one of the mutations by DNA sequencing analysis. It has been proposed that the c.574+1G > A mutation would cause exon 5 skipping during NTRK1 mRNA splicing. We could confirm this prediction and, more importantly, we provide evidence that the novel c.2206-2A > G mutation also disrupts normal NTRK1 splicing, leading to the use of an alternative splice acceptor site within exon 17. As a consequence, this mutation would result in the production of a mutant NTRK1 protein with a seven aminoacid in-frame deletion in its tyrosine kinase domain. --Conclusions: We present the first description of a CIPA-associated NTRK1 mutation causing a short interstitial deletion in the tyrosine kinase domain of the receptor. The possible phenotypical implications of this mutation are discussed.

Towards multilingual domain module acquisition

Máster y Doctorado en Sistemas Informáticos Avanzados, Informatika Fakultatea - Facultad de Informática

Semi-automatic generation of learning domain modules for technology supported learning systems

In a time when Technology Supported Learning Systems are being widely used, there is a lack of tools that allows their development in an automatic or semi-automatic way. Technology Supported Learning Systems require an appropriate Domain Module, ie. the pedagogical representation of the domain to be mastered, in order to be effective. However, content authoring is a time and effort consuming task, therefore, efforts in automatising the Domain Module acquisition are necessary.Traditionally, textbooks have been used as the main mechanism to maintain and transmit the knowledge of a certain subject or domain. Textbooks have been authored by domain experts who have organised the contents in a means that facilitate understanding and learning, considering pedagogical issues.Given that textbooks are appropriate sources of information, they can be used to facilitate the development of the Domain Module allowing the identification of the topics to be mastered and the pedagogical relationships among them, as well as the extraction of Learning Objects, ie. meaningful fragments of the textbook with educational purpose.Consequently, in this work DOM-Sortze, a framework for the semi-automatic construction of Domain Modules from electronic textbooks, has been developed. DOM-Sortze uses NLP techniques, heuristic reasoning and ontologies to fulfill its work. DOM-Sortze has been designed and developed with the aim of automatising the development of the Domain Module, regardless of the subject, promoting the knowledge reuse and facilitating the collaboration of the users during the process.