Facial Expression Space Research Articles

Interactive facial animation with deep neural networks

Creating realistic animations of human faces is still a challenging task in computer graphics. While computer graphics (CG) models capture much variability in a small parameter vector, they usually do not meet the necessary visual quality. This is due to the fact, that geometry‐based animation often does not allow fine‐grained deformations and fails in difficult areas (mouth, eyes) to produce realistic renderings. Image‐based animation techniques avoid these problems by using dynamic textures that capture details and small movements that are not explained by geometry. This comes at the cost of high‐memory requirements and limited flexibility in terms of animation because dynamic texture sequences need to be concatenated seamlessly, which is not always possible and prone to visual artefacts. In this study, the authors present a new hybrid animation framework that exploits recent advances in deep learning to provide an interactive animation engine that can be used via a simple and intuitive visualisation for facial expression editing. The authors describe an automatic pipeline to generate training sequences that consist of dynamic textures plus sequences of consistent three‐dimensional face models. Based on this data, they train a variational autoencoder to learn a low‐dimensional latent space of facial expressions that is used for interactive facial animation.

The semantic space for facial communication

ABSTRACTWe can learn a lot about someone by watching their facial expressions and body language. Harnessing these aspects of non‐verbal communication can lend artificial communication agents greater depth and realism but requires a sound understanding of the relationship between cognition and expressive behaviour. Here, we extend traditional word‐based methodology to use actual videos and then extract the semantic/cognitive space of facial expressions. We find that depending on the specific expressions used, either a four‐dimensional or a two‐dimensional space is needed to describe the variance in the stimuli. The shape and structure of the 4D and 2D spaces are related to each other and very stable to methodological changes. The results show that there is considerable variance between how different people express the same emotion. The recovered space can well capture the full range of facial communication and is very suitable for semantic‐driven facial animation. Copyright © 2014 John Wiley & Sons, Ltd.

Online modeling for realtime facial animation

We present a new algorithm for realtime face tracking on commodity RGB-D sensing devices. Our method requires no user-specific training or calibration, or any other form of manual assistance, thus enabling a range of new applications in performance-based facial animation and virtual interaction at the consumer level. The key novelty of our approach is an optimization algorithm that jointly solves for a detailed 3D expression model of the user and the corresponding dynamic tracking parameters. Realtime performance and robust computations are facilitated by a novel subspace parameterization of the dynamic facial expression space. We provide a detailed evaluation that shows that our approach significantly simplifies the performance capture workflow, while achieving accurate facial tracking for realtime applications.

Exploring Non‐Linear Relationship of Blendshape Facial Animation

AbstractHuman face is a complex biomechanical system and non‐linearity is a remarkable feature of facial expressions. However, in blendshape animation, facial expression space is linearized by regarding linear relationship between blending weights and deformed face geometry. This results in the loss of reality in facial animation. To synthesize more realistic facial animation, aforementioned relationship should be non‐linear to allow the greatest generality and fidelity of facial expressions. Unfortunately, few existing works pay attention to the topic about how to measure the non‐linear relationship. In this paper, we propose an optimization scheme that automatically explores the non‐linear relationship of blendshape facial animation from captured facial expressions. Experiments show that the explored non‐linear relationship is consistent with the non‐linearity of facial expressions soundly and is able to synthesize more realistic facial animation than the linear one.

Laying the foundations for an in-depth investigation of the whole space of facial expressions

Laying the foundations for an in-depth investigation of the whole space of facial expressions

A computational shape-based model of anger and sadness justifies a configural representation of faces

Research suggests that configural cues (second-order relations) play a major role in the representation and classification of face images; making faces a “special” class of objects, since object recognition seems to use different encoding mechanisms. It is less clear, however, how this representation emerges and whether this representation is also used in the recognition of facial expressions of emotion. In this paper, we show how configural cues emerge naturally from a classical analysis of shape in the recognition of anger and sadness. In particular our results suggest that at least two of the dimensions of the computational (cognitive) space of facial expressions of emotion correspond to pure configural changes. The first of these dimensions measures the distance between the eyebrows and the mouth, while the second is concerned with the height–width ratio of the face. Under this proposed model, becoming a face “expert” would mean to move from the generic shape representation to that based on configural cues. These results suggest that the recognition of facial expressions of emotion shares this expertise property with the other processes of face processing.

Animating cartoon faces by multi‐view drawings

AbstractIn this paper, we present a novel framework for creating cartoon facial animation from multi‐view hand‐drawn sketches. The input sketches are first employed to construct a base mesh model by using a hybrid sketch‐based method. The model is then deformed for each key viewpoint, yielding a set of models that closely match the corresponding sketches. We introduce a view‐dependent facial expression space defined by the key viewpoints and the basic emotions to generate various facial expressions viewed from arbitrary angles. The output facial animation conforms to the input sketches and maintains frame‐to‐frame correspondence. We demonstrate the potential of our approach through an easy‐to‐use system, where the animating of cartoon faces is automated once the user accomplishes sketching and configuration. Copyright © 2010 John Wiley & Sons, Ltd.

An emotion generation model based on random graphs

Emotions are an indispensable aspect in harmonious human computer interaction and artificial intelligence. In this paper, we present an algorithm to generate emotions and show them with 3D facial expression animation. The Pleasure-Arousal-Dominance (PAD) emotion space is used to define the affective elements of exterior stimuli and interior emotional states. A stochastic graphical model is designed to represent the relationships between the emotion, the mood and the personality. Finally, we build a mapping from the emotion space to the facial expression space with a competitive learning network. The facial expression is vividly shown by a 3D virtual human head. The experimental results demonstrate that our emotion generation model works effectively and meets the basic principle of human emotion generation.

Affective interaction based on person-independent facial expression space

This paper proposes a person-independent facial expression space (PIFES) to analyze and synthesize facial expressions based on supervised locality preserving projections (SLPP), which aligns different subjects and different intensities of facial expressions on one generalized expression manifold. Interactive curves of different patterns are generated according to the input facial expression image sequence, and target responsive expression images are synthesized for different emotions. In order to synthesize arbitrary expressions for a new person with natural details, a novel approach based on local geometry preserving between the input face image and the target expression image is proposed. Experimental results clearly demonstrate the efficiency of the proposed algorithm.

FCM 클러스터링을 이용한 표정공간의 단계적 가시화

본 논문은 사용자로 하여금 표정공간으로부터 일련의 표정들을 선택하게 함으로써 3차원 아바타의 표정을 제어할 수 있는 표정공간의 단계적 가시화 기법을 기술한다. 본 기법에 의한 시스템은 무표정 상태를 포함하여 11개의 서로 다른 모션들로 구성된 2400여개의 표정 프레임으로 2차원 표정공간을 구성하였으며, 3차원 아바타의 표정 제어는 사용자가 표정공간을 항해함으로서 수행되어진다. 그러나 표정공간에서는 과격한 표정 변화에서부터 세밀한 표정 변화까지 다양한 표정 제어를 수행할 수 있어야하기 때문에 단계적 가시화 기법이 필요하다. 표정공간을 단계적으로 가시화하기 위해서는 퍼지 클러스터링을 이용한다. 초기 단계에서는 11개의 클러스터 센터를 가지도록 클러스터링하고, 단계가 증가될 때 마다 클러스터 센터의 수를 두 배씩 증가시켜 표정들을 클러스터링한다. 이때 클러스터 센터와 표정공간에 분포된 표정들의 위치는 서로 다른 경우가 많기 때문에, 클러스터 센터에서 가장 가까운 표정상태를 찾아 클러스터 센터로 간주한다. 본 논문은 본 시스템이 어떤 효과가 있는지를 알기 위해 사용자들로 하여금 본 시스템을 사용하여 3차원 아바타의 단계적 표정 제어를 수행하게 하였으며, 그 결과를 평가한다. This paper presents a phased visualization method of facial expression space that enables the user to control facial expression of 3D avatars by select a sequence of facial frames from the facial expression space. Our system based on this method creates the 2D facial expression space from approximately 2400 facial expression frames, which is the set of neutral expression and 11 motions. The facial expression control of 3D avatars is carried out in realtime when users navigate through facial expression space. But because facial expression space can phased expression control from radical expressions to detail expressions. So this system need phased visualization method. To phased visualization the facial expression space, this paper use fuzzy clustering. In the beginning, the system creates 11 clusters from the space of 2400 facial expressions. Every time the level of phase increases, the system doubles the number of clusters. At this time, the positions of cluster center and expression of the expression space were not equal. So, we fix the shortest expression from cluster center for cluster center. We let users use the system to control phased facial expression of 3D avatar, and evaluate the system based on the results.

얼굴 표정공간에서 최적의 표정전이경로 자동 설정 방법

얼굴 표정공간에서 최적의 표정전이경로 자동 설정 방법

LLE 알고리즘을 사용한 얼굴 모션 데이터의 투영 및 실시간 표정제어

본 논문은 얼굴 모션 캡쳐 데이터를 재사용하여 실시간 표정 제어 및 표정 애니메이션을 생성하기 위한 방법론을 기술한다. 이 방법의 핵심요소는 얼굴 표정들을 정의할 수 있는 표정상태 표현법을 정하고, 이를 LLE 알고리즘에 적용하여 표정들을 적당한 공간에 분포시키는 방법론과, 이 공간을 사용하여 실시간 표정 애니메이션 생성 및 표정제어를 수행하기 위한 사용자 인터페이스 기법이다. 본 논문에서는 약 2400개의 얼굴 표정 프레임 데이터를 이용하여 공간을 생성하고, 애니메이터가 이 공간을 자유롭게 항해할 때, 항해경로 상에 위치한 얼굴 표정 프레임 데이터들이 연속적으로 선택되어 하나의 애니메이션이 생성되거나 표정제어가 가능하도록 하였다. 약 2400개의 얼굴 표정 프레임 데이터들을 직관적인 공간상에 분포하기 위해서는 얼굴 표정 프레임 데이터로부터 얼굴 표정상태를 표현할 필요가 있고, 이를 위해서는 임의의 두 마커 사이의 거리들로 구성된 거리행렬 벡터를 이용한다. 직관적인 공간에서의 데이터 배치는 얼굴 표정상태벡터들의 집합을 LLE 알고리즘에 적용하고, 이로부터 2차원 평면에 균일하게 분포하였다. 본 논문에서는 애니메이터로 하여금 사용자 인터페이스를 사용하여 실시간으로 표정 애니메이션을 생성하거나 표정제어를 수행하도록 하였으며, 그 결과를 평가한다. This paper describes methodology that enables animators to create the facial expression animations and to control the facial expressions in real-time by reusing motion capture datas. In order to achieve this, we fix a facial expression state expression method to express facial states based on facial motion data. In addition, by distributing facial expressions into intuitive space using LLE algorithm, it is possible to create the animations or to control the expressions in real-time from facial expression space using user interface. In this paper, approximately 2400 facial expression frames are used to generate facial expression space. In addition, by navigating facial expression space projected on the 2-dimensional plane, it is possible to create the animations or to control the expressions of 3-dimensional avatars in real-time by selecting a series of expressions from facial expression space. In order to distribute approximately 2400 facial expression data into intuitional space, there is need to represents the state of each expressions from facial expression frames. In order to achieve this, the distance matrix that presents the distances between pairs of feature points on the faces, is used. In order to distribute this datas, LLE algorithm is used for visualization in 2-dimensional plane. Animators are told to control facial expressions or to create animations when using the user interface of this system. This paper evaluates the results of the experiment.

Facial-Expression Affective Attributes and their Configural Correlates: Components and Categories

The present study investigates the perception of facial expressions of emotion, and explores the relation between the configural properties of expressions and their subjective attribution. Stimuli were a male and a female series of morphed facial expressions, interpolated between prototypes of seven emotions (happiness, sadness, fear, anger, surprise and disgust, and neutral) from Ekman and Friesen (1976). Topographical properties of the stimuli were quantified using the Facial Expression Measurement (FACEM) scheme. Perceived dissimilarities between the emotional expressions were elicited using a sorting procedure and processed with multidimensional scaling. Four dimensions were retained in the reconstructed facial-expression space, with positive and negative expressions opposed along D1, while the other three dimensions were interpreted as affective attributes distinguishing clusters of expressions categorized as "Surprise-Fear," "Anger," and "Disgust." Significant relationships were found between these affective attributes and objective facial measures of the stimuli. The findings support a componential explanatory scheme for expression processing, wherein each component of a facial stimulus conveys an affective value separable from its context, rather than a categorical-gestalt scheme. The findings further suggest that configural information is closely involved in the decoding of affective attributes of facial expressions. Configural measures are also suggested as a common ground for dimensional as well as categorical perception of emotional faces.

모션 데이터를 이용한 3차원 아바타 얼굴 표정 제어

This paper propose a method that controls facial expression of 3D avatar by having the user select a sequence of facial expressions in the space of facial expressions. And we setup its system. The space of expression is created from about 2400 frames consist of motion captured data of facial expressions. To represent the state of each expression, we use the distance matrix that represents the distances between pairs of feature points on the face. The set of distance matrices is used as the space of expressions. But this space is not such a space where one state can go to another state via the straight trajectory between them. We derive trajectories between two states from the captured set of expressions in an approximate manner. First, two states are regarded adjacent if the distance between their distance matrices is below a given threshold. Any two states are considered to have a trajectory between them If there is a sequence of adjacent states between them. It is assumed . that one states goes to another state via the shortest trajectory between them. The shortest trajectories are found by dynamic programming. The space of facial expressions, as the set of distance matrices, is multidimensional. Facial expression of 3D avatar Is controled in real time as the user navigates the space. To help this process, we visualized the space of expressions in 2D space by using the multidimensional scaling(MDS). To see how effective this system is, we had users control facial expressions of 3D avatar by using the system. As a result of that, users estimate that system is very useful to control facial expression of 3D avatar in real-time.

모션 데이터의 계층적 가시화에 의한 3차원 아바타의 표정 제어

본 논문은 사용자로 하여금 계층적 가시화 기법에 의한 표정들의 공간으로부터 일련의 표정을 선택하게 함으로써 3차원 아바타의 표정 제어기법을 기술한다. 본 시스템에서는 2,40P0여개의 표정 프레임을 이용하여 2차원 표정공간을 구성하였으며, 3차원 아바타의 표정 제어는 사용자가 표정공간을 항해하면서 수행한다. 그러나 표정의 수가 너무 많아 사용자가 항해를 하는데 어려움이 많기 때문에 계층적 가시화 기법이 필요하다. 표정공간을 계층적으로 분할하기 위해, 퍼지 클러스터링을 이용한다. 초기 단계에서는 2,400여개의 표정들을 이용하여 약 11개의 클러스터센터를 가지도록 클러스터링한다. 클러스터 센터들은 2차원 평면에 표시되며 후보 키 프레임으로 사용된다 사용자는 후보 키 프레임들 중에서 특정 키 프레임들을 선택하여 초기 항해경로를 생성한다. 사용자가 줌 인(이산적인 단계를 가지고 있음)을 하면 상세한 단계를 보기를 원한다는 의미이므로 줌 인된 단계에 적합한 표정 클러스터들을 생성한다. 단계가 증가될 때 마다 클러스터의 수를 두 배로 하고, 클러스터의 수만큼 표정들을 클러스터링한다. 사용자는 현재 단계에서 그전 단계의 항해경로를 따라 새로운 키 프레임(클러스터 센터)들을 선택하여 항해경로를 갱신한다. 줌 인을 최대로 한 마지막 단계에서 항해경로를 갱신하면 표정 제어 설정이 끝난다. 사용자는 언제든지 줌 아웃을 통해 그 전단계로 돌아가서 항해경로를 수정할 수 있다. 본 논문은 본 시스템이 어떤 효과가 있는지를 알기 위해 사용자들로 하여금 본 시스템을 사용하여 3차원 아바타의 표정 제어를 수행하게 하였으며, 그 결과를 평가한다. This paper presents a facial expression control method of 3D avatar that enables the user to select a sequence of facial frames from the facial expression space, whose level of details the user can select hierarchically. Our system creates the facial expression spare from about 2,400 captured facial frames. But because there are too many facial expressions to select from, the user faces difficulty in navigating the space. So, we visualize the space hierarchically. To partition the space into a hierarchy of subspaces, we use fuzzy clustering. In the beginning, the system creates about 11 clusters from the space of 2,400 facial expressions. The cluster centers are displayed on 2D screen and are used as candidate key frames for key frame animation. When the user zooms in (zoom is discrete), it means that the user wants to see mort details. So, the system creates more clusters for the new level of zoom-in. Every time the level of zoom-in increases, the system doubles the number of clusters. The user selects new key frames along the navigation path of the previous level. At the maximum zoom-in, the user completes facial expression control specification. At the maximum, the user can go back to previous level by zooming out, and update the navigation path. We let users use the system to control facial expression of 3D avatar, and evaluate the system based on the results.

Emotion Disc and Emotion Squares: Tools to Explore the Facial Expression Space

Abstract In the paper we present two novel interactive tools, Emotion Disc and Emotion Squares, to explore the facial expression space. They map navigation in a two‐dimensional circle, by the first tool, or in two two‐dimensional squares, by the second tool, to the high‐dimensional parameter space of facial expressions, by using a small number of predefined reference expressions. They can be used as exploration tools by researchers, or as control devices by end‐users to put expressions on the face of embodied agents or avatars in applications like games, telepresence and education.

FEATURE-BASED FACIAL EXPRESSION RECOGNITION: SENSITIVITY ANALYSIS AND EXPERIMENTS WITH A MULTILAYER PERCEPTRON

In this paper, we report our experiments on feature-based facial expression recognition within an architecture based on a two-layer perceptron. We investigate the use of two types of features extracted from face images: the geometric positions of a set of fiducial points on a face, and a set of multiscale and multiorientation Gabor wavelet coefficients at these points. They can be used either independently or jointly. The recognition performance with different types of features has been compared, which shows that Gabor wavelet coefficients are much more powerful than geometric positions. Furthermore, since the first layer of the perceptron actually performs a nonlinear reduction of the dimensionality of the feature space, we have also studied the desired number of hidden units, i.e. the appropriate dimension to represent a facial expression in order to achieve a good recognition rate. It turns out that five to seven hidden units are probably enough to represent the space of facial expressions. Then, we have investigated the importance of each individual fiducial point to facial expression recognition. Sensitivity analysis reveals that points on cheeks and on forehead carry little useful information. After discarding them, not only the computational efficiency increases, but also the generalization performance slightly improves. Finally, we have studied the significance of image scales. Experiments show that facial expression recognition is mainly a low frequency process, and a spatial resolution of 64 pixels × 64 pixels is probably enough.

Recognition of facial expressions using 2D DCT and neural network

The recognition of human facial expressions has been expected to have applications in various fields, such as psychology, engineering, and so forth, and many techniques have been proposed to date. However, most of the techniques are based on the motions of local feature-bearing blocks, such as the eyes and mouth, which are supposed to be closely associated with facial expressions. It is thus required to segment these blocks from the face image and to track their motions in real-time applications, resulting in high complexity. Furthermore, not all information on the motions of the facial expressions has been utilized. In this paper, a new recognition technique is proposed, which uses the 2D DCT of the entire facial image and a neural network. The segmentation of local feature-bearing blocks is not needed. The differences of the DCT coefficients in the lower frequency area between neutral and expression-bearing images are given to a neural network to realize a mapping into the facial expression space. The new technique has been applied to a database of normalized facial images of 60 persons, with images of 40 persons used for network training and the remaining images for testing. The recognition rate is 100% for the training images. A maximum recognition rate of 95% is achieved for the testing images. © 1999 Scripta Technica, Electron Comm Jpn Pt 3, 82(7): 1–11, 1999

20-1 FACSに基づく顔表情空間における表情記述と合成

20-1 FACSに基づく顔表情空間における表情記述と合成

Multidimensional scaling of hierarchical sorting data applied to facial expressions

A new algorithm for multidimensional scaling analysis of sorting data and hierarchical‐sorting data is tested by applying it to facial expressions of emotion. We construct maps in “facial expression space” for two sets of still photographs: the I‐FEEL series (expressions displayed spontaneously by infants and young children), and a subset of the Lightfoot series (posed expressions, all from one actress). The analysis avoids potential artefacts by fitting a map directly to the subject's judgments, rather than transforming the data into a matrix of estimated dissimilarities as an intermediate step. The results for both stimulus sets display an improvement in the extent to which they agree with existing maps. Some points emerge about the limitations of sorting data and the need for caution when interpreting MDS configurations derived from them.