3D Gaze Estimation from Remote RGB-D Sensors

The development of systems able to retrieve and characterise the state of humans is important for many applications and fields of study. In particular, as a display of attention and interest, gaze is a fundamental cue in understanding people activities, behaviors, intentions, state of mind and personality. Moreover, gaze plays a major role in the communication process, like for showing attention to the speaker, indicating who is addressed or averting gaze to keep the floor. Therefore, many applications within the fields of human-human, human-robot and human-computer interaction could benefit from gaze sensing. However, despite significant advances during more than three decades of research, current gaze estimation technologies can not address the conditions often required within these fields, such as remote sensing, unconstrained user movements and minimum user calibration. Furthermore, to reduce cost, it is preferable to rely on consumer sensors, but this usually leads to low resolution and low contrast images that current techniques can hardly cope with. In this thesis we investigate the problem of automatic gaze estimation under head pose variations, low resolution sensing and different levels of user calibration, including the uncalibrated case. We propose to build a non-intrusive gaze estimation system based on remote consumer RGB-D sensors. In this context, we propose algorithmic solutions which overcome many of the limitations of previous systems. We thus address the main aspects of this problem: 3D head pose tracking, 3D gaze estimation, and gaze based application modeling. First, we develop an accurate model-based 3D head pose tracking system which adapts to the participant without requiring explicit actions. Second, to achieve a head pose invariant gaze estimation, we propose a method to correct the eye image appearance variations due to head pose. We then investigate on two different methodologies to infer the 3D gaze direction. The first one builds upon machine learning regression techniques. In this context, we propose strategies to improve their generalization, in particular, to handle different people. The second methodology is a new paradigm we propose and call geometric generative gaze estimation. This novel approach combines the benefits of geometric eye modeling (normally restricted to high resolution images due to the difficulty of feature extraction) with a stochastic segmentation process (adapted to low-resolution) within a Bayesian model allowing the decoupling of user specific geometry and session specific appearance parameters, along with the introduction of priors, which are appropriate for adaptation relying on small amounts of data. The aforementioned gaze estimation methods are validated through extensive experiments in a comprehensive database which we collected and made publicly available. Finally, we study the problem of automatic gaze coding in natural dyadic and group human interactions. The system builds upon the thesis contributions to handle unconstrained head movements and the lack of user calibration. It further exploits the 3D tracking of participants and their gaze to conduct a 3D geometric analysis within a multi-camera setup. Experiments on real and natural interactions demonstrate the system is highly accuracy. Overall, the methods developed in this dissertation are suitable for many applications, involving large diversity in terms of setup configuration, user calibration and mobility.

Error handling in multimodal voice-enabled interfaces of tour-guide robots using graphical models

Plamen Prodanov

Mobile service robots are going to play an increasing role in the society of humans. Voice-enabled interaction with service robots becomes very important, if such robots are to be deployed in real-world environments and accepted by the vast majority of potential human users. The research presented in this thesis addresses the problem of speech recognition integration in an interactive voice-enabled interface of a service robot, in particular a tour-guide robot. The task of a tour-guide robot is to engage visitors to mass exhibitions (users) in dialogue providing the services it is designed for (e.g. exhibit presentations) within a limited time. In managing tour-guide dialogues, extracting the user goal (intention) for requesting a particular service at each dialogue state is the key issue. In mass exhibition conditions speech recognition errors are inevitable because of noisy speech and uncooperative users of robots with no prior experience in robotics. They can jeopardize the user goal identification. Wrongly identified user goals can lead to communication failures. Therefore, to reduce the risk of such failures, methods for detecting and compensating for communication failures in human-robot dialogue are needed. During the short-term interaction with visitors, the interpretation of the user goal at each dialogue state can be improved by combining speech recognition in the speech modality with information from other available robot modalities. The methods presented in this thesis exploit probabilistic models for fusing information from speech and auxiliary modalities of the robot for user goal identification and communication failure detection. To compensate for the detected communication failures we investigate multimodal methods for recovery from communication failures. To model the process of modality fusion, taking into account the uncertainties in the information extracted from each input modality during human-robot interaction, we use the probabilistic framework of Bayesian networks. Bayesian networks are graphical models that represent a joint probability function over a set of random variables. They are used to model the dependencies among variables associated with the user goals, modality related events (e.g. the event of user presence that is inferred from the laser scanner modality of the robot), and observed modality features providing evidence in favor of these modality events. Bayesian networks are used to calculate posterior probabilities over the possible user goals at each dialogue state. These probabilities serve as a base in deciding if the user goal is valid, i.e. if it can be mapped into a tour-guide service (e.g. exhibit presentation) or is undefined – signaling a possible communication failure. The Bayesian network can be also used to elicit probabilities over the modality events revealing information about the possible cause for a communication failure. Introducing new user goal aspects (e.g. new modality events and related features) that provide auxiliary information for detecting communication failures makes the design process cumbersome, calling for a systematic approach in the Bayesian network modelling. Generally, introducing new variables for user goal identification in the Bayesian networks can lead to complex and computationally expensive models. In order to make the design process more systematic and modular, we adapt principles from the theory of grounding in human communication. When people communicate, they resolve understanding problems in a collaborative joint effort of providing evidence of common shared knowledge (grounding). We use Bayesian network topologies, tailored to limited computational resources, to model a state-based grounding model fusing information from three different input modalities (laser, video and speech) to infer possible grounding states. These grounding states are associated with modality events showing if the user is present in range for communication, if the user is attending to the interaction, whether the speech modality is reliable, and if the user goal is valid. The state-based grounding model is used to compute probabilities that intermediary grounding states have been reached. This serves as a base for detecting if the the user has reached the final grounding state, or wether a repair dialogue sequence is needed. In the case of a repair dialogue sequence, the tour-guide robot can exploit the multiple available modalities along with speech. For example, if the user has failed to reach the grounding state related to her/his presence in range for communication, the robot can use its move modality to search and attract the attention of the visitors. In the case when speech recognition is detected to be unreliable, the robot can offer the alternative use of the buttons modality in the repair sequence. Given the probability of each grounding state, and the dialogue sequence that can be executed in the next dialogue state, a tour-guide robot has different preferences on the possible dialogue continuation. If the possible dialogue sequences at each dialogue state are defined as actions, the introduced principle of maximum expected utility (MEU) provides an explicit way of action selection, based on the action utility, given the evidence about the user goal at each dialogue state. Decision networks, constructed as graphical models based on Bayesian networks are proposed to perform MEU-based decisions, incorporating the utility of the actions to be chosen at each dialogue state by the tour-guide robot. These action utilities are defined taking into account the tour-guide task requirements. The proposed graphical models for user goal identification and dialogue error handling in human-robot dialogue are evaluated in experiments with multimodal data. These data were collected during the operation of the tour-guide robot RoboX at the Autonomous System Lab of EPFL and at the Swiss National Exhibition in 2002 (Expo.02). The evaluation experiments use component and system level metrics for technical (objective) and user-based (subjective) evaluation. On the component level, the technical evaluation is done by calculating accuracies, as objective measures of the performance of the grounding model, and the resulting performance of the user goal identification in dialogue. The benefit of the proposed error handling framework is demonstrated comparing the accuracy of a baseline interactive system, employing only speech recognition for user goal identification, and a system equipped with multimodal grounding models for error handling.

EPFL2006

A collaborative approach to image segmentation and behavior recognition from image sequences

Laura Ioana Gui

Visual behavior recognition is currently a highly active research area. This is due both to the scientific challenge posed by the complexity of the task, and to the growing interest in its applications, such as automated visual surveillance, human-computer interaction, medical diagnosis or video indexing/retrieval. A large number of different approaches have been developed, whose complexity and underlying models depend on the goals of the particular application which is targeted. The general trend followed by these approaches is the separation of the behavior recognition task into two sequential processes. The first one is a feature extraction process, where features which are considered relevant for the recognition task are extracted from the input image sequence. The second one is the actual recognition process, where the extracted features are classified in terms of the pre-defined behavior classes. One problematic issue of such a two-pass procedure is that the recognition process is highly dependent on the feature extraction process, and does not have the possibility to influence it. Consequently, a failure of the feature extraction process may impair correct recognition. The focus of our thesis is on the recognition of single object behavior from monocular image sequences. We propose a general framework where feature extraction and behavior recognition are performed jointly, thereby allowing the two tasks to mutually improve their results through collaboration and sharing of existing knowledge. The intended collaboration is achieved by introducing a probabilistic temporal model based on a Hidden Markov Model (HMM). In our formulation, behavior is decomposed into a sequence of simple actions and each action is associated with a different probability of observing a particular set of object attributes within the image at a given time. Moreover, our model includes a probabilistic formulation of attribute (feature) extraction in terms of image segmentation. Contrary to existing approaches, segmentation is achieved by taking into account the relative probabilities of each action, which are provided by the underlying HMM. In this context, we solve the joint problem of attribute extraction and behavior recognition by developing a variation of the Viterbi decoding algorithm, adapted to our model. Within the algorithm derivation, we translate the probabilistic attribute extraction formulation into a variational segmentation model. The proposed model is defined as a combination of typical image- and contour-dependent energy terms with a term which encapsulates prior information, offered by the collaborating recognition process. This prior information is introduced by means of a competition between multiple prior terms, corresponding to the different action classes which may have generated the current image. As a result of our algorithm, the recognized behavior is represented as a succession of action classes corresponding to the images in the given sequence. Furthermore, we develop an extension of our general framework, that allows us to deal with a common situation encountered in applications. Namely, we treat the case where behavior is specified in terms of a discrete set of behavior types, made up of different successions of actions, which belong to a shared set of action classes. Therefore, the recognition of behavior requires the estimation of the most probable behavior type and of the corresponding most probable succession of action classes which explains the observed image sequence. To this end, we modify our initial model and develop a corresponding Viterbi decoding algorithm. Both our initial framework and its extension are defined in general terms, involving several free parameters which can be chosen so as to obtain suitable implementations for the targeted applications. In this thesis, we demonstrate the viability of the proposed framework by developing particular implementations for two applications. Both applications belong to the field of gesture recognition and concern finger-counting and finger-spelling. For the finger-counting application, we use our original framework, whereas for the finger-spelling application, we use its proposed extension. For both applications, we instantiate the free parameters of the respective frameworks with particular models and quantities. Then, we explain the training of the obtained models from specific training data. Finally, we present the results obtained by testing our trained models on new image sequences. The test results show the robustness of our models in difficult cases, including noisy images, occlusions of the gesturing hand and cluttered background. For the finger-spelling application, a comparison with the traditional sequential approach to image segmentation and behavior recognition illustrates the superiority of our collaborative model.

EPFL2008

Exploring Spatial Perspective Taking in Human-Robot Interaction

Elmira Yadollahi

Humans often rely on their perspective taking skills to thrive within the world's complex relations and connections. An adequate understanding of others' spatial perspectives can increase the quality of the interaction, not only perceptually but also cognitively. This thesis is dedicated to exploring children and adults' spatial perspective taking abilities emerging from the interaction with embodied and virtual robots in different contexts. While most previous approaches were limited to particular circumstances and targeted adults, the proposed approach developed a cognitive model incorporated in agents to foster perspective taking abilities in different contexts. The developments of the model also detail the processes used by humans to infer spatial connections from other's viewpoints and it is adaptable to add or remove processes based on the context of interaction.First, this thesis explores different interaction modalities and cognitive abilities through user studies with children. Each interaction outlines a set of components and processes required to develop a perspective taking model for robots and agents. The platform developed for the first study aims at evaluating the effect of a robot's non-verbal gestures, such as pointing, on children's joint attention during reading activity. The second study evaluates children's perspective adaptation to the robot in the context of collaborative activity. We expanded our observations to include children's first perspective choice, how they tried to accommodate the robot's perspective, and how they updated their mental model during the interaction. The third study explores children's spatial perspective taking abilities using game-based interaction and non-verbal channels.Inspired by previous studies, the thesis proposes a cognitive model that uses automatic and cognitive controlled processes to generate behaviors and decisions for the robot. The model focuses on processes linked with taking spatial perspectives and can be integrated into any agent architecture that deals with decision making and reasoning. The agent processes are then adapted to a new interaction scenario. Finally, the thesis evaluates the model through a user study with adult participants and within a virtual platform, a drastic change from the exploratory study designs caused by the pandemic. The final studies are designed as two-player games with two virtual robots that interact with each other in two contexts of competition and cooperation. In the competitive version of the game, the robot guided by the human plays against the robot equipped with the perspective taking model, while in the cooperative version, the robots guide each other to win the game as a team. We performed two between-subject studies with more than 180 adults to evaluate the participants' perception of a robot with complete perspective taking abilities, compared to one with limited abilities. Participants were more influenced by the robot's perspective taking abilities in the cooperative game compared to the competitive one, which was reflected in their ratings of the robot's intelligence and game fun. Experimental results on the model evaluation can open up future possibilities for exploring links between the perspective developments in children through cooperation and competition. Furthermore, the model can be extended to study other perceptional, cognitive, and affective dimensions, such as prosocial behavior and and transparency.

EPFL2021