Machine Learning, Robotic, and Toy Computing
Permanent URI for this collection
1 - 5 of 5
ItemAutomatic Segmentation of Grammatical Facial Expressions in Sign Language: Towards an Inclusive Communication Experience( 2020-01-07)Nowadays, natural language processing techniques enable the development of applications that promote communication between humans and between humans and machines. Although the technology related to automated oral communication is mature and affordable, there are currently no appropriate solutions for visual-spatial languages. In the scarce efforts to automatically process sign languages, studies on non-manual gestures are rare, making it difficult to properly interpret the speeches uttered in those languages. In this paper, we present a solution for the automatic segmentation of grammatical facial expressions in sign language. This is a low-cost computational solution designed to integrate a sign language processing framework that supports the development of simple but high value-added applications for the context of universal communication. Moreover, we present a discussion of the difficulties faced by this solution to guide future research in this area.
ItemIoT4Fun Rapid Prototyping Toolkit for Smart Toys( 2020-01-07)Rapid prototyping tools turn the design of smart toys faster and easier for creative teams. Appropriate tools for smart toys should meet a list of requirements, which include distributed data collection and adaptability for assorted toy shapes and size. The IoT4Fun toolkit innovates by mixing the embedded, modular, and plug-and-play approaches. It supports motion tracking data, wireless communication, and contactless identification. IoT4Fun demonstrates its effectiveness to design a variety of smart toy solutions by fitting into a hula-hoop toy until spherical, cubic, and wearable shapes. Solutions connect with either mobile applications or other toys and play rules range from open-ended to closed behaviors. End-users exhaustively tested developed solutions, and technical assessment evaluates their integrity after playtesting sessions. Results show comparative data on battery consumption and vulnerabilities threats for data security and privacy of each design. Future versions of IoT4Fun can benefit from miniaturization, robustness, and reliability improvements.
ItemA Literature Survey on Smart Toy-related Children's Privacy Risks( 2020-01-07)Smart toys have become popular as technological solutions offer a better experience for children. However, the technology used increases the risks to children's privacy, which does not seem to have become a real concern for toy makers. Most researchers in this domain are vague in defining their motivations due to lack of an expert survey to support them. We conducted a literature survey to find papers on smart toy-related children's privacy risks and mitigation solutions. We analyzed 26 papers using a taxonomy for privacy principles and preserving techniques adapted from the IoT context. Our analysis shows that some types of risks received more attention, especially (a) confidentiality, (b) use, retention and disclosure limitation, (c) authorization, (d) consent and choice, (e) openness, transparency and notice and (f) authentication. As for solutions, few were effectively presented; the vast majority related to data restriction -- (a) access control and (b) cryptographic.
ItemAssessing Mission Performance for Technology Reliant Missions( 2020-01-07)Operators today increasingly rely on technology to accomplish objectives. Although technology can increase mission success and efficiency in a majority of operations, it can simultaneously increase vulnerability prevalence, resulting in a higher exploitation likelihood. Defense methods have been proposed and evaluated based on their ability to ensure network security. However, these evaluation metrics do not fully quantify how network exploitation impacts mission task completion. Our mission performance model links cyber devices to mission tasks utilizing a mission’s mission map and evaluates a mission’s performance as the proportion of completed mission tasks in an agent based simulation. Our model allows for mission mappings with varying degrees of completion to enable a generic and adaptable model. We investigate the impact differing levels of mission map completion have on the mission performance metric for the same mission. Experiments serve to provide quantitative assessment for mission performance in cyber-network mission systems.