Two Stage Transform Vector Quantization of LSFs for Wideband Speech Coding

We investigate the use of a two stage transform vector quantizer (TSTVQ) for coding of line spectral frequency (LSF) parameters in wideband speech coding. The first stage quantizer of TSTVQ, provides better matching of source distribution and the second stage quantizer provides additional coding gain through using an individual cluster specific decorrelating transform and variance normalization. Further coding gain is shown to be achieved by exploiting the slow time-varying nature of speech spectra and thus using inter-frame cluster continuity (ICC) property in the first stage of TSTVQ method. The proposed method saves 3-4 bits and reduces the computational complexity by 58-66%, compared to the traditional split vector quantizer (SVQ), but at the expense of 1.5-2.5 times of memory.

Human action recognition from video sequences

This PhD research has proposed new machine learning techniques to improve human action recognition based on local features. Several novel video representation and classification techniques have been proposed to increase the performance with lower computational complexity. The major contributions are the construction of new feature representation techniques, based on advanced machine learning techniques such as multiple instance dictionary learning, Latent Dirichlet Allocation (LDA) and Sparse coding. A Binary-tree based classification technique was also proposed to deal with large amounts of action categories. These techniques are not only improving the classification accuracy with constrained computational resources but are also robust to challenging environmental conditions. These developed techniques can be easily extended to a wide range of video applications to provide near real-time performance.

Recognition of nonstructural protein-peptides by cytotoxic t-lymphocytes raised against japanese encephalitis-virus

We have previously reported that Lyt(2+) cytotoxic T lymphocytes (CTL) can be raised against Japanese encephalitis virus (JEV) in BALB/c mice. In order to confirm the presence of H-2K(d)-restricted CTL and to examine their cross-recognition of West Wile virus (WNV), we tested the capacity of anti-JEV CTL to lyse uninfected syngeneic target cells that were pulsed with synthetic peptides. The sequence of the synthetic peptides was predicted based upon the H-2K(d) binding consensus motif. We show here that preincubation of uninfected syngeneic targets (P388D1) with JEV NS1- and NS3-derived peptides [NS1 (891-899) and NS3 (1804-1812)], but not with JEV NS5-derived peptide [NS5 (3370-3378)], partially sensitized them for lysis by polyclonal anti-JEV CTL. These results indicate the CTL recognition of NS1- and NS3-derived peptides of JEV.

Elliptical summary randomisation for sensor-based human activity recognition

Many conventional statistical machine learning al- gorithms generalise poorly if distribution bias ex- ists in the datasets. For example, distribution bias arises in the context of domain generalisation, where knowledge acquired from multiple source domains need to be used in a previously unseen target domains. We propose Elliptical Summary Randomisation (ESRand), an efficient domain generalisation approach that comprises of a randomised kernel and elliptical data summarisation. ESRand learns a domain interdependent projection to a la- tent subspace that minimises the existing biases to the data while maintaining the functional relationship between domains. In the latent subspace, ellipsoidal summaries replace the samples to enhance the generalisation by further removing bias and noise in the data. Moreover, the summarisation enables large-scale data processing by significantly reducing the size of the data. Through comprehensive analysis, we show that our subspace-based approach outperforms state-of-the-art results on several activity recognition benchmark datasets, while keeping the computational complexity significantly low.

Recognition of Interaction Interface Residues in Low-Resolution Structures of Protein Assemblies Solely from the Positions of C alpha Atoms

Background: The number of available structures of large multi-protein assemblies is quite small. Such structures provide phenomenal insights on the organization, mechanism of formation and functional properties of the assembly. Hence detailed analysis of such structures is highly rewarding. However, the common problem in such analyses is the low resolution of these structures. In the recent times a number of attempts that combine low resolution cryo-EM data with higher resolution structures determined using X-ray analysis or NMR or generated using comparative modeling have been reported. Even in such attempts the best result one arrives at is the very course idea about the assembly structure in terms of trace of the C alpha atoms which are modeled with modest accuracy. Methodology/Principal Findings: In this paper first we present an objective approach to identify potentially solvent exposed and buried residues solely from the position of C alpha atoms and amino acid sequence using residue type-dependent thresholds for accessible surface areas of C alpha. We extend the method further to recognize potential protein-protein interface residues. Conclusion/Significance: Our approach to identify buried and exposed residues solely from the positions of C alpha atoms resulted in an accuracy of 84%, sensitivity of 83-89% and specificity of 67-94% while recognition of interfacial residues corresponded to an accuracy of 94%, sensitivity of 70-96% and specificity of 58-94%. Interestingly, detailed analysis of cases of mismatch between recognition of interface residues from C alpha positions and all-atom models suggested that, recognition of interfacial residues using C alpha atoms only correspond better with intuitive notion of what is an interfacial residue. Our method should be useful in the objective analysis of structures of protein assemblies when positions of only C alpha positions are available as, for example, in the cases of integration of cryo-EM data and high resolution structures of the components of the assembly.

Recognition and alignment of homologous DNA sequences between minichromosomes and single-stranded DNA promoted by RecA protein

The incorporation of DNA into nucleosomes and higher-order forms of chromatin in vivo creates difficulties with respect to its accessibility for cellular functions such as transcription, replication, repair and recombination. To understand the role of chromatin structure in the process of homologous recombination, we have studied the interaction of nucleoprotein filaments, comprised of RecA protein and ssDNA, with minichromosomes. Using this paradigm, we have addressed how chromatin structure affects the search for homologous DNA sequences, and attempted to distinguish between two mutually exclusive models of DNA-DNA pairing mechanisms. Paradoxically, we found that the search for homologous sequences, as monitored by unwinding of homologous or heterologous duplex DNA, was facilitated by nucleosomes, with no discernible effect on homologous pairing. More importantly, unwinding of minichromosomes required the interaction of nucleoprotein filaments and led to the accumulation of circular duplex DNA sensitive to nuclease P1. Competition experiments indicated that chromatin templates and naked DNA served as equally efficient targets for homologous pairing. These and other findings suggest that nucleosomes do not impede but rather facilitate the search for homologous sequences and establish, in accordance with one proposed model, that unwinding of duplex DNA precedes alignment of homologous sequences at the level of chromatin. The potential application of this model to investigate the role of chromosomal proteins in the alignment of homologous sequences in the context of cellular recombination is considered.

CD4+ alpha beta T cell and gamma delta T cell responses to Mycobacterium tuberculosis. Similarities and differences in Ag recognition, cytotoxic effector function, and cytokine production.

CD4+ and gamma delta T cells are activated readily by Mycobacterium tuberculosis. To examine their role in the human immune response to M. tuberculosis, CD4+ and gamma delta T cells from healthy tuberculin-positive donor were studied for patterns of Ag recognition, cytotoxicity, and cytokine production in response to M. tuberculosis-infected mononuclear phagocytes. Both T cell subsets responded to intact M. tuberculosis and its cytosolic Ags. However, CD4+ and gamma delta T cells differed in the range of cytosolic Ags recognized: reactivity to a wide m.w. range of Ags for CD4+ T cells, and a restricted pattern for gamma delta T cells, with dominance of Ags of 10 to 15 kDa. Both T cell subsets were equally cytotoxic for M. tuberculosis-infected monocytes. Furthermore, both CD4+ and gamma delta T cells produced large amounts of IFN-gamma: mean pg/ml of IFN-gamma in supernatants was 2458 +/- 213 for CD4+ and 2349 +/- 245 for gamma delta T cells. By filter-spot ELISA (ELISPOT), the frequency of IFN-gamma-secreting gamma delta T cells was one-half of that of CD4+ T cells in response to M. tuberculosis, suggesting that gamma delta T cells on a per cell basis were more efficient producers of IFN-gamma than CD4+ T cells. In contrast, CD4+ T cells produced more IL-2 than gamma delta T cells, which correlated with diminished T cell proliferation of gamma delta T cells compared with CD4+ T cells. These results indicate that CD4+ and gamma delta T cell subsets have similar effector functions (cytotoxicity, IFN-gamma production) in response to M. tuberculosis-infected macrophages, despite differences in the Ags recognized, IL-2 production, and efficiency of IFN-gamma production.

Recognition, comorbidity, and outcome of DSM-IV bipolar I and II disorders in psychiatric care

This study is part of an ongoing collaborative bipolar research project, the Jorvi Bipolar Study (JoBS). The JoBS is run by the Department of Mental Health and Alcohol Research of the National Public Health Institute, Helsinki, and the Department of Psychiatry, Jorvi Hospital, Helsinki University Central Hospital (HUCH), Espoo, Finland. It is a prospective, naturalistic cohort study of secondary level care psychiatric in- and outpatients with a new episode of bipolar disorder (BD). The second report also included 269 major depressive disorder (MDD) patients from the Vantaa Depression Study (VDS). The VDS was carried out in collaboration with the Department of Psychiatry of the Peijas Medical Care District. Using the Mood Disorder Questionnaire (MDQ), all in- and outpatients at the Department of Psychiatry at Jorvi Hospital who currently had a possible new phase of DSM-IV BD were sought. Altogether, 1630 psychiatric patients were screened, and 490 were interviewed using a semistructured interview (SCID-I/P). The patients included in the cohort (n=191) had at intake a current phase of BD. The patients were evaluated at intake and at 6- and 18-month interviews. Based on this study, BD is poorly recognized even in psychiatric settings. Of the BD patients with acute worsening of illness, 39% had never been correctly diagnosed. The classic presentations of BD with hospitalizations, manic episodes, and psychotic symptoms lead clinicians to correct diagnosis of BD I in psychiatric care. Time of follow-up elapsed in psychiatric care, but none of the clinical features, seemed to explain correct diagnosis of BD II, suggesting reliance on cross- sectional presentation of illness. Even though BD II was clearly less often correctly diagnosed than BD I, few other differences between the two types of BD were detected. BD I and II patients appeared to differ little in terms of clinical picture or comorbidity, and the prevalence of psychiatric comorbidity was strongly related to the current illness phase in both types. At the same time, the difference in outcome was clear. BD II patients spent about 40% more time depressed than BD I patients. Patterns of psychiatric comorbidity of BD and MDD differed somewhat qualitatively. Overall, MDD patients were likely to have more anxiety disorders and cluster A personality disorders, and bipolar patients to have more cluster B personality disorders. The adverse consequences of missing or delayed diagnosis are potentially serious. Thus, these findings strongly support the value of screening for BD in psychiatric settings, especially among the major depressive patients. Nevertheless, the diagnosis must be based on a clinical interview and follow-up of mood. Comorbidity, present in 59% of bipolar patients in a current phase, needs concomitant evaluation, follow-up, and treatment. To improve outcome in BD, treatment of bipolar depression is a major challenge for clinicians.

An Insight to the Dynamics of Conserved Water Molecular Triad in IMPDH II (Human): Recognition of Cofactor and Substrate to Catalytic Arg 322

Inosine 5' monophosphate dehydrogenase (IMPDH II) is a key enzyme involved in the de novo biosynthesis pathway of purine nucleotides and is also considered to be an excellent target for cancer inhibitor design. The conserve R 322 residue (in human) is thought to play some role in the recognition of inhibitor and cofactor through the catalytic D 364 and N 303. The 15 ns simulation and the water dynamics of the three different PDB structures (1B3O, 1NF7, and 1NFB) of human IMPDH by CHARMM force field have clearly indicated the involvement of three conserved water molecules (W-L, W-M, and W-C) in the recognition of catalytic residues (R 322, D 364, and N 303) to inhibitor and cofactor. Both the guanidine nitrogen atoms (NH1 and NH 2) of the R 322 have anchored the di- and mono-nucleotide (cofactor and inhibitor) binding domains via the conserved W-C and W-L water molecules. Another conserved water molecule W-M seems to bridge the two domains including the R 322 and also the W-C and W-L through seven centers H-bonding coordination. The conserved water molecular triad (W-C - W-M - W-L) in the protein complex may thought to play some important role in the recognition of inhibitor and cofactor to the protein through R 322 residue.

Seeing and hearing speech, sounds, and signs: functional magnetic resonance studies on fluent and dyslexic readers

Speech has both auditory and visual components (heard speech sounds and seen articulatory gestures). During all perception, selective attention facilitates efficient information processing and enables concentration on high-priority stimuli. Auditory and visual sensory systems interact at multiple processing levels during speech perception and, further, the classical motor speech regions seem also to participate in speech perception. Auditory, visual, and motor-articulatory processes may thus work in parallel during speech perception, their use possibly depending on the information available and the individual characteristics of the observer. Because of their subtle speech perception difficulties possibly stemming from disturbances at elemental levels of sensory processing, dyslexic readers may rely more on motor-articulatory speech perception strategies than do fluent readers. This thesis aimed to investigate the neural mechanisms of speech perception and selective attention in fluent and dyslexic readers. We conducted four functional magnetic resonance imaging experiments, during which subjects perceived articulatory gestures, speech sounds, and other auditory and visual stimuli. Gradient echo-planar images depicting blood oxygenation level-dependent contrast were acquired during stimulus presentation to indirectly measure brain hemodynamic activation. Lip-reading activated the primary auditory cortex, and selective attention to visual speech gestures enhanced activity within the left secondary auditory cortex. Attention to non-speech sounds enhanced auditory cortex activity bilaterally; this effect showed modulation by sound presentation rate. A comparison between fluent and dyslexic readers' brain hemodynamic activity during audiovisual speech perception revealed stronger activation of predominantly motor speech areas in dyslexic readers during a contrast test that allowed exploration of the processing of phonetic features extracted from auditory and visual speech. The results show that visual speech perception modulates hemodynamic activity within auditory cortex areas once considered unimodal, and suggest that the left secondary auditory cortex specifically participates in extracting the linguistic content of seen articulatory gestures. They are strong evidence for the importance of attention as a modulator of auditory cortex function during both sound processing and visual speech perception, and point out the nature of attention as an interactive process (influenced by stimulus-driven effects). Further, they suggest heightened reliance on motor-articulatory and visual speech perception strategies among dyslexic readers, possibly compensating for their auditory speech perception difficulties.

GMM based Bayesian approach to speech enhancement in signal transform domain

Considering a general linear model of signal degradation, by modeling the probability density function (PDF) of the clean signal using a Gaussian mixture model (GMM) and additive noise by a Gaussian PDF, we derive the minimum mean square error (MMSE) estimator. The derived MMSE estimator is non-linear and the linear MMSE estimator is shown to be a special case. For speech signal corrupted by independent additive noise, by modeling the joint PDF of time-domain speech samples of a speech frame using a GMM, we propose a speech enhancement method based on the derived MMSE estimator. We also show that the same estimator can be used for transform-domain speech enhancement.

Learning Feature Trajectories Using Gabor Filter Bank for Human Activity Segmentation and Recognition

We describe a novel method for human activity segmentation and interpretation in surveillance applications based on Gabor filter-bank features. A complex human activity is modeled as a sequence of elementary human actions like walking, running, jogging, boxing, hand-waving etc. Since human silhouette can be modeled by a set of rectangles, the elementary human actions can be modeled as a sequence of a set of rectangles with different orientations and scales. The activity segmentation is based on Gabor filter-bank features and normalized spectral clustering. The feature trajectories of an action category are learnt from training example videos using dynamic time warping. The combined segmentation and the recognition processes are very efficient as both the algorithms share the same framework and Gabor features computed for the former can be used for the later. We have also proposed a simple shadow detection technique to extract good silhouette which is necessary for good accuracy of an action recognition technique.

Recognition of active and inactive catalytic triads: A template based approach

It Is well established that a sequence template along with the database is a powerful tool for identifying the biological function of proteins. Here, we describe a method for predicting the catalytic nature of certain proteins among the several protein structures deposited in the Protein Data Bank (PDB) For the present study, we considered a catalytic triad template (Ser-His-Asp) found in serine proteases We found that a geometrically optimized active site template can be used as a highly selective tool for differentiating an active protein among several inactive proteins, based on their Ser-His-Asp interactions. For any protein to be proteolytic in nature, the bond angle between Ser O-gamma-Ser H-gamma His N-epsilon 2 in the catalytic triad needs to be between 115 degrees and 140 degrees The hydrogen bond distance between Ser H-gamma His N-epsilon 2 is more flexible in nature and it varies from 2 0 angstrom to 27 angstrom while in the case of His H-delta 1 Asp O-delta 1, it is from 1.6 angstrom to 2.0 angstrom In terms of solvent accessibility, most of the active proteins lie in the range of 10-16 angstrom(2), which enables easy accessibility to the substrate These observations hold good for most catalytic triads and they can be employed to predict proteolytic nature of these catalytic triads (C) 2010 Elsevier B V All rights reserved.

Machine recognition of online handwritten Devanagari characters

In this paper, we describe a system for the automatic recognition of isolated handwritten Devanagari characters obtained by linearizing consonant conjuncts. Owing to the large number of characters and resulting demands on data acquisition, we use structural recognition techniques to reduce some characters to others. The residual characters are then classified using the subspace method. Finally the results of structural recognition and feature-based matching are mapped to give final output. The proposed system Ifs evaluated for the writer dependent scenario.