Figure 3 - uploaded by Marc Hanheide
Content may be subject to copyright.
Example images from the selected reference subsequences. Top row: signaling success via head gestures (left) and gaze direction (right). Bottom row: signaling failure via facial expressions. In each case, the first, middle, and last image of a reference subsequence is shown. Please refer to Sec. 6 and Sec. 7.
Source publication
Facial communicative signals (FCSs) such as head gestures, eye gaze, and facial expressions can provide useful feedback in conversations between people and also in human-robot interaction. This paper presents a pattern recognition approach for the interpretation of FCSs in terms of valence, based on the selection of discriminative subsequences in v...
Context in source publication
Context 1
... is only slightly below the average human recognition performance on this task (Sec. 4) and better than our previous results (Sec. 7.2). The standard deviation of the classification rates is very high and even systematic misclassifications occur (persons six and nine), both holds for the human performance and the previous results as well [23,25]. Fig. 3 depicts example images taken from the most discrimative reference subsequence of some ...
Citations
... For example, Droeschel et al. (2011) used two cameras and two laser range finders to detect human gazes or pointing gestures, while Anjum et al. (2014) used a Microsoft Kinect camera and Support Vector Machine to recognize eight activities (e.g., sit and drink or wave hello) with extreme accuracy. Lang et al. (2013) used a recognition mechanism able to localize faces and extract their features to attribute different emotional states (happiness, sadness, fear, etc.) in an object teaching scenario. Finally, Kulic and Croft (2007) used a classifier based on a Hidden Markov Model in order to estimate affective states from physiological data collected in HRI. ...
The vast expansion of research in human-robot interactions (HRI) these last decades has been accompanied by the design of increasingly skilled robots for engaging in joint actions with humans. However, these advances have encountered significant challenges to ensure fluent interactions and sustain human motivation through the different steps of joint action. After exploring current literature on joint action in HRI, leading to a more precise definition of these challenges, the present article proposes some perspectives borrowed from psychology and philosophy showing the key role of communication in human interactions. From mutual recognition between individuals to the expression of commitment and social expectations, we argue that communicative cues can facilitate coordination, prediction, and motivation in the context of joint action. The description of several notions thus suggests that some communicative capacities can be implemented in the context of joint action for HRI, leading to an integrated perspective of robotic communication.
... Active appearance models individualized for each participant enabled the development of highly discriminant feature vectors from subsegments of each video sequence, which were cross-validated and then tested. Accuracy of the system in recognizing 'success' versus 'failure' displays, despite considerable individual and intertrial variation, equaled average human recognition performance [90]. ...
Based on modern theories of signal evolution and animal communication, the behavioral ecology view of facial displays (BECV) reconceives our 'facial expressions of emotion' as social tools that serve as lead signs to contingent action in social negotiation. BECV offers an externalist, functionalist view of facial displays that is not bound to Western conceptions about either expressions or emotions. It easily accommodates recent findings of diversity in facial displays, their public context-dependency, and the curious but common occurrence of solitary facial behavior. Finally, BECV restores continuity of human facial behavior research with modern functional accounts of non-human communication , and provides a non-mentalistic account of facial displays well-suited to new developments in artificial intelligence and social robotics.
I argue in this paper that the claimed universal recognition of basic emotions corresponds to the recognition of conventionalized representations of emotions common in our culture. Section one presents some of the faces that people make in different circumstances, and argues that making faces is a form of action. Faces made function as narrative tools and as conversational tools. Section two compares and contrasts two conceptions of facial displays: basic emotion theories (BET) and the behavioral ecology view (BECV). The next section analyzes and evaluates BET’s claim concerning the universal expression of emotions. Section four argues that the still pictures of posed emotions used by Ekman correspond to conventionalized iconographic representations of emotions in our culture. The last section asks whether present day social robots can make faces. They cannot for two reasons, I argue. First because of the dominance of BET in robotic research, second because robots do not need to enter into strategic negotiation with their human partners. The faces of robots simply reproduce conventionalized expressions of emotions, that they do paradoxically bear witness to the central relevance of the behavioral ecology view of facial displays.
This paper investigates facial communicative signals (head gestures, eye gaze, and facial expressions) as nonverbal feedback in human-robot interaction. Motivated by a discussion of the literature, we suggest scenario-specific investigations due to the complex nature of these signals and present an object-teaching scenario where subjects teach the names of objects to a robot, which in turn shall term these objects correctly afterwards. The robot’s verbal answers are to elicit facial communicative signals of its interaction partners. We investigated the human ability to recognize this spontaneous facial feedback and also the performance of two automatic recognition approaches. The first one is a static approach yielding baseline results, whereas the second considers the temporal dynamics and achieved classification rates comparable to the human performance.
