A morphable model for the synthesis of 3D faces

Although emotions were, for a long time, considered undesirable for rational behaviour, there is now evidence in Neuroscience and Psychology that shows emotions to be an important factor in problem solving capabilities and intelligence in general [8]. As a result, a strong new field is emerging in computer science: Affective Computing, i.e. “computing that relates to, arises from or deliberately influences emotions” [21].

Affective interactions : Toward a new generation of computer interfaces?

Microsoft Kinect 虛擬復健系統設計

A novel hybrid bidirectional unidirectional LSTM network for dynamic hand gesture recognition with Leap Motion

In many cases, visitors come to a museum in small groups. In such cases, the visitors’ social context has an impact on their museum visit experience. Knowing the social context may allow a system to provide socially aware services to the visitors. Evidence of the social context can be gained from observing/monitoring the visitors’ social behavior. However, automatic identification of a social context requires, on the one hand, identifying typical social behavior patterns and, on the other, using relevant sensors that measure various signals and reason about them to detect the visitors’ social behavior. We present such typical social behavior patterns of visitor pairs, identified by observations, and then the instrumentation, detection process, reasoning, and analysis of measured signals that enable us to detect the visitors’ social behavior. Simple sensors’ data, such as proximity to other visitors, proximity to museum points of interest, and visitor orientation are used to detect social synchronization, attention to the social companion, and interest in museum exhibits. The presented approach may allow future research to offer adaptive services to museum visitors based on their social context to support their group visit experience better.

Automatic Detection of Social Behavior of Museum Visitor Pairs

In this paper, human facial emotions are detected through normalized minimal feature vectors using semi-supervised Twin Support Vector Machine (TWSVM) learning. In this study, face detection and tracking are carried out using the Constrained Local Model (CLM), which has 66 entire feature vectors. Based on Facial Animation Parameter’s (FAPs) definition, entire feature vectors are those things that visibly affect human emotion. This paper proposes the 13 minimal feature vectors that have high variance among the entire feature vectors are sufficient to identify the six basic emotions. Using the Max & Min and Z-normalization technique, two types of normalized minimal feature vectors are formed. The novelty of this study is methodological in that the normalized data of minimal feature vectors fed as input to the semi-supervised multi-class TWSVM classifier to classify the human emotions is a new contribution. The macro facial expression datasets are used by a standard database and several real-time datasets. 10-fold and hold out cross-validation is applied with the cross-database (combining standard and real-time). In the experimental result, using ‘One vs One’ and ‘One vs All’ multi-class techniques with 3 kernel functions produce a 36 trained model of each emotion and their validation parameters are calculated. The overall accuracy achieved for 10-fold cross-validation is 93.42 ± 3.25% and hold out cross-validation is 92.05 ± 3.79%. The overall performance (Precision, Recall, F1-score, Error rate and Computation Time) of the proposed model was also calculated. The performance of the proposed model and existing methods were compared and results indicate them to be more reliable than existing models.

Detecting facial emotions using normalized minimal feature vectors and semi-supervised twin support vector machines classifier

Particle filtering is known as a robust approach for motion tracking by vision, at the cost of heavy computation in a high dimensional pose space. In this work, we describe a number of heuristics that we demonstrate to jointly improve robustness and real-time for motion capture. 3D human motion capture by monocular vision without markers can be achieved in realtime by registering a 3D articulated model on a video. First, we search the high-dimensional space of 3D poses by generating new hypotheses (or particles) with equivalent 2D projection by kinematic flipping. Second, we use a semi-deterministic particle prediction based on local optimization. Third, we deterministi-cally resample the probability distribution for a more efficient selection of particles. Particles (or poses) are evaluated using a match cost function and penalized with a Gaussian probability pose distribution learned off-line. In order to achieve real-time, measurement step is parallelized on GPU using the OpenCL API. We present experimental results demonstrating robust real-time 3D motion capture with a consumer computer and webcam.

/pdf/real-time-particle-filtering-with-heuristics-for-3d-motion-3yol2fx4tn.pdf

Real-time particle filtering with heuristics for 3D motion capture by monocular vision

Discriminative feature embedding is about vital appreciation within the research area of deep face identification. During this paper, we would suggest a remaining attention based convolutional neural network (ResNet) because differs from facial characteristic implanting, who objectives in conformity with locate outdoors the long-range dependencies regarding rear images through reducing the knowledge redundancy amongst channels and as specialize of the important informative factors on spatial feature maps (SFM). More specifically, the proposed interest module consists about self channel attention (SCA) block then self spatial attention (SSA) barrier who adaptively aggregates the characteristic maps within each duct and spatial domains after find outdoors the inter-channel affinity form yet consequently the interspatial kindred matrix, and then matrix multiplications are conducted because a refined yet robust back feature. With the sight module we proposed, we wish make grade convolutional neural networks (CNNs), like ResNet-50 have greater discriminative government for deep face recognition. The experiments on labelled faces of SMIC-HS, SMIC-NIR, SMIC-VIS, CASME II and CASME I are show that our developed ResNet shape constantly outperforms naive CNNs and achieved the state-of-the-art performance. The proposed work on these micro-expression datasets yields better results with a state-of-the-art accuracy of quite 98% in each dataset. For verification purposes, the videos collected real-time manually from different gender people were tested and an accuracy score of 99.88% and an F1_score of 99.88% was achieved.

Residual attention network for deep face recognition using micro-expression image analysis

Our goal is to capture style from real human motion so it can be rendered with a virtual agent that represents this human user. We used expressivity parameters to describe motion style. As a first contribution, we propose an approach to estimate a subset of expressivity parameters defined in the literature (namely spatial extent and temporal extent) from captured motion trajectories. Second, we capture the expressivity of real users and then output it to the Greta engine that animates a virtual agent representing the user. We experimentally demonstrate that expressivity can be another clue for identifiable virtual clones of real humans.

https://hal.archives-ouvertes.fr/hal-00724449/document

Virtually Cloning Real Human with Motion Style

This paper aims to finding happiness in human face with minimal feature vectors. In this system, the face detection and tracking are carried out by Constrained Local Model (CLM). Using CLM grid node, the entire and minimal feature vector displacement is obtained through extracted features. The feature vector displacements are computed in multi-classes of TwinSupport Vector Machines (TWSVM) classifier to evaluate the happiness. In training and testing phases, the following databases are used such as MMI database, Cohn-Kanade (CK), Extended-CK, Mahnob-Laughter and also Real Time data. Also, this paper compares the Supervised Support Vector Machines and Unsupervised Twin Support Vector Machines classifier with cross datavalidation. Using the normalization of Min-max and Znorm technique, the overall accuracy of finding happiness are computed as 86.29% and 83.79% respectively.

/pdf/detecting-happiness-in-human-face-using-unsupervised-twin-4qzs0z0zm1.pdf

Manoj Kumar Rajagopal

Papers

Detecting facial emotions using normalized minimal feature vectors and semi-supervised twin support vector machines classifier

Real-time particle filtering with heuristics for 3D motion capture by monocular vision

Residual attention network for deep face recognition using micro-expression image analysis

Virtually Cloning Real Human with Motion Style

Detecting Happiness in Human Face using Unsupervised Twin-Support Vector Machines