Skip to main content

Über dieses Buch

This book constitutes the refereed proceedings of the Second International Workshop on Human Behavior Understanding, HBU 2011, held in Amsterdam, The Netherlands, in November 2011, in conjunction with AmI-11, the International Joint Conference on Ambient Intelligence.

The 13 revised full papers presented together with 2 keynote talks and one summarizing paper were carefully reviewed and selected from 32 submissions. The papers are organized in topical sections on analysis of human actions and activities, face and gesture analysis, persuasive technologies, and social interactions.



Human Behavior Understanding for Inducing Behavioral Change: Application Perspectives

Human Behavior Understanding for Inducing Behavioral Change: Application Perspectives

Pervasive sensing and human behavior understanding can help us in implementing or improving systems that can induce behavioral change. In this introductory paper of the 2nd International Workshop on Human Behavior Understanding (HBU’11), which has a special focus theme of “Inducing Behavioral Change”, we provide a taxonomy to describe where and how HBU technology can be harnessed to this end, and supply a short survey of the area from an application perspective. We also consider how social signals and settings relate to this concept.
Albert Ali Salah, Bruno Lepri, Fabio Pianesi, Alex Sandy Pentland

Analysis of Human Actions and Activities

Urban Computing and Smart Cities: Opportunities and Challenges in Modelling Large-Scale Aggregated Human Behavior

City-wide urban infrastructures are increasingly reliant on networked technology to improve and expand their services. As a side effect of this digitalization, large amounts of data –digital footprints– can be sensed and analyzed to uncover patterns of human urban behavior and to augment the city experience of its citizens. In my talk, I will introduce the main concepts, opportunities and challenges in the emerging area of urban computing/smart cities, which focuses on improving the quality of life of an urban environment by understanding the city dynamics through the data provided by ubiquitous technologies. This is a human-centric, data-rich area that spans multiple disciplines, including sociology, computer science and urban planning. From a computer science perspective, there are challenges in a variety of domains, including data visualization, storage, security, privacy, machine learning, data mining and pattern recognition. Some of the applications of smart cities include traffic forecasting, modeling of the spread of biological viruses, urban and transportation design and location-based services.
Nuria Oliver

Human Action Categorization Using Ultrasound Micro-Doppler Signatures

The spectrotemporal representation of an ultrasonar wave reflected by an object contains frequency shifts corresponding to the velocity of the object’s moving parts, also known as the micro-Doppler signature. The present study describes how the micro-Doppler signature of human subjects, collected in two experiments, can be used to categorize the action performed by the subject. The proposed method segments the spectrogram into temporal events, learns prototypes and categorizes the events using a Nearest Neighbour approach. Results show an average accuracy above 95%, with some categories reaching 100%, and a strong robustness to variations in the model parameters. The low computational cost of the system, together with its high accuracy, even for short length inputs, make it appropriate for a real-time implementation with applications to intelligent surveillance, monitoring and related disciplines.
Salvador Dura-Bernal, Guillaume Garreau, Charalambos Andreou, Andreas Andreou, Julius Georgiou, Thomas Wennekers, Susan Denham

Sequential Deep Learning for Human Action Recognition

We propose in this paper a fully automated deep model, which learns to classify human actions without using any prior knowledge. The first step of our scheme, based on the extension of Convolutional Neural Networks to 3D, automatically learns spatio-temporal features. A Recurrent Neural Network is then trained to classify each sequence considering the temporal evolution of the learned features for each timestep. Experimental results on the KTH dataset show that the proposed approach outperforms existing deep models, and gives comparable results with the best related works.
Moez Baccouche, Franck Mamalet, Christian Wolf, Christophe Garcia, Atilla Baskurt

One-Sequence Learning of Human Actions

In this paper we address the problem of human action recognition from a single training sequence per class using a modified version of the Hidden Markov Model. Inspired by codebook approaches in object and scene categorization, we first construct a codebook of possible discrete observations by applying a clustering algorithm to all samples from all classes. The number of clusters defines the size of the codebook. Given a new observation, we assign to it a probability to belong to every cluster, i.e., to correspond to a discrete value of the codebook. In this sense, we change the ‘winner takes all’ rule in the discrete-observation HMM for a distributed probability of membership. It implies the modification of the Baum-Welch algorithm for training discrete HMM to be able to deal with fuzzy observations. We compare our approach with other models such as, dynamic time warping (DTW), continuous-observation HMM, Conditional Random Fields (CRF) and Hidden Conditional Random Fields (HCRF) for human action recognition.
Carlos Orrite, Mario Rodríguez, Miguel Montañés

Face and Gesture Analysis

Analyzing Facial Behavioral Features from Videos

Face analysis from videos can be approached using two different strategies, depending on whether the temporal information is used or not. The most straightforward strategy applies still image based techniques to some selected (or all) frames and then fuses the results over the sequence. In contrast, an emerging strategy consists of encoding both facial structure and dynamics through spatiotemporal representations. To gain insight into the usefulness of facial dynamics, this paper considers two baseline systems and compares static versus spatiotemporal approaches to face analysis from videos. The first approach is based only on static images and uses spatial Local Binary Pattern features as inputs to SVM classifiers, while the second baseline system combines facial appearance and motion through a spatiotemporal representation using Volume LBP features as inputs to SVM classifiers. Preliminary experiments on classifying face patterns into different categories based on gender, identity, age, and ethnicity point out very interesting findings on the role of facial dynamics in face analysis from videos.
Abdenour Hadid

Adaptive Integration of Multiple Cues for Contingency Detection

Critical to natural human-robot interaction is the capability of robots to detect the contingent reactions by humans. In various interaction scenarios, a robot can recognize a human’s intention by detecting the presence or absence of a human response to its interactive signal. In our prior work [1], we addressed the problem of detecting visible reactions by developing a method of detecting changes in human behavior resulting from a robot signal. We extend the previous behavior change detector by integrating multiple cues using a mechanism that operates at two levels of information integration and then adaptively applying these cues based on their reliability. We propose a new method for evaluating reliability of cues online during interaction. We perform a data collection experiment with help of the Wizard-of-Oz methodology in a turn-taking scenario in which a humanoid robot plays the turn-taking imitation game “Simon says” with human partners. Using this dataset, which includes motion and body pose cues from a depth and color image, we evaluate our contingency detection module with the proposed integration mechanisms and show the importance of selecting the appropriate level of cue integration.
Jinhan Lee, Crystal Chao, Andrea L. Thomaz, Aaron F. Bobick

DTW Based Clustering to Improve Hand Gesture Recognition

Vision based hand gesture recognition systems track the hands and extract their spatial trajectory and shape information, which are then classified with machine learning methods. In this work, we propose a dynamic time warping (DTW) based pre-clustering technique to significantly improve hand gesture recognition accuracy of various graphical models used in the human computer interaction (HCI) literature. A dataset of 1200 samples consisting of the ten digits written in the air by 12 people is used to show the efficiency of the method. Hidden Markov model (HMM), input-output HMM (IOHMM), hidden conditional random field (HCRF) and explicit duration model (EDM), which is a type of hidden semi Markov model (HSMM) are trained on the raw dataset and the clustered dataset. Optimal model complexities and recognition accuracies of each model for both cases are compared. Experiments show that the recognition rates undergo substantial improvement, reaching perfect accuracy for most of the models, and the optimal model complexities are significantly reduced.
Cem Keskin, Ali Taylan Cemgil, Lale Akarun

Persuasive Technologies

Augmenting Social Interactions: Experiments in Socio-emotional Computing

In recent decades, research on affective computing, social signal processing, and mediated communication has flourished. Combining these diverse fields leads to the new, multidisciplinary area of socio-emotional computing, where computing technologies are applied to transform and enrich communication between people, either mediated or face-to-face. As a research field, socio-emotional computing serves a number of goals. First, it aims to inform the design of communication media through identifying, implementing and validating those socio-emotional elements that enable or augment awareness, mutual understanding, empathy, and intimacy between people. Augmented social interactions can be beneficial to many application areas, including mental healthcare, training and coaching, behavior change, negotiation, and intimate social interactions. Secondly, research in socio-emotional computing allows us to obtain a more fundamental understanding of the impact of mediated communication on human intimacy and social connectedness. Finally, media tools developed to augment social interactions can, at the same time, serve as research tools to extend and improve research on the fundamental emotional and interpersonal processes underlying intimate communication. In this presentation, I will highlight some of the exciting research opportunities that emerge in this multidisciplinary field, and will present a number of experiments that exemplify socio-emotional computing as applied to intimacy, empathy, and persuasion.
Wijnand Ijsselsteijn

An Energy-Saving Support System for Office Environments

We present a system that helps office workers to save energy at work. The system features two concepts which are differing from current smart metering systems. It takes the special characteristics of office environments into account where saving energy has lower priority than the actual working processes. Firstly, the system uses unobtrusive technology in order not to interrupt the normal working processes of office workers. Secondly, the system minimizes the effort of workers to deal with the topic of saving energy so that it can be done en passant. In an explorative user study, we examine if the system is considered useful by users.
Marc Jentsch, Marco Jahn, Ferry Pramudianto, Jonathan Simon, Amro Al-Akkad

From Stress Awareness to Coping Strategies of Medical Staff: Supporting Reflection on Physiological Data

Nurses and physicians on a stroke unit constantly face pressure and emotional stress. Physiological sensors can create awareness of one’s own stress and persuade medical staff to reflect on their own behavior and coping strategies. In this study, eight nurses and physicians of a stroke unit were equipped with a wearable electrocardiography (ECG) and acceleration sensor during their everyday work in order to (a) make them aware of stress and (b) support the re-calling of experiences to identify stressors. In an interview one week later, the participants were asked to recollect stress related events through the examination of the sensor data. Although high activity levels diminished the expressiveness of the data, physicians and nurses could recall stressful events and were interested in their physiological signals. However, existing coping strategies turned out as barriers to the adoption of new tools. Future persuasive applications should focus on integration with existing coping strategies to scaffold the reflection process.
Lars Müller, Verónica Rivera-Pelayo, Christine Kunzmann, Andreas Schmidt

Why Won’t You Do What’s Good for You? Using Intelligent Support for Behavior Change

Human health depends to a large extent on their behavior. Adopting a healthy lifestyle often requires behavior change. This paper presents a computational model of behavior change that describes formal relations between the determinants of behavior change, based on existing psychological theories. This model is developed to function as the core of a reasoning mechanism of an intelligent support system that is able to create theory-based intervention messages. The system first tries to determine the reason of the occurrence of the unwanted behavior by asking short questions via a mobile phone application and by gathering information from an online lifestyle diary. The system then attempts to influence the user using tailored information and persuasive motivational messages.
Michel Klein, Nataliya Mogles, Arlette van Wissen

A Research Framework for Playful Persuasion Based on Psychological Needs and Bodily Interaction

This paper presents a research framework that relates interactive systems to behavioral change with psychological needs and bodily interaction as intermediating variables. The framework is being developed in a multidisciplinary research project that focuses on how to design intelligent play environments that promote physical and social activities. Here, the framework serves to generate design relevant research questions and to guide communication amongst group members.
Marco Rozendaal, Arnold Vermeeren, Tilde Bekker, Huib de Ridder

Social Interactions

Automatic Modeling of Dominance Effects Using Granger Causality

We propose the use of Granger Causality to model the effects that dominant people induce on the other participants’ behavioral patterns during small group interactions. We test the proposed approach on a dataset of brainstorming and problem solving tasks collected using the sociometric badges’ accelerometers. The expectation that more dominant people have generalized higher influence is not borne out; however some more nuanced patterns emerge. In the first place, more dominant people tend to behave differently according to the nature of the task: during brainstorming they engage in complex relations where they simultaneously play the role of influencer and of influencee, whereas during problem solving they tend to be influenced by less dominant people. Moreover, dominant people adopt a complementarity stance, increasing or decreasing their body activity in an opposite manner to their influencers. On the other hand, less dominant people react (almost) as frequently with mimicry as with complementary. Finally, we can also see that the overall level of influence in a group can be associated with the group’s performance, in particular for problem solving task.
Kyriaki Kalimeri, Bruno Lepri, Taemie Kim, Fabio Pianesi, Alex Sandy Pentland

Abnormal Crowd Behavior Detection by Social Force Optimization

We propose a new scheme for detecting and localizing the abnormal crowd behavior in video sequences. The proposed method starts from the assumption that the interaction force, as estimated by the Social Force Model (SFM), is a significant feature to analyze crowd behavior. We step forward this hypothesis by optimizing this force using Particle Swarm Optimization (PSO) to perform the advection of a particle population spread randomly over the image frames. The population of particles is drifted towards the areas of the main image motion, driven by the PSO fitness function aimed at minimizing the interaction force, so as to model the most diffused, normal, behavior of the crowd. In this way, anomalies can be detected by checking if some particles (forces) do not fit the estimated distribution, and this is done by a RANSAC-like method followed by a segmentation algorithm to finely localize the abnormal areas. A large set of experiments are carried out on public available datasets, and results show the consistent higher performances of the proposed method as compared to other state-of-the-art algorithms, proving the goodness of the proposed approach.
R. Raghavendra, Alessio Del Bue, Marco Cristani, Vittorio Murino

Understanding the Influence of Social Interactions on Individual’s Behavior Pattern in a Work Environment

In this work, we study social interactions in a work environment and investigate how the presence of other people changes personal behavior patterns. We design the visual processing algorithms to track multiple people in the environment and detect dyadic interactions using a discriminative classifier. The locations of the users are associated with semantic tasks based on the functions of the areas. Our learning method then deduces patterns from the trajectories of people and their interactions. We propose an algorithm to compare the patterns of a user in the presence and absence of social interactions. We evaluate our method on a video dataset collected in a real office. By detecting interactions, we gain insights in not only how often people interact, but also in how these interactions affect the usual routines of the users.
Chih-Wei Chen, Asier Aztiria, Somaya Ben Allouch, Hamid Aghajan


Weitere Informationen

Premium Partner