Data-driven modeling of human motions is ubiquitous in computer graphics and computer vision applications, such as synthesizing realistic motions or recognizing actions. Recent research has shown that such problems can be approached by learning a natural motion manifold using deep learning on a large amount data, to address the shortcomings of traditional data-driven approaches. However, previous deep learning methods can be sub-optimal for two reasons. First, the skeletal information has not been fully utilized for feature extraction. Unlike images, it is difficult to define spatial proximity in skeletal motions in the way that deep networks can be applied for feature extraction. Second, motion is time-series data with strong multi-modal temporal correlations between frames. On the one hand, a frame could be followed by several candidate frames leading to different motions; on the other hand, long-range dependencies exist where a number of frames in the beginning correlate to a number of frames later. Ineffective temporal modeling would either under-estimate the multi-modality and variance, resulting in featureless mean motion or over-estimate them resulting in jittery motions, which is a major source of visual artifacts. In this paper, we propose a new deep network to tackle these challenges by creating a natural motion manifold that is versatile for many applications. The network has a new spatial component for feature extraction. It is also equipped with a new batch prediction model that predicts a large number of frames at once, such that long-term temporally-based objective functions can be employed to correctly learn the motion multi-modality and variances. With our system, long-duration motions can be predicted/synthesized using an open-loop setup where the motion retains the dynamics accurately. It can also be used for denoising corrupted motions and synthesizing new motions with given control signals. We demonstrate that our system can create superior results comparing to existing work in multiple applications.

No abstract

Biocompatible, biodegradable, and luminescent nano material can be used as an alternative bioimaging agent for early cancer diagnosis, which is crucial to achieve successful treatment. Hydroxyapatite (HAP) nanocyrstals have good biocompatibility and biodegradability, and can be used as an excellent host for luminescent rare earth elements. In this study, based on the energy transfer from Gd(3+) to Eu(3+), the luminescence enhanced imaging agent of Eu/Gd codoping HAP (HAP:Eu/Gd) nanocrystals are obtained via coprecipitation with plate-like shape and no change in crystal phase composition. The luminescence can be much elevated (up to about 120%) with a nonlinear increase versus Gd doping content, which is due to the energy transfer ((6)PJ of Gd(3+) → (5)HJ of Eu(3+)) under 273 nm and the possible combination effect of the cooperative upconversion and the successive energy transfer under 394 nm, respectively. Results demonstrate that the biocompatible HAP:Eu/Gd nanocrystals can successfully perform cell labeling and in vivo imaging. The intracellular HAP:Eu/Gd nanocrystals display good biodegradability with a cumulative degradation of about 65% after 72 h. This biocompatible, biodegradable, and luminescence enhanced HAP:Eu/Gd nanocrystal has the potential to act as a fluorescent imaging agent in vitro and in vivo.

Abstract-We propose a new semantic-level crowd evaluation metric in this paper. Crowd simulation has been an active and important area for several decades. However, only recently has there been an increased focus on evaluating the fidelity of the results with respect to real-world situations. The focus to date has been on analyzing the properties of low-level features such as pedestrian trajectories, or global features such as crowd densities. We propose the first approach based on finding semantic information represented by latent Path Patterns in both real and simulated data in order to analyze and compare them. Unsupervised clustering by non-parametric Bayesian inference is used to learn the patterns, which themselves provide a rich visualization of the crowd behavior. To this end, we present a new Stochastic Variational Dual Hierarchical Dirichlet Process (SV-DHDP) model. The fidelity of the patterns is computed with respect to a reference, thus allowing the outputs of different algorithms to be compared with each other and/or with real data accordingly. Detailed evaluations and comparisons with existing metrics show that our method is a good alternative for comparing crowd data at a different level and also works with more types of data, holds fewer assumptions and is more robust to noise.

Pleckstrin homology (PH) domains can recruit proteins to membranes by recognition of phosphatidylinositol phosphate (PIP) lipids. Several family members are linked to diseases including cancer. We report the systematic simulation of the interactions of 100 mammalian PH domains with PIP-containing membranes. The observed PIP interaction hotspots recapitulate crystallographic binding sites and reveal a number of insights: (i) The β1 and β2 strands and their connecting loop constitute the primary PIP interaction site but are typically supplemented by interactions at the β3-β4 and β5-β6 loops; (ii) we reveal exceptional cases such as the Exoc8 PH domain; (iii) PH domains adopt different membrane-bound orientations and induce clustering of anionic lipids; and (iv) beyond family-level insights, our dataset sheds new light on individual PH domains, e.g., by providing molecular detail of secondary PIP binding sites. This work provides a global view of PH domain/membrane association involving multivalent association with anionic lipids.

Background: It is clinically important to develop innovative techniques that can accurately measure blood pressures (BP) automatically. Objectives: This study aimed to present and evaluate a novel automatic BP measurement method based on deep learning method, and to confirm the effects on measured BPs of the position and contact pressure of stethoscope. Methods: 30 healthy subjects were recruited. 9 BP measurements (from three different stethoscope contact pressures and three repeats) were performed on each subject. The convolutional neural network (CNN) was designed and trained to identify the Korotkoff sounds at a beat-by-beat level. Next, a mapping algorithm was developed to relate the identified Korotkoff beats to the corresponding cuff pressures for systolic and diastolic BP (SBP and DBP) determinations. Its performance was evaluated by investigating the effects of the position and contact pressure of stethoscope on measured BPs in comparison with reference manual auscultatory method. Results: The overall measurement errors of the proposed method were 1.4±2.4 mmHg for SBP and 3.3±2.9 mmHg for DBP from all the measurements. In addition, the method demonstrated that there were small SBP differences between the 2 stethoscope positions, respectively at the 3 stethoscope contact pressures, and that DBP from the stethoscope under the cuff was significantly lower than that from outside the cuff by 2.0 mmHg (P < 0.01). Conclusion: Our findings suggested that the deep learning based method was an effective technique to measure BP, and could be developed further to replace the current oscillometric based automatic blood pressure measurement method.

Electroencephalograph (EEG) has been widely applied for brain-computer interface (BCI) which enables paralyzed people to directly communicate with and control of external devices, due to its portability, high temporal resolution, ease of use and low cost. Of various EEG paradigms, steady-state visual evoked potential (SSVEP)-based BCI system which uses multiple visual stimuli (such as LEDs or boxes on a computer screen) flickering at different frequencies has been widely explored in the past decades due to its fast communication rate and high signalto-noise ratio. In this paper, we review the current research in SSVEP-based BCI, focusing on the data analytics that enables continuous, accurate detection of SSVEPs and thus high information transfer rate. The main technical challenges, including signal pre-processing, spectrum analysis, signal decomposition, spatial filtering in particular canonical correlation analysis and its variations, and classification techniques are described in this paper. Research challenges and opportunities in spontaneous brain activities, mental fatigue, transfer learning as well as hybrid BCI are also discussed.

No abstract