ABSTRACT Animating realistic avatars requires using high‐quality animations for every possible state the avatar can be in. This includes actions like walking or running, but also subtle movements that convey emotions and personality. Idle animations, such as standing, breathing, or looking around, are crucial for realism and believability. In virtual applications, these are often handcrafted or recorded with actors, but this is costly. Furthermore, recording realistic idle animations may be complex, since the actor being aware of the recording could interfere with the genuineness of the movements. Currently, there are no large‐scale idle animation datasets for deep learning, and this recording challenge may partly explain this. Nevertheless, this paper concludes that both acted and genuine idle animations are perceived as real, and users are not able to distinguish between them. It also states that raw recorded idle animations and artist‐retouched ones are perceived differently. These conclusions mean that recording idle animations should be easier than expected, implying that actors can be instructed to act the movements, significantly simplifying the recording process. This should help future efforts to record idle animation datasets. Finally, we publish ReActIdle, the first three dimensional idle animation dataset containing long sequences of real and acted idle motions.

Natural gesturing is very important for the credibility of social robots. It is even more crucial for storytelling robots since the expression, emotion and emphasis must be highlighted. In this paper we propose a hybrid gesture generation approach for a storytelling robot that combines beats automatically generated by a GAN with a probabilistic semantic related gesture insertion system. Beats are executed according to a probability based on the duration of the sentences and semantic gesture insertions are dependent of the previous occurrences of the gestures associated to the words. The polarity of the text is extracted and affects several features of the motion to arouse emotion. A qualitative evaluation of robot behavior is conducted and confirms the approach as a promising one as storytelling system.

Abstract Natural gestures are a desirable feature for a humanoid robot, as they are presumed to elicit a more comfortable interaction in people. With this aim in mind, we present in this paper a system to develop a natural talking gesture generation behavior. A Generative Adversarial Network (GAN) produces novel beat gestures from the data captured from recordings of human talking. The data is obtained without the need for any kind of wearable, as a motion capture system properly estimates the position of the limbs/joints involved in human expressive talking behavior. After testing in a Pepper robot, it is shown that the system is able to generate natural gestures during large talking periods without becoming repetitive. This approach is computationally more demanding than previous work, therefore a comparison is made in order to evaluate the improvements. This comparison is made by calculating some common measures about the end effectors’ trajectories (jerk and path lengths) and complemented by the Fréchet Gesture Distance (FGD) that aims to measure the fidelity of the generated gestures with respect to the provided ones. Results show that the described system is able to learn natural gestures just by observation and improves the one developed with a simpler motion capture system. The quantitative results are sustained by questionnaire based human evaluation .

This paper presents a novel approach for iris dissimilarity computation based on Machine Learning paradigms and Computer Vision transformations. Based on the training dataset given by the MICHE II Challenge organizers, a set of classifiers has been constructed and tested, aiming at classifying a single image. The main novelty of this paper remains in the used approach to iris dissimilarity computation: given two iris images, both of them are classified using the same paradigm, obtaining the a posteriori probability for each of the considered class values. Hence, two distributions are obtained, one for each iris image, and the dissimilarity is computed as the distance between these two distributions. Experimental results indicate the appropriateness of this new approach, even though more research and experiments are needed to obtain some improvements and to accelerate the classification process.

Bertsolaritza, Basque improvised contest poetry, offers another sphere to develop robot body language and robot communication capabilities, that shares some similarities with theatrical performances. It is also a new area to work on social robotics. The work presented in this paper makes some steps forward in designing and implementing the set of behaviors the robots need to show in the stage to increase, on the one hand robot autonomy and on the other hand, credibility and sociability.

Class binarization strategies decompose the original multi-class problem into several binary sub-problems. One versus One (OVO) is one of the most popular class binarization techniques, which considers every pair of classes as a different sub-problem. Usually, the same classifier is applied to every sub-problem and then all the outputs are combined by some voting scheme. In this paper we present a novel idea where for each test instance we try to assign the best classifier in each sub-problem of OVO. To do so, we have used two simple Dynamic Classifier Selection (DCS) strategies that have not been yet used in this context. The two DCS strategies use K-NN to obtain the local region of the test-instance, and the classifier that performs the best for those instances in the local region, is selected to classify the new test instance. The difference between the two DCS strategies remains in the weight of the instance. In this paper we have also proposed a novel approach in those DCS strategies. We propose to use the K-Nearest Neighbor Equality (K-NNE) method to obtain the local accuracy. K-NNE is an extension of K-NN in which all the classes are treated independently: the K nearest neighbors belonging to each class are selected. In this way all the classes take part in the final decision. We have carried out an empirical study over several UCI databases, which shows the robustness of our proposal.