50175 research outputs found
Sort by
Unsupervised feature learning of human actions as trajectories in pose embedding manifold
No full textAn unsupervised human action modeling framework can provide useful pose-sequence representation, which can be utilized in a variety of pose analysis applications. In this work we propose a novel temporal pose-sequence modeling framework, which can embed the dynamics of 3D human-skeleton joints to a continuous latent space in an efficient manner. In contrast to end-to-end framework explored by previous works, we disentangle the task of individual pose representation learning from the task of learning actions as a trajectory in pose embedding space. In order to realize a continuous pose embedding manifold with improved reconstructions, we propose an unsupervised, manifold learning procedure named Encoder GAN, (or EnGAN). Further we use the pose embeddings generated by EnGAN to model human actions using a bidirectional RNN auto-encoder architecture, PoseRNN. We introduce first-order gradient loss to explicitly enforce temporal regularity in the predicted motion sequence. A hierarchical feature fusion technique is also investigated for simultaneous modeling of local skeleton joints along with global pose variations. We demonstrate state-of-the-art transfer-ability of the learned representation against other supervisedly and unsupervisedly learned motion embeddings for the task of fine-grained action recognition on SBU interaction dataset. Further, we show the qualitative strengths of the proposed framework by visualizing skeleton pose reconstructions and interpolations in pose-embedding space, and low dimensional principal component projections of the reconstructed pose trajectories. © 2019 IEEE
Phase Retrieval - A Deconvolution Perspective
No full textPhase retrieval finds applications in various optical imaging modalities such as X-ray crystallography, holography, frequency-domain optical-coherence tomography, etc. The sensors used in optical imaging can measure only the magnitudes of incoming wavefronts and the phase information is not measured directly. This necessitates developing appropriate phase retrieval algorithms to reconstruct the object as phase contains most of the structural information. The phase retrieval problem naturally arises in the Fourier imaging context, where the measurement is the Fourier magnitude/intensity spectrum. Reconstruction from the Fourier intensity results in the autocorrelation and not the signal. We therefore address the equivalent problem of signal retrieval from the autocorrelation. Since the signal autocorrelation can be expressed as a convolution of the signal with its flipped version, we propose to solve the phase retrieval problem within a deconvolution framework. We consider a non-convex cost in two vector variables, the signal and its flipped version. An alternating minimization (Alt. Min.) strategy is employed to arrive at an optimal estimate of the signal, given the autocorrelation. Due to non-convexity of the cost function, the accuracy of the estimation is critically dependent on the initialization. We establish that the Alt. Min. iterates ensure that the cost is nonincreasing. For the specific case of causal, delta-dominant signals, the proposed framework results in exact reconstruction with an all zero-phase initialization. We shall also consider the effect of random initialization on the estimation accuracy. © 2018 APSIPA organization
High-speed waveguide integrated silicon photodetector on a SiN-SOI platform for short reach datacom
No full textWe present a waveguide integrated high-speed Si photodetector integrated with a silicon nitride (SiN) waveguide on an silicon-on-insulator (SOI) platform for short reach data communication in a 850 nm wavelength band. We demonstrate a waveguide couple Si pin photodetector responsivity of 0.44 A/W at 25 V bias. The frequency response of the photodetector is evaluated by the coupling of a femtosecond laser source through an SiN grating coupler of the integrated photodetector. We estimate a 3 dB bandwidth of 14 GHz at 20 V bias which, to the best of our knowledge, is the highest reported bandwidth for a waveguide integrated Si photodetector. We also present detailed optoelectronic DC and AC characterization of the fabricated devices. The demonstrated integrated photodetector could enable an integrated solution for scaling of short reach data communication and connectivity. (C) 2019 Optical Society of Americ
Dirichlet Latent Variable Model: A Dynamic Model Based on Dirichlet Prior for Audio Processing
No full textWe propose a dynamic latent variable model for learning latent bases from time varying, non-negative data. We take a probabilistic approach to modeling the temporal dependence in data by introducing a dynamic Dirichlet prior-a Dirichlet distribution with dynamic parameters. This new distribution allows us to assure non-negativity and avoid intractability when sequential updates are performed (otherwise encountered in using Dirichlet prior). We refer to the proposed model as the Dirichlet latent variable model (DLVM). We develop an expectation maximization algorithm for the proposed model, and also derive a maximum a posteriori estimate of the parameters. Furthermore, we connect the proposed DLVM to two popular latent basis learning methods- probabilistic latent component analysis (PLCA) and non-negative matrix factorization (NMF). We show that 1) PLCA is a special case of our DLVM, and 2) DLVM can be interpreted as a dynamic version of NMF. The usefulness of DLVM is demonstrated for three audio processing applications-speaker source separation, denoising, and bandwidth expansion. To this end, a new algorithm for source separation is also proposed. Through extensive experiments on benchmark databases, we show that the proposed model outperforms several relevant existing methods in all three applications
Evidence of a Nodal Line in the Superconducting Gap Symmetry of Noncentrosymmetric ThCoC2
No full textThe newly discovered noncentrosymmetric superconductor ThCoC2 exhibits numerous types of unconventional behavior in the field dependent heat capacity data. Here we present the first measurement of the gap symmetry of ThCoC2 by muon spin rotation and relaxation (mu SR) measurements. The temperature dependence of the magnetic penetration depth measured using the transverse field mu SR experiment reveals the evidence of a nodal pairing symmetry. To understand this finding, we carry out calculations of the superconducting pairing eigenvalue and eigenfunction (pairing symmetry) due to the spin-fluctuation mechanism by directly implementing the ab initio band structures. We find that the system possesses a single Fermi surface with considerable three dimensionality and a strong nesting along the k(z) direction. Such nesting promotes a superconducting state with a cos k(z)-like pairing symmetry with a prominent nodal line on the k(z) = +/-pi/2 plane. The result agrees well with the experimental data
MODIFIED BERGMAN MINIMAL MODEL FOR GLUCOSE-INSULIN DYNAMICS AND ESTIMATION OF MODEL PARAMETERS FOR INDIAN POPULATION
No full textDevelopment and characterization of supramolecular calcitonin assembly and assessment of its interactions with the bone remodelling process
No full textOsteoporosis is the most common metabolic bone disease, which poses an immense socio-economic burden on the society. Human calcitonin, though safe, is not considered as a therapeutic option because of its high tendency to self-associate to form amyloid fibrils thereby affecting its potency. To circumvent this issue we harnessed the inherent capacity of aggregation and developed an assemblage of human calcitonin monomers, Supramolecular Calcitonin Assembly (SCA-I)], which releases biologically active calcitonin monomers in a sustained manner for a period of at least three weeks. AFM and FT-IR analysis showed that SCA-I is amorphous aggregates of calcitonin monomers. Both SCA-I and monomer released from it demonstrated superior anti-osteoclast activity and proteolytic stability in-vitro. SCA-I upon single injection significantly improved bone formation markers and reduced bone resorption markers in ovariectomized (OVX) rat model of postmenopausal osteoporosis. Micro-CT analysis revealed that calcitonin released from SCA-I exhibits its beneficial effect on cortical bone more profoundly compared to trabecular bone. This study demonstrates that SCA-I is more effective compared to the human calcitonin monomers on osteoclasts and has site-specific effect on bone in a model of post-menopausal osteoporosis. This approach opens up an innovative way to use and study the function of human calcitonin
Pathways linked by hydrogen bonds with redox-dependent breaks implicated in electron transfer in human cytochrome c protein
No full textPathways of hydrogen-bond-linked peptide units, polar side chains of the amino acid residues and buried water molecules have been traced in human cytochrome c protein. These connect heme-Fe to the surface through axially coordinated Met80-S and His18-N on the two sides of the heme plate. Oxygen atoms of the heme-propionate side chain and of the internal invariant water molecules form hydrogen bonds in connecting these pathways. With 28 out of the 37 amino acid residues being in the conserved list, these pathways are likely to be common in the highly conserved cytochrome proteins. Selective breaks appear in hydrogen bonds on the Hisl8 side in the oxidized form and on the Met80 side in the reduced form consequent to the accompanying structural changes consistent with a regulatory role. These changes are defined by phi, psi angles of the backbone and dihedral angles of the side chains, between the redox states. The pathways are identical in both the redox forms. They are suitable for intramolecular atom-to-atom electron transfer with hydrogen bond now experimentally found to transfer electrons better than covalent sigma-bond, hitherto used for making the paths
Foetal Acoustic Simulator
No full textFoetal phonocardiography is the passive recording of foetal heart sounds. It is a valuable alternative to existing expensive technologies like foetal electrocardiography, cardiotochography etc. for ensuring the wellbeing of the foetus. To test and validate foetal heart monitoring system based on phonocardiography, a foetal acoustic simulator model was developed which simulates acoustical states of foetus such as bradycardia, tachycardia, etc. during pregnancy. The generated foetal heart sounds and noises are played from the speaker and recorded using electronic stethoscope from the surface of the water filled balloon. The acquired signals from foetal acoustic simulator are then analyzed to determine whether the signals resemble that of the real foetal phonocardiogram recordings
Expansion of Preisach density in magnetic hysteresis using general basis functions
No full textThe Preisach model of hysteresis has two parts: a geometrical staircase and a density or weighting function. In typical applications, the underlying density function of hysteresis operators is estimated through partial derivatives of first order reversal curves, or a priori assumed to obey simple functional forms like Gaussian, Lorenzian etc. Here we take a more agnostic and empirical approach, and expand the density in a general form using the spectrum of the Laplace operator on a bounded triangular domain. Transforming the input to the same bounded domain, we have a nonlinear parameter fitting problem. We fit parameters to our own magnetic hysteresis data directly using complex waveforms, for both soft and hard loops. For hard loops, if the Preisach density is to be kept strictly nonnegative (as is usual), a nonlinear transformation of the Preisach output is needed. Additionally, the Preisach density is consists of a single hump for soft loops and three distinct humps for hard loops. Our fitted density function, based on a general expansion, contains several coefficients, but subsequent simulation is quick. The main contribution of this paper is a direct demonstration of fitting the density function without making a priori assumptions about the functional form. (c) 2018 Elsevier Inc. All rights reserved