Real Human Face Research Articles

Improved AdaBoost Algorithm for Face Detection and Its Application

Face detection plays a crucial role in developing human-robot interaction (HRI) for Intelligent Educational Robot (IER) to recognize users or speakers. In this paper, we introduce an intelligent vision algorithm that is able to detect human face from complex scene and filter out all the non-face but face-like images. The human face is detected in real time using the approach called AdaBoost-based Haar-Cascade Classifier[1-2], and the real human face detection is improved to implement from single-face detection to multi-face detection. Furthermore, variable head pose is taken into account, such as pitch, roll, yaw, etc. The proposed robot vision algorithm for human detection is tested to be effective and robust through real-time experiments.

Generating various composite human faces from real 3D facial images

Generating large human crowds of distinguishable individuals is one of the challenges in the gaming industry. When the scene contains many characters, it becomes impracticable to create all the individual characters manually. However, the requirement for the different appearances of individuals in a crowd, namely their faces, is now in greater demand. Therefore, this paper describes our solution to the automatic generation of human faces that are created as a composite of facial parts of 3D scans of real human faces. However, the user has the possibility to further adjust the composite by designing replacements, leading to a desired appearance. The final composite can be exported and attached to a given avatar. To evaluate the usability of our solution, we performed two case studies. The conducted perception study performed with 104 participants aimed to confirm the decreasing human ability to recognize morphologically modified faces. The morphological study focused on the quantification of the extent of facial modifications. Both studies were performed by domain experts from psychology and anthropology.

How Realistic Should Avatars Be?

Abstract. Increased interaction with characters in games and online necessitates a better understanding of how different characteristics of these agents impact media users. This paper investigates a possible neurological underpinning for a common research finding – namely, that animated characters designed to be comparatively more human, more real, and more similar to the people they represent elicit more positive self-reported evaluations. The goal of this study was to examine the extent to which these results might be due to differential processing of character features in brain networks recruited for face recognition. There is some evidence that parts of the face network may be specifically tuned for real human faces. An experiment was conducted where participants viewed photographs of faces of actual agents (humans and animals) or colored drawings of matched agents (cartoon humans and animals). Using functional magnetic resonance imaging (fMRI) to measure blood oxygen-level dependent (BOLD) activation in the whole brain and specifically in the face network, we investigated the variation in patterns of activation with human and animal faces that were more or less real. The results were consistent with previous reports that the core regions of the face network are sensitive to the humanness of faces. However, our results extended previous work by showing that regions of the core and extended regions of the face network – and some regions outside the network – were sensitive to realism, but only realism of human faces.

3D-Model-Based Video Analysis for Computer Generated Faces Identification

Modern computer graphics technologies brought realism in computer-generated characters, making them achieve truly natural appearance. Besides traditional virtual reality applications such as avatars, games, or cinema, these synthetic characters may be used to generate realistic fakes, which may lead to improper use of the technology. This fact raises the demand for advanced tools able to discriminate real and artificial human faces in digital media. In this paper, we propose a method to distinguish between computer generated and natural faces by modeling and evaluating their dynamic behavior. Because of a 3D-model-based video analysis, the proposed technique allows identifying synthetic characters by detecting their more limited variability over time. Experimental results demonstrate the effectiveness of the proposed approach also on very challenging and realistic video sequences.

Group sparse feature selection on local learning based clustering

Feature selection plays an important role in many machine learning applications. By extracting meaningful features and eliminating both redundancies and noises, it effectively improves the accuracy and efficiency of the learning algorithm. In this paper, an unsupervised feature selection method called GSFS-llc (group sparse feature selection on local learning based clustering) is proposed. GSFS-llc first retrieves the cluster structure in a dataset using LLC and then selects important features that best preserve the cluster structure by L2,1-norm regularized regression. By combining group sparsity regularization with locality based learning algorithm, GSFS-llc leads a more robust result than other methods and selects features that respect the underlying geometric structure in the dataset. Extensive experimental results over real world handwritten digits, human faces, voices and object images datasets demonstrate the superiority of the GSFS-llc algorithm.

Statistical Analysis of the Characteristic Features of Standard Image for Racial and Ethnic Identity of a Person

The paper provides statistical analysis of the characteristic features used to determine racial and ethnic identity of a person basing on the images. Standard characteristic features of the persons of each ethnic group have been set accordingly. Characteristic features of random image have been compared with the characteristic features of standard image to determine belonging of a person to this or that racial and ethnic group. Comparing the characteristic features of the random image with the characteristic features of the standard image of each ethnic group, it is important to define the reliability of the interdependence between them. Note that, characteristic features of standard image were determined basing on the characteristic features of a real human face belonging to each ethnic group. The dependence was defined among the characteristic features of racial and ethnic standards set up according to the images with the characteristic features of a human face. A Fischer criterion was used to determine the dependence.

Artificial faces are harder to remember

Observers interact with artificial faces in a range of different settings and in many cases must remember and identify computer-generated faces. In general, however, most adults have heavily biased experience favoring real faces over synthetic faces. It is well known that face recognition abilities are affected by experience such that faces belonging to “out-groups” defined by race or age are more poorly remembered and harder to discriminate from one another than faces belonging to the “in-group”. Here, we examine the extent to which artificial faces form an “out-group” in this sense when other perceptual categories are matched. We rendered synthetic faces using photographs of real human faces and compared performance in a memory task and a discrimination task across real and artificial versions of the same faces. We found that real faces were easier to remember, but only slightly more discriminable than artificial faces. Artificial faces were also equally susceptible to the well-known face inversion effect, suggesting that while these patterns are still processed by the human visual system in a face-like manner, artificial appearance does compromise the efficiency of face processing.

Dogs Can Discriminate Emotional Expressions of Human Faces

The question of whether animals have emotions and respond to the emotional expressions of others has become a focus of research in the last decade [1-9]. However, to date, no study has convincingly shown that animals discriminate between emotional expressions of heterospecifics, excluding the possibility that they respond to simple cues. Here, we show that dogs use the emotion of a heterospecific as a discriminative cue. After learning to discriminate between happy and angry human faces in 15 picture pairs, whereby for one group only the upper halves of the faces were shown and for the other group only the lower halves of the faces were shown, dogs were tested with four types of probe trials: (1) the same half of the faces as in the training but of novel faces, (2) the other half of the faces used in training, (3) the other half of novel faces, and (4) the left half of the faces used in training. We found that dogs for which the happy faces were rewarded learned the discrimination more quickly than dogs for which the angry faces were rewarded. This would be predicted if the dogs recognized an angry face as an aversive stimulus. Furthermore, the dogs performed significantly above chance level in all four probe conditions and thus transferred the training contingency to novel stimuli that shared with the training set only the emotional expression as a distinguishing feature. We conclude that the dogs used their memories of real emotional human faces to accomplish the discrimination task.

3-D Reconstruction of Human Face Using the Derivative Moiré Topography

A new 3-D reconstruction algorithm for the human face is proposed using the derivative Moir<TEX>$\acute{e}$</TEX> topography which ensures fast and robust reconstruction even for rough surfaces. The Moir<TEX>$\acute{e}$</TEX> interference fringe pattern is initially obtained through the projection Moir<TEX>$\acute{e}$</TEX> topography based on phase shifting, and then differentiated to provide a full unwrapped phase map for a human face. <TEX>$2{\pi}$</TEX> ambiguity, which has been a chronically unsolved problem with Moir<TEX>$\acute{e}$</TEX> topography, is successfully surmounted by differentiating the Moir<TEX>$\acute{e}$</TEX> fringe patterns both in x- and y-directions when the object is located in the x-y plane. A real human face is used for verifying the proposed derivative Moir<TEX>$\acute{e}$</TEX> topography. A human face of 4 different phase-shifted images taken in the fixed plane is almost fully reconstructed in 3-D format in 0.1 mm lateral resolution.

Exaggerating Facial Expressions: A Way to Intensify Emotion or a Way to the Uncanny Valley?

Exaggeration of facial expressions is used in animation and robotics to intensify emotions. However, modifying a human-like face can lead to an unsettling outcome. This phenomenon is known as uncanny valley. The goal of this study was to identify the realism level and magnitude of facial expression that produce the maximum amount of emotional intensity and the minimum amount of perceived strangeness. We studied the perceived intensity of emotion and perceived strangeness of faces with varying levels of realism (from schematic to photorealistic) and magnitude of facial expressions (from neutral to extremely exaggerated). We found that less realistic faces required more exaggeration to reach the emotional intensity of a real human face. While there is a range of emotional intensity that can be expressed by real human faces (from neutral to full intensity), we found that the same range of emotional intensity could be expressed by artificial faces when exaggeration was used. However, attempts to express emotional intensities outside this range using exaggeration led to strange-looking faces at all levels of realism.

A Critical Perspective of Information Technology (IT) Industry Growth in India

It is very much evident that the growth agenda being prescribed for the sustainable development of many emerging economies like India has a consistent reliance on the abnormality factor which is very popularly identified with the services sector in general and Information Technology sector in particular. The mushrooming growth rates and boom factors are no doubt, phenomenal in nature with respect to software sector in recent times and has drastically changed the economic landscape of India, in a way, providing new mantra for the growth of our economy. What makes it inevitable to depend on the inconsistencies and abnormality that form the major trends of information technology (IT) industry that penetrated deep into the nook and corners of the country? The most popular theory that caught the imagination of both entrepreneurs and government is that IT and IT Enabled Services (ITES) is the growth engine of economy, capable of meeting the foreign exchange reserves demand and people purchasing power at one stroke. At the same time, by systematically developing the inflow of foreign investments, the IT industry leads to the infrastructural advancement in various forms. Behind the conceiving of Software Technology Parks (STPs), IT Special Economic Zones (SEZs) as a major tool to trigger the economic growth rates, one has to look at the changing configurations of the industry in the light of many human-related issues and implications. As such, the article tries to recap the penetration of IT industry in India from a critical perspective by supporting the idea of adding a real human face to its further development in the country. Also, the basic issues like employee satisfaction and commitment get highlighted as part of narrating the growth story of IT industry in India.

Computing Extremal Teichmüller Map of Multiply-Connected Domains Via Beltrami Holomorphic Flow

A numerical method for computing the extremal Teichmuller map between multiply-connected domains is presented. Given two multiply-connected domains, there exists a unique Teichmuller map (T-Map) between them minimizing the conformality distortion. The extremal T-Map can be considered as the `most conformal' map between multiply-connected domains. In this paper, we propose an iterative algorithm to compute the extremal T-Map using the Beltrami holomorphic flow (BHF). The BHF procedure iteratively adjusts the initial map based on a sequence of Beltrami coefficients, which are complex-valued functions defined on the source domain. It produces a sequence of quasi-conformal maps, which converges to the T-Map minimizing the conformality distortion. We test our method on synthetic data together with real human face data. Results show that our algorithm computes the extremal T-Map between two multiply-connected domains of the same topology accurately and efficiently.

Categorization difficulty is associated with negative evaluation in the “uncanny valley” phenomenon

AbstractHuman observers often experience strongly negative impressions of human‐like objects falling within a particular range of visual similarity to real humans (the “uncanny valley” phenomenon). We hypothesized that negative impressions in the uncanny valley phenomenon are related to a difficulty in object categorization. We produced stimulus images by morphing two of each of real, stuffed, and cartoon human face images (Experiment 1). Observers were asked to categorize each of these images as either category and evaluated the likability of the image. The results revealed that the longest latency, the highest ambiguity in categorization, and the lowest likability score co‐occurred at consistent morphing percentages. Similar results were obtained even when we employed stimulus images that were created by morphing two of each of real, stuffed, and cartoon dog images (Experiment 2). However, the effect of categorization difficulty on evaluation was weak when two real human faces were morphed (Experiment 3). These results suggest that the difficulty in categorizing an object as either of two dissimilar categories is linked to negative evaluation regardless of whether the object is human‐related or not.

Physical face cloning

We propose a complete process for designing, simulating, and fabricating synthetic skin for an animatronics character that mimics the face of a given subject and its expressions. The process starts with measuring the elastic properties of a material used to manufacture synthetic soft tissue. Given these measurements we use physics-based simulation to predict the behavior of a face when it is driven by the underlying robotic actuation. Next, we capture 3D facial expressions for a given target subject. As the key component of our process, we present a novel optimization scheme that determines the shape of the synthetic skin as well as the actuation parameters that provide the best match to the target expressions. We demonstrate this computational skin design by physically cloning a real human face onto an animatronics figure.

Heterogeneous Specular and Diffuse 3-D Surface Approximation for Face Recognition Across Pose

This paper proposes a novel heterogeneous specular and diffuse (HSD) 3-D surface approximation which considers spatial variability of specular and diffuse reflections in face modelling and recognition. Traditional 3-D face modelling and recognition methods constrain human faces with either the Lambertian assumption or the homogeneity assumption, resulting in suboptimal shape and texture models. The proposed HSD approach allows both specular and diffuse reflectance coefficients to vary spatially to better accommodate surface properties of real human faces. From a small number of face images of a person under different lighting conditions, 3-D shape and surface reflectivity property are estimated using a localized stochastic optimization method. The resultant personalized 3-D face model is used to render novel gallery views under different poses for recognition across pose. The proposed approach is evaluated on both synthetic and real face datasets and benchmarked against the state-of-the-art approaches. Experimental results demonstrated that it can achieve a higher level of performances in modelling accuracy, algorithm reliability, and recognition accuracy, which suggests that face modelling and recognition beyond the Lambertian and homogeneity assumptions is a feasible and better solution towards pose-invariant face recognition.

Real Time Face Recognition Using AdaBoost Improved Fast PCA Algorithm

This paper presents an automated system for human face recognition in a real time background world for a large homemade dataset of persons face. The task is very difficult as the real time background subtraction in an image is still a challenge. Addition to this there is a huge variation in human face image in terms of size, pose and expression. The system proposed collapses most of this variance. To detect real time human face AdaBoost with Haar cascade is used and a simple fast PCA and LDA is used to recognize the faces detected. The matched face is then used to mark attendance in the laboratory, in our case. This biometric system is a real time attendance system based on the human face recognition with a simple and fast algorithms and gaining a high accuracy rate..

The development of the uncanny valley in infants

When adults view very realistic humanoid robots or computer avatars they often exhibit an aversion to them. This phenomenon, known as the "uncanny valley," is assumed to be evolutionary in origin, perhaps tapping into modules for disgust or attractiveness that detect violations of our normal expectations regarding social signals. Here, we test an alternative hypothesis that the uncanny valley is developmental in origin and, thus, that specific early experience with real human faces leads to its eventual emergence. To test this idea, we measured visual preferences in response to all possible pairs of a human face, realistic avatar face, and an unrealistic avatar face in groups of 6-, 8-, 10-, and 12-month-old infants. Consistent with the developmental hypothesis, we found that the uncanny valley effect emerges at 12 months of age suggesting that perceptual experience with real human faces is critical to its emergence.

The Cadaver on the Cover

In Reply: From Michelle Paff: While Dr. Champney's opinion that we owe respect to cadavers is actually my own also, I feel he misunderstands a significant fact about my drawing: The drawing does not depict the real face of the cadaver. When I created the drawing, the cadaver's face was deformed from the embalming process and from lying facedown on the dissecting table during the first few weeks of dissection. I had to use a substantial amount of creativity and imagination to derive a warm human face from what was left. The face was unrecognizable as any real human face, and so I added features of my own grandmother's face to reflect how I imagined the woman looked during life. Furthermore, viewing the cadaver's face early in the course was a choice I made for the purpose of this work, which was to see the cadaver not as an object but as a person who donated her body as a gift for my benefit. Dr. DePace adds: The students at Drexel University College of Medicine have two exposures to the cadavers in the lab prior to any dissection. During the week of orientation, students visit the gross anatomy lab in small groups, where they are told about the cadaver donor program here in Pennsylvania and are shown a cadaver. During this orientation we stress the concept that the cadaver is a gift to be revered and respected. When the course begins in earnest, we have an initial lab exercise entitled “Evaluation of the Cadaver.” During this lab, the assembled lab groups meet at their assigned locations. They are asked to uncover their cadavers for the first time and to examine them for surgical scars and medical appliances. Each group is then instructed to sponge down their cadaver and to place the cadaver in the prone position in preparation for the first dissection. Before they leave the lab, the students completely replace the covering sheets. Many might argue that this somewhat ritualistic hour is a waste of curricular time. However, we feel that it is a gentle method of introducing our students to their first patients. It is a way of getting them to lay hands-on for the first time, a way to help each student become involved in the care and preservation of their cadaver, and a way of stressing the humanity of the gift that is to be part of their education as physicians. Michelle Paff Second-year medical student, Drexel University College of Medicine, Philadelphia, Pennsylvania; [email protected]. Dennis DePace, PhD Associate professor of neurobiology and anatomy and course director of medical gross anatomy, Drexel University College of Medicine, Philadelphia, Pennsylvania.

A face emotion tree structure representation with probabilistic recursive neural network modeling

This paper describes a novel structural approach to recognize the human facial features for emotion recognition. Conventionally, features extracted from facial images are represented by relatively poor representations, such as arrays or sequences, with a static data structure. In this study, we propose to extract facial expression features vectors as Localized Gabor Features (LGF) and then transform these feature vectors into FacE Emotion Tree Structures (FEETS) representation. It is an extension of the Human Face Tree Structures (HFTS) representation presented in (Cho and Wong in Lecture notes in computer science, pp 1245–1254, 2005). This facial representation is able to simulate as human perceiving the real human face and both the entities and relationship could contribute to the facial expression features. Moreover, a new structural connectionist architecture based on a probabilistic approach to adaptive processing of data structures is presented. The so-called probabilistic based recursive neural network (PRNN) model extended from Frasconi et al. (IEEE Trans Neural Netw 9:768–785, 1998) is developed to train and recognize human emotions by generalizing the FEETS representation. For empirical studies, we benchmarked our emotion recognition approach against other well known classifiers. Using the public domain databases, such as Japanese Female Facial Expression (JAFFE) (Lyons et al. in IEEE Trans Pattern Anal Mach Intell 21(12):1357–1362, 1999; Lyons et al. in third IEEE international conference on automatic face and gesture recognition, 1998) database and Cohn–Kanade AU-Coded Facial Expression (CMU) Database (Cohn et al. in 7th European conference on facial expression measurement and meaning, 1997), our proposed system might obtain an accuracy of about 85–95% for subject-dependent and subject-independent conditions. Moreover, by testing images having artifacts, the proposed model significantly supports the robust capability to perform facial emotion recognition.

Effect of Title on Eye-Movement Exploration of Cubist Paintings by Fernand Léger

Franklin et al (1993, Leonardo 26 103-108) reported that title information influenced the interpretation of paintings but not the way observers explore and look at the paintings; in their study subjects used a hand pointer to indicate where they looked. We used eye-movement recording and examined the effect of title on eye-movement exploration of nonrealistic cubist paintings giving rise to free interpretation. Three paintings by Fernand Léger were used: The Wedding contained high density of small fragments of real human faces, limbs, or arbitrary fragments mixed with large plane surfaces; The Alarm Clock consisted of arbitrary fragments creating perception of a person; Contrast of Forms contained forms and cylinders. Different groups of naive subjects explored paintings without knowing the title (spontaneous condition), with the instruction to invent a title (active condition), and after announcement of the authentic title (driven condition). Exploration time was unrestricted and eye movements were recorded by Chronos video-oculography. Fixation duration was found to increase in the driven condition relative to active condition; such increase occurred for all paintings. In contrast, fixation-duration variability remained stable over all title conditions. Saccade amplitude increased in the driven condition for Contrast of Forms. Increase of fixation duration and of saccade size are attributed to additional cognitive analysis, ie search fitting between the title and the painting. When comparing paintings within each title condition, The Wedding produced different results than the other paintings: longer exploration time (in spontaneous condition), higher fixation duration variability (in spontaneous and driven conditions), but smaller saccade sizes (in active and driven conditions). The differences are attributed to visual aspects (high density of small fragments) but also to complex semantic analysis of multiple segments of faces and limbs contained by this painting. Spatial distribution of fixation time was highly selective, with a preponderance of the central area that was the most fixated for all paintings and all title conditions. In the driven condition, however, loci of most frequent fixations were different than in the other conditions from the first 5 s; particularly for The Alarm Clock the title drove the eyes rapidly on the inconspicuous fragment of the clock. Our findings go against Franklin's conclusions. We conclude that title information influences both physiological parameters of eye movements and the distribution of fixation time over different selected areas of the painting.