Style-Controllable Speech-Driven Gesture Synthesis Using Normalising Flows

Automatic synthesis of realistic gestures promises to transform the fields of animation, avatars and communicative agents. In off-line applications, novel tools can alter the role of an animator to that of a director, who provides only high-level input for the desired animation; a learned network then translates these instructions into an appropriate sequence of body poses. In interactive scenarios, systems for generating natural animations on the fly are key to achieving believable and relatable characters. In this paper we address some of the core issues towards these ends. By adapting a deep learning-based motion synthesis method called MoGlow, we propose a new generative model for generating state-of-the-art realistic speech-driven gesticulation. Owing to the probabilistic nature of the approach, our model can produce a battery of different, yet plausible, gestures given the same input speech signal. Just like humans, this gives a rich natural variation of motion. We additionally demonstrate the ability to exert directorial control over the output style, such as gesture level, speed, symmetry and spacial extent. Such control can be leveraged to convey a desired character personality or mood. We achieve all this without any manual annotation of the data. User studies evaluating upper-body gesticulation confirm that the generated motions are natural and well match the input speech. Our method scores above all prior systems and baselines on these measures, and comes close to the ratings of the original recorded motions. We furthermore find that we can accurately control gesticulation styles without unnecessarily compromising perceived naturalness. Finally, we also demonstrate an application of the same method to full-body gesticulation, including the synthesis of stepping motion and stance.     

Beyond graph neural networks with lifted relational neural networks

Machine Learning - We introduce a declarative differentiable programming framework, based on the language of Lifted Relational Neural Networks, where small parameterized logic programs are used to...     

Convex optimization of measurement allocation for magnetic tracking systems

Magnetic tracking is a popular technique that exploits static and low-frequency magnetic fields for positioning of quasi-stationary objects. One important system design aspect, which substantially influences the performance of the tracking system, is how to collect as much information as possible with a given number of measurements. In this work, we optimize the allocation of measurements given a large number of possible measurements of a generic magnetic tracking system that exploits time-division multiplexing. We exploit performance metrics based on the Fisher information matrix. In particular, the performance metrics measure worst-case or average performance in a measurement domain, i.e. the domain where the tracking is to be performed. An optimization problem with integer variables is formulated. By relaxing the constraint that the variables should be integer, a convex optimization problem is obtained. The two performance metrics are compared for several realistic measurement scenarios with planar transmitter constellations. The results show that the worst performance is obtained in the most distant parts of the measurement domain. Furthermore, measurement allocations optimized for worst-case performance require measurements in a larger area than measurement allocations optimized for average performance.     

Initial rapid wetting in metallic systems

The initial rapid wetting of a solid surface by a liquid phase is an important step in many industrial processes. Liquid-phase sintering of powder metallurgical steels is one such industrial process, where metallic powders of micrometer size are used. Investigating the dynamic wetting of a high-temperature metallic drop of micrometer size experimentally is very challenging. Here, a phase-field-based numerical model is first implemented and verified by accurately capturing the initial dynamic wetting of millimeter-sized metal drops and then the model is extended to predict the dynamic wetting of a micrometer-sized metal drop. We found, in accordance with recent observations, that contact line friction is required for accurate simulation of dynamic wetting. Our results predict the wetting time for a micrometer-sized metal drop and also indicate that the dynamic wetting patterns at the micro- and millimeter length scales are qualitatively similar. We also found that the wetting process is much faster for a micrometer-sized metal drop compared to a millimeter-sized metal drop.     

Leisurely Lesbians in a Small City in South Africa

Over the past two decades, particularly in post-industrial societies, the geographies of gay men’s and, to a lesser extent, gay women’s lives have received considerable research attention. A central research interest has been the mapping of gay geographies of leisure. There are, however, fewer studies in the academic record about gay leisure geographies in the developing world context, particularly in Africa. This investigation aims to address an aspect of this investigatory oversight by looking through the lens of white gay women in the city of Bloemfontein, South Africa. The study maps white gay women’s leisure-seeking and positions their (sexual) desire(s) relative to other homosexual and heterosexual leisure geographies. A highly complex and contradictory set of spatial and temporal patterns and practices emerge in which desire and repulsion are often in close proximity located in a number of leisure, retail, work, and private spaces. In addition, these spaces are inhabited by a range of different identity cohorts according to gender, gender performativity, class, and race.     

Efficiency and effectiveness of promotion systems for electricity generation from renewable energy sources - Lessons from EU countries

Currently, a wide range of strategies is implemented in different countries to increase the share of electricity from renewable energy sources (RES-E). A still controversial discussion is whether quantity-driven (like Tradable Green Certificates (TGCs) based on quotas) or price-driven (like feed-in-tariffs (FIT)) instruments lead to preferable solutions for society. The core objective of this paper is to compare the perspectives of quota-based certificate trading systems for an efficient and effective increase of RES-E with FIT. The major results of this analysis are: (i) The success stories of growth in RES-E in EU member states in recent years has been triggered by FIT implemented in a technology-specific manner at modest costs for European citizens; (ii) At present, TGC systems in most countries applied show a low effectiveness with respect to RES-E deployment of less mature technologies such as solar PV (with improving tendencies in e.g. the UK or Italy with respect to certain technologies); (iii) Compared to short term trading in TGC markets the intrinsic stability of FIT systems appears to be a key element for success; (iv) Hence, currently a well-designed (dynamic) FIT system provides a certain deployment of RES-E in the shortest time and at lowest costs for society.     

Energy cost and error performance of range-aware, anchor-free localization algorithms

Many applications that use sensor data from a wireless sensor network (WSN) require corresponding node position information as well. Therefore, it is not surprising that a common figure of merit for localization algorithms is the accuracy of the position estimate produced. Similarly, the amount of communication required by a localization algorithm is often of paramount interest as well since it is common knowledge that communication expends the most energy in a WSN. However, localization algorithms seldom characterize their communication cost. Furthermore, when they do it is often merely qualitative and is typically described as “expensive”. For two types of range-aware, anchor-free localization algorithms we found the opposite to be true. Rather than being expensive, the communication costs were quite modest. So much so that we maintain range-aware, anchor-free localization algorithms should be chosen on the basis of the accuracy required by the intended application independent of the communication cost.In this paper, we examine the effect of node degree, node distribution, range error and network size on distance error and communication cost for both incremental and concurrent versions of range-aware, anchor-free algorithms. The concurrent algorithm is twice as accurate as the incremental, but less efficient. Furthermore, node degree influences the energy cost of the algorithms the most, but neither algorithm uses more than a surprisingly small 0.8% of a 560 mA h battery. This result indicates less energy efficient localization algorithms can be tolerated, especially if they provide better accuracy. Furthermore, if energy does need to be conserved, there is not much savings available within the localization algorithm and savings must be found in other areas such as the MAC protocol or routing algorithm.     

Design options for cooperation mechanisms under the new European renewable energy directive

In June 2009, a new EU directive on the promotion of renewable energy sources (RES) entered into effect. The directive 2009/28/EC, provides for three cooperation mechanisms that will allow member states to achieve their national RES target in cooperation with other member states: statistical transfer, joint projects, and joint support schemes. This article analyses the pros and cons of the three mechanisms and explores design options for their implementation through strategic and economic questions: How to counterbalance the major drawbacks of each mechanism? How to reflect a balance of costs and benefits between the involved member states? The analysis identifies a number of design options that respond to these questions, e.g. long term contracts to ensure sufficient flexibility for statistical transfers, a coordinated, standardised joint project approach to increase transparency in the European market, and a stepwise harmonisation of joint support schemes that is based on a cost-effective accounting approach. One conclusion is that the three cooperation mechanisms are closely interlinked. One can consider their relation to be a gradual transition from member state cooperation under fully closed national support systems in case of statistical transfers, to cooperation under fully open national support systems in a joint support scheme.