Capture Dataset Research Articles

An Approach for Selecting the Most Explanatory Features for Facial Expression Recognition

The objective of this work is to analyze which features are most important in the recognition of facial expressions. To achieve this, we built a facial expression recognition system that learns from a controlled capture data set. The system uses different representations and combines them from a learned model. We studied the most important features by applying different feature extraction methods for facial expression representation, transforming each obtained representation into a sparse representation (SR) domain, and trained combination models to classify signals, using the extended Cohn–Kanade (CK+), BU-3DFE, and JAFFE data sets for validation. We compared 14 combination methods for 247 possible combinations of eight different feature spaces and obtained the most explanatory features for each facial expression. The results indicate that the LPQ (83%), HOG (82%), and RAW (82%) features are those features most able to improve the classification of expressions and that some features apply specifically to one expression (e.g., RAW for neutral, LPQ for angry and happy, LBP for disgust, and HOG for surprise).

Sequential Gesture Learning for Continuous Labanotation Generation Based on the Fusion of Graph Neural Networks

Labanotation is a symbolic recording system for human movements, and also a powerful tool for protecting and spreading folk dances and other performing arts. State-of-the-art automatic Labanotation uses end-to-end methods with sequence-based skeleton representation, which cannot capture the relationship between joints and bones in the skeleton for accurate descriptions of continuous lower limb movements such as dance steps. In this paper, we propose a novel double-stream fusion method of directed graph neural networks (DGNN), combined with connectionist temporal classification (CTC), namely DFGNN-CTC, for sequential fine-grained motion recognition, such as the Labanotation generation of unsegmented dance movement. First, we extract double-stream directed graph feature, employing an orientation-normalized directed acyclic graph (ON-DAG) and an orientation-normalized temporal directed acyclic graph (ON-TDAG), to jointly model spatiotemporal properties of movement recorded in motion capture data. Then, we design a CTC-based fusion-pooling module to fuse the spatial and temporal streams encoded by two DGNNs. It concatenates and fuses the two streams to generate discriminative descriptions of each time step, and concentrates them to make per-time-step predictions of Laban gesture type, from which the CTC searches the optimal Laban symbol sequence, corresponding to elemental motions composing the movement. In this way, the new method enables much finer discrimination for similar Laban gestures with subtle differences in spatial and temporal properties through joint contextual spatiotemporal modeling so that it achieves much superior performance in continuous Labanotation generation to existing methods, which only have single-stream analysis either spatially or temporally. The experiments on two Labanotation-labelled motion capture datasets demonstrate the effectiveness of the components in the proposed method and its superiority comparing with the state-of-the-art methods, especially for lower limb movements.

Dynamic Motion Matching: Design and Implementation of a Context-Aware Animation System for Games

Despite modern game systems adopting motion matching to retrieve an appropriate short motion clip from a database in real-time, existing methods struggle to support complex gaming scenes due to their inability to adapt live the motion retrieval based on the context. This paper presents the design and implementation of a context-aware character animation system, synthesizing realistic animations suitable for complex game scenes from a large-scale motion database. This system, called dynamic motion matching (DyMM), enables geometry and objects aware motion synthesis by introducing a two-phase context computation: an offline subspace decomposition of motion clips for creating a set of retrieval sub-spaces tailored to specific contexts and a subspace ensemble matching to compare relevant sub-features to determine the most appropriate motion clip. We also show the system architecture and implementation details applicable to a production-grade game engine. We verified the effectiveness of our method with industry-level motion data captured by professional game artists for multiple configurations and character controllers. The results of this study show that, by finding motion clips that comply well with the scene context, one can leverage large motion capture datasets to create practical systems that generate believable and controllable animations for games.

Generative model-enhanced human motion prediction.

The task of predicting human motion is complicated by the natural heterogeneity and compositionality of actions, necessitating robustness to distributional shifts as far as out‐of‐distribution (OoD). Here, we formulate a new OoD benchmark based on the Human3.6M and Carnegie Mellon University (CMU) motion capture datasets, and introduce a hybrid framework for hardening discriminative architectures to OoD failure by augmenting them with a generative model. When applied to current state‐of‐the‐art discriminative models, we show that the proposed approach improves OoD robustness without sacrificing in‐distribution performance, and can theoretically facilitate model interpretability. We suggest human motion predictors ought to be constructed with OoD challenges in mind, and provide an extensible general framework for hardening diverse discriminative architectures to extreme distributional shift. The code is available at: https://github.com/bouracha/OoDMotion.

Target sequence capture of Barnadesioideae (Compositae) demonstrates the utility of low coverage loci in phylogenomic analyses

Target sequence capture has emerged as a powerful method to sequence hundreds or thousands of genomic regions in a cost- and time-efficient approach. In most cases, however, targeted regions lack full sequence information for certain samples, due to taxonomic, laboratory, or stochastic factors. Loci lacking molecular data for a large number of samples are commonly excluded from downstream analyses, even though they may still contain valuable information. On the other hand, including data-poor loci may bias phylogenetic analyses. Here we use a target sequence capture dataset of an ecologically and taxonomically diverse group of spiny sunflowers (Asteraceae, or Compositae: Barnadesioideae) to test how the inclusion or exclusion of such data-poor loci affects phylogenetic inference. We investigate the sensitivity of concatenation and coalescent approaches to missing data with matrices of varying taxonomic completeness by filtering loci with different proportions of missing samples prior to data analysis. We find that missing data affect both the topology and branch support of the resulting phylogenies. The matrix containing all loci yielded the overall highest node support values, independently of the amount of missing nucleotides. These results provide empirical support to earlier suggestions based on single genes and data simulations that taxa with high amounts of missing data should not be readily dismissed as they can provide essential information for phylogenomic reconstruction.

Deep Learning-Based Speech Emotion Recognition Using Multi-Level Fusion of Concurrent Features

The detection and classification of emotional states in speech involves the analysis of audio signals and text transcriptions. There are complex relationships between the extracted features at different time intervals which ought to be analyzed to infer the emotions in speech. These relationships can be represented as spatial, temporal and semantic tendency features. In addition to emotional features that exist in each modality, the text modality consists of semantic and grammatical tendencies in the uttered sentences. Spatial and temporal features have been extracted sequentially in deep learning-based models using convolutional neural networks (CNN) followed by recurrent neural networks (RNN) which may not only be weak at the detection of the separate spatial-temporal feature representations but also the semantic tendencies in speech. In this paper, we propose a deep learning-based model named concurrent spatial-temporal and grammatical (CoSTGA) model that concurrently learns spatial, temporal and semantic representations in the local feature learning block (LFLB) which are fused as a latent vector to form an input to the global feature learning block (GFLB). We also investigate the performance of multi-level feature fusion compared to single-level fusion using the multi-level transformer encoder model (MLTED) that we also propose in this paper. The proposed CoSTGA model uses multi-level fusion first at the LFLB level where similar features (spatial or temporal) are separately extracted from a modality and secondly at the GFLB level where the spatial-temporal features are fused with the semantic tendency features. The proposed CoSTGA model uses a combination of dilated causal convolutions (DCC), bidirectional long short-term memory (BiLSTM), transformer encoders (TE), multi-head and self-attention mechanisms. Acoustic and lexical features were extracted from the interactive emotional dyadic motion capture (IEMOCAP) dataset. The proposed model achieves 75.50% and 75.82% of weighted and unweighted accuracy, 75.32% and 75.57% of recall and F1 score respectively. These results imply that concurrently learned spatial-temporal features with semantic tendencies learned in a multi-level approach improve the model’s effectiveness and robustness.

Speech Emotion Recognition Based on Self-Attention Weight Correction for Acoustic and Text Features

Speech emotion recognition (SER) is essential for understanding a speaker’s intention. Recently, some groups have attempted to improve SER performance using a bidirectional long short-term memory (BLSTM) to extract features from speech sequences and a self-attention mechanism to focus on the important parts of the speech sequences. SER also benefits from combining the information in speech with text, which can be accomplished automatically using an automatic speech recognizer (ASR), further improving its performance. However, ASR performance deteriorates in the presence of emotion in speech. Although there is a method to improve ASR performance in the presence of emotional speech, it requires the fine-tuning of ASR, which has a high computational cost and leads to the loss of cues important for determining the presence of emotion in speech segments, which can be helpful in SER. To solve these problems, we propose a BLSTM-and-self-attention-based SER method using self-attention weight correction (SAWC) with confidence measures. This method is applied to acoustic and text feature extractors in SER to adjust the importance weights of speech segments and words with a high possibility of ASR error. Our proposed SAWC reduces the importance of words with speech recognition error in the text feature while emphasizing the importance of speech segments containing these words in acoustic features. Our experimental results on the Interactive Emotional Dyadic Motion Capture (IEMOCAP) dataset reveal that our proposed method achieves a weighted average accuracy of 76.6%, outperforming other state-of-the-art methods. Furthermore, we investigated the behavior of our proposed SAWC in each of the feature extractors.

Kinematics-vis: A Visualization Tool for the Mathematics of Human Motion.

Kinematic analysis studies characterize human motion in healthy, pathologic, and rehabilitated subjects. These studies provide rich datasets which enable clinicians and researchers to understand disease progression, and the effects of surgical intervention and physical therapy. A variety of techniques - ranging from optical skin marker tracking to biplane fluoroscopy - are utilized to collect these datasets. The quantification and physical interpretation of these motion capture datasets (i.e. kinematic analysis) are predominantly conducted using Euler/Cardan angles (Diebel, 2006; Wikipedia, 2021; Zatsiorsky, 1998), whose output - time series of 3 angles - is presented in peer-reviewed manuscripts in 2D graphs (Ludewig et al., 2009). This rudimentary presentation, however, obscures the physical meaning behind Euler/Cardan analysis. As other authors have noted, the visualization of these angles enables meaningful discussion between biomechanics researchers and clinicians (Baker, 2011). A ubiquitous open-source method to concurrently visualize recorded human motion and the results of kinematic analysis is presently lacking in the biomechanics community. 'Kinematics-vis' is a JavaScript web application that visualizes kinematic analysis output resulting from motion capture studies.

Modelling and identification of characteristic kinematic features preceding freezing of gait with convolutional neural networks and layer-wise relevance propagation

BackgroundAlthough deep neural networks (DNNs) are showing state of the art performance in clinical gait analysis, they are considered to be black-box algorithms. In other words, there is a lack of direct understanding of a DNN’s ability to identify relevant features, hindering clinical acceptance. Interpretability methods have been developed to ameliorate this concern by providing a way to explain DNN predictions.MethodsThis paper proposes the use of an interpretability method to explain DNN decisions for classifying the movement that precedes freezing of gait (FOG), one of the most debilitating symptoms of Parkinson’s disease (PD). The proposed two-stage pipeline consists of (1) a convolutional neural network (CNN) to model the reduction of movement present before a FOG episode, and (2) layer-wise relevance propagation (LRP) to visualize the underlying features that the CNN perceives as important to model the pathology. The CNN was trained with the sagittal plane kinematics from a motion capture dataset of fourteen PD patients with FOG. The robustness of the model predictions and learned features was further assessed on fourteen PD patients without FOG and fourteen age-matched healthy controls.ResultsThe CNN proved highly accurate in modelling the movement that precedes FOG, with 86.8% of the strides being correctly identified. However, the CNN model was unable to model the movement for one of the seven patients that froze during the protocol. The LRP interpretability case study shows that (1) the kinematic features perceived as most relevant by the CNN are the reduced peak knee flexion and the fixed ankle dorsiflexion during the swing phase, (2) very little relevance for FOG is observed in the PD patients without FOG and the healthy control subjects, and (3) the poor predictive performance of one subject is attributed to the patient’s unique and severely flexed gait signature.ConclusionsThe proposed pipeline can aid clinicians in explaining DNN decisions in clinical gait analysis and aid machine learning practitioners in assessing the generalization of their models by ensuring that the predictions are based on meaningful kinematic features.

Descending 13 real world steps: A dataset and analysis of stair descent

Background:Stair descent analysis has been typically limited to laboratory staircases of 4 or 5 steps. To date there has been no report of gait parameters during unconstrained stair descent outside of the laboratory, and few motion capture datasets are publicly available. Research question:We aim to collect a dataset and perform gait analysis for stair descent outside of the laboratory. We aim to measure basic kinematic and kinetic gait parameters and foot placement behavior. Methods:We present a public stair descent dataset from 101 unimpaired participants aged 18–35 on an unconstrained 13-step staircase collected using wearable sensors. The dataset consists of kinematics (full-body joint angle and position), kinetics (plantar normal forces, acceleration), and foot placement for 30,609 steps. Results:We report the lower limb joint angle ranges (30° and 8° for hip flexion and extension, 85° and −11° for knee flexion and extension, and 31° and 28° for ankle dorsi- and plantar-flexion). The self-selected speed was 0.79 ± 0.16 m/s, with cycle duration of 0.97 ± 0.18 s. Mean foot overhang as a percentage of foot length was 17.07 ± 6.66 %, and we calculate that foot size explains only 6% of heel placement variation, but 79% of toe placement variation. We also find a minor but significant asymmetry between left and right maximum hip flexion angle, though all other measured parameters were symmetrical. Significance:This is the first quantitative observation of gait data from a large number (n = 101) of participants descending an unconstrained staircase outside of a laboratory. This study enables analysis of gait characteristics including self-selected walking speed and foot placement to better understand typical stair gait behavior. The dataset is a public resource for understanding typical stair descent.

Development of the Wheat Practical Haplotype Graph database as a resource for genotyping data storage and genotype imputation.

To improve the efficiency of high-density genotype data storage and imputation in bread wheat (Triticum aestivum L.), we applied the Practical Haplotype Graph (PHG) tool. The Wheat PHG database was built using whole-exome capture sequencing data from a diverse set of 65 wheat accessions. Population haplotypes were inferred for the reference genome intervals defined by the boundaries of the high-quality gene models. Missing genotypes in the inference panels, composed of wheat cultivars or recombinant inbred lines genotyped by exome capture, genotyping-by-sequencing (GBS), or whole-genome skim-seq sequencing approaches, were imputed using the Wheat PHG database. Though imputation accuracy varied depending on the method of sequencing and coverage depth, we found 92% imputation accuracy with 0.01× sequence coverage, which was slightly lower than the accuracy obtained using the 0.5× sequence coverage (96.6%). Compared to Beagle, on average, PHG imputation was ∼3.5% (P-value < 2 × 10−14) more accurate, and showed 27% higher accuracy at imputing a rare haplotype introgressed from a wild relative into wheat. We found reduced accuracy of imputation with independent 2× GBS data (88.6%), which increases to 89.2% with the inclusion of parental haplotypes in the database. The accuracy reduction with GBS is likely associated with the small overlap between GBS markers and the exome capture dataset, which was used for constructing PHG. The highest imputation accuracy was obtained with exome capture for the wheat D genome, which also showed the highest levels of linkage disequilibrium and proportion of identity-by-descent regions among accessions in the PHG database. We demonstrate that genetic mapping based on genotypes imputed using PHG identifies SNPs with a broader range of effect sizes that together explain a higher proportion of genetic variance for heading date and meiotic crossover rate compared to previous studies.

Extracting ‘legacy loci’ from an invertebrate sequence capture data set

AbstractSequence capture studies result in rich data sets comprising hundreds to thousands of targeted genomic regions that are superseding Sanger‐based data sets comprised of a few well‐known loci with historical uses in phylogenetics (‘legacy loci’). However, integrating sequence capture and Sanger‐based data sets is of interest as legacy loci can include different types of loci (e.g. mitochondrial and nuclear) across a potentially larger sample of species from past studies. Sequence capture data sets include nontargeted sequences, and there has been recent interest in extracting legacy loci from invertebrate data sets. Here, we use published legacy data from leaf‐footed bugs (Hemiptera: Coreoidea) to recover 15 mitochondrial and seven nuclear legacy loci from off‐target sequences in a sequence capture data set, explore approaches to improve legacy locus recovery, and combine these loci with sequence capture data for phylogenetic analysis. Two nuclear loci were determined to already be targeted by sequence capture baits. Most of the remaining loci were successfully recovered from off‐target sequences, but this recovery varied greatly. Additionally, complementing complete mitogenomes with additional reference mitochondrial sequences from a genetic depository did not offer improvement for most of our taxa; however, supplementing these reference sequences with extracted legacy loci offered ≥6% improvement across taxa for a given mitochondrial locus (negligible improvement for nuclear loci). Phylogenetic analysis of legacy and sequence capture data produced a topology generally congruent with recent studies, but support was lower. Thus, future studies may employ the approaches used in this study to integrate legacy data with newly generated sequence capture data sets without added expenses.

Unsupervised Feature Learning for Speech Emotion Recognition Based on Autoencoder

Speech signals contain abundant information on personal emotions, which plays an important part in the representation of human potential characteristics and expressions. However, the deficiency of emotion speech data affects the development of speech emotion recognition (SER), which also limits the promotion of recognition accuracy. Currently, the most effective approach is to make use of unsupervised feature learning techniques to extract speech features from available speech data and generate emotion classifiers with these features. In this paper, we proposed to implement autoencoders such as a denoising autoencoder (DAE) and an adversarial autoencoder (AAE) to extract the features from LibriSpeech for model pre-training, and then conducted experiments on the Interactive Emotional Dyadic Motion Capture (IEMOCAP) datasets for classification. Considering the imbalance of data distribution in IEMOCAP, we developed a novel data augmentation approach to optimize the overlap shift between consecutive segments and redesigned the data division. The best classification accuracy reached 78.67% (weighted accuracy, WA) and 76.89% (unweighted accuracy, UA) with AAE. Compared with state-of-the-art results to our knowledge (76.18% of WA and 76.36% of UA with the supervised learning method), we achieved a slight advantage. This suggests that using unsupervised learning benefits the development of SER and provides a new approach to eliminate the problem of data scarcity.

INFIMA leverages multi-omics model organism data to identify effector genes of human GWAS variants

Genome-wide association studies reveal many non-coding variants associated with complex traits. However, model organism studies largely remain as an untapped resource for unveiling the effector genes of non-coding variants. We develop INFIMA, Integrative Fine-Mapping, to pinpoint causal SNPs for diversity outbred (DO) mice eQTL by integrating founder mice multi-omics data including ATAC-seq, RNA-seq, footprinting, and in silico mutation analysis. We demonstrate INFIMA’s superior performance compared to alternatives with human and mouse chromatin conformation capture datasets. We apply INFIMA to identify novel effector genes for GWAS variants associated with diabetes. The results of the application are available at http://www.statlab.wisc.edu/shiny/INFIMA/.

Genetic Continuity of Bronze Age Ancestry with Increased Steppe-Related Ancestry in Late Iron Age Uzbekistan.

Although Uzbekistan and Central Asia are known for the well-studied Bronze Age civilization of the Bactria–Margiana Archaeological Complex (BMAC), the lesser-known Iron Age was also a dynamic period that resulted in increased interaction and admixture among different cultures from this region. To broaden our understanding of events that impacted the demography and population structure of this region, we generated 27 genome-wide single-nucleotide polymorphism capture data sets of Late Iron Age individuals around the Historical Kushan time period (∼2100–1500 BP) from three sites in South Uzbekistan. Overall, Bronze Age ancestry persists into the Iron Age in Uzbekistan, with no major replacements of populations with Steppe-related ancestry. However, these individuals suggest diverse ancestries related to Iranian farmers, Anatolian farmers, and Steppe herders, with a small amount of West European Hunter Gatherer, East Asian, and South Asian Hunter Gatherer ancestry as well. Genetic affinity toward the Late Bronze Age Steppe herders and a higher Steppe-related ancestry than that found in BMAC populations suggest an increased mobility and interaction of individuals from the Northern Steppe in a Southward direction. In addition, a decrease of Iranian and an increase of Anatolian farmer-like ancestry in Uzbekistan Iron Age individuals were observed compared with the BMAC populations from Uzbekistan. Thus, despite continuity from the Bronze Age, increased admixture played a major role in the shift from the Bronze to the Iron Age in southern Uzbekistan. This mixed ancestry is also observed in other parts of the Steppe and Central Asia, suggesting more widespread admixture among local populations.

Vector coherence imaging for enhancement of small omni-directional scatterers and suppression of geometric reflections

Ultrasonic imaging with full-matrix capture (FMC) datasets using the total focusing method (TFM) has been demonstrated to improve the detection, sizing and characterization of small defects in the area of Nondestructive Evaluation (NDE). Despite the general improvements in resolution and signal to noise ratio (SNR) offered by FMC/TFM, detection of small flaws remains challenging - particularly when flaws are in immediate proximity to the bounding surfaces of a test piece. In this paper, it is shown that by retaining only phase information during the application of the TFM algorithm, it is possible increase the image intensity of small omni-directional reflectors with respect to large specular reflectors (i.e. backwall/frontwall). The proposed algorithm, called Vector Coherence Imaging, is derived from the Vector Coherence Factor phase coherence metric and represents the amount of statistical (Shannon) information in favour of the presence of a coherent scatterer everywhere in the imaging domain. The VCI algorithm is shown to aid in detection of small defects located close to the boundaries of the specimen. Notably, VCI is seen to facilitate detection tip diffraction/scattering originating from surface connected fatigue cracks (less than 2 ultrasonic wavelengths in height) and early stage high temperature hydrogen attack (HTHA) damage (void nucleation and micro-fissuring).

Keypose synthesis from 3D motion capture data by using evolutionary clustering

SummaryMotion Capture datasets are captured as high‐frequency discrete samples using photogrammetric computer vision or various industrial geodetic measurement methods over the relevant model. Because Motion Capture data are inherently large‐datasets, expressing Motion Capture data without employing a motion data abstraction approach such as keypose is challenging. Keypose synthesis is a serious problem in many applications. Unfortunately, the local statistical features of Motion Capture data over time periods vary often, making it difficult to determine a key point summing the motion activation involved. Conventional clustering methods, such as Fuzzy C‐Means (FCM), that are sensitive to initial conditions, can be fitted to a local solution rather than producing a highly reliable keypose. Data clustering can be performed using evolutionary search methods without being limited to relatively local solutions. In this paper, the Motion Capture data were clustered using evolutionary search algorithms, and related keyposes were synthesized using the “minimum‐distance to cluster‐centers” principle. In Experiments section of this paper, motion data containing sportive movements were clustered by using evolutionary algorithms (i.e., Particle Swarm Optimization, Artificial Bee Colony, and Differential Search Algorithm) and classical clustering algorithms (i.e., Self‐Organizing Neural Network, FCM) to obtain related keyposes. The computed statistics exposed that evolutionary methods were more successful in obtaining keypose than classical methods.

HybPhaser: A workflow for the detection and phasing of hybrids in target capture data sets.

PREMISEHybrids contain divergent alleles that can confound phylogenetic analyses but can provide insights into reticulated evolution when identified and phased. We developed a workflow to detect hybrids in target capture data sets and phase reads into parental lineages using a similarity and phylogenetic framework.METHODSWe used Angiosperms353 target capture data for Nepenthes, including known hybrids to test the novel workflow. Reference mapping was used to assess heterozygous sites across the data set and to detect hybrid accessions and paralogous genes. Hybrid samples were phased by mapping reads to multiple references and sorting reads according to similarity. Phased accessions were included in the phylogenetic framework.RESULTSAll known Nepenthes hybrids and nine additional samples had high levels of heterozygous sites, had reads associated with multiple divergent clades, and were phased into accessions resembling divergent haplotypes. Phylogenetic analysis including phased accessions increased clade support and confirmed parental lineages of hybrids.DISCUSSIONHybPhaser provides a novel approach to detect and phase hybrids in target capture data sets, which can provide insights into reticulations by revealing origins of hybrids and reduce conflicting signal, leading to more robust phylogenetic analyses.

Using target sequence capture to improve the phylogenetic resolution of a rapid radiation in New Zealand Veronica.

Recent, rapid radiations present a challenge for phylogenetic reconstruction. Fast successive speciation events typically lead to low sequence divergence and poorly resolved relationships with standard phylogenetic markers. Target sequence capture of many independent nuclear loci has the potential to improve phylogenetic resolution for rapid radiations. Here we applied target sequence capture with 353 protein-coding genes (Angiosperms353 bait kit) to Veronica sect. Hebe (common name hebe) to determine its utility for improving the phylogenetic resolution of rapid radiations. Veronica section Hebe originated 5-10 million years ago in New Zealand, forming a monophyletic radiation of ca 130 extant species. We obtained approximately 150 kbp of 353 protein-coding exons and an additional 200 kbp of flanking noncoding sequences for each of 77 hebe and two outgroup species. When comparing coding, noncoding, and combined data sets, we found that the latter provided the best overall phylogenetic resolution. While some deep nodes in the radiation remained unresolved, our phylogeny provided broad and often improved support for subclades identified by both morphology and standard markers in previous studies. Gene-tree discordance was nonetheless widespread, indicating that additional methods are needed to disentangle fully the history of the radiation. Phylogenomic target capture data sets both increase phylogenetic signal and deliver new insights into the complex evolutionary history of rapid radiations as compared with traditional markers. Improving methods to resolve remaining discordance among loci from target sequence capture is now important to facilitate the further study of rapid radiations.

UNOC: Understanding Occlusion for Embodied Presence in Virtual Reality.

Tracking body and hand motions in 3D space is essential for social and self-presence in augmented and virtual environments. Unlike the popular 3D pose estimation setting, the problem is often formulated as egocentric tracking based on embodied perception (e.g., egocentric cameras, handheld sensors). In this article, we propose a new data-driven framework for egocentric body tracking, targeting challenges of omnipresent occlusions in optimization-based methods (e.g., inverse kinematics solvers). We first collect a large-scale motion capture dataset with both body and finger motions using optical markers and inertial sensors. This dataset focuses on social scenarios and captures ground truth poses under self-occlusions and body-hand interactions. We then simulate the occlusion patterns in head-mounted camera views on the captured ground truth using a ray casting algorithm and learn a deep neural network to infer the occluded body parts. Our experiments show that our method is able to generate high-fidelity embodied poses by applying the proposed method to the task of real-time egocentric body tracking, finger motion synthesis, and 3-point inverse kinematics.