Verbal and non-verbal activities convey insightful information about people’s affect, empathy, and engagement during social interactions. In this paper, we investigate the usage of inertial sensors to recognize verbal (e.g., speaking), non-verbal (e.g., head nodding, shaking) and other activities (e.g., eating, no movement). We implement an end-to-end deep neural network to distinguish among these activities. We then explore the generalizability of the approach in three scenarios: (1) using new data to detect a known activity from a known user, (2) detecting a novel activity of a known user and (3) detecting the activity of an unknown user. Results show that using accelerometer and gyroscope sensors, the model achieves a balanced accuracy of 55% when tested on data from a new user, 41% on a new activity of an existing user, and 80% on new data of a known activity from an existing user. The results are between 7-47 percentage points higher than baseline classifiers.

A large plethora of models to predict human mobility exists in the literature. The problem of how to select the most appropriate model to solve a specific mobility prediction task has however received only little attention. Yet a wrong model choice may lead to severe performance losses. In this paper, we address the model selection problem in human mobility prediction and make the following contributions. We present SELECTOR, a generic framework to explore human mobility data and compute both population models and individual models to predict human mobility. The former are models that are adapted to the characteristics of an entire population of users and can be used to overcome the cold-start problem. The latter are prediction models optimized for individual users. We present and analyze the results obtainable using SELECTOR on the Nokia data set, which is one of the largest and richest, publicly available data sets of human mobility data. We show that for many users generic population models can be used in place of individual models with negligible performance losses. Yet for about 25% of the users individual models perform at least 3 percentage points better than population models. Thereby, we show that the use of phone context data does not lead to significantly better performance of human mobility predictors with respect to the case in which only temporal and spatial features are used. We further observe that the population models we derive are robust against the demographics of the users and that building different population models for different periods of the day leads to performance improvements. We make SELECTOR publicly available to allow other researchers and practitioners to explore further mobility data sets and to embed the code base of SELECTOR in their applications.

Robot systems largely depend on embedded systems to operate. The interfaces of those embedded systems, e.g. motor controllers or laser scanners, are often vendor-specific and therefore require a component that translates from/to the Robot Operating System (ROS) Middleware interface. In this work we present an implementation and evaluation of a ROS Middleware client based on the Contiki operating systems, which is suitable for constrained embedded devices, like wireless sensor nodes. We show that in-buffer processing of ROS messages without relying on dynamic memory allocation is possible. That message contents can be accessed conveniently via well-known concepts of the C language (structs) with negligible processing overhead compared to a C++-based client. And that the message-passing middle- ware concept of ROS fits nicely in Contiki's event-based nature. Furthermore, in order for an environment enriched with wireless sensor network to help robots in navigating, understanding and manipulating environments a direct integration is mandatory

Electrodermal activity (EDA) is a physiological signal that can be used to infer humans' affective states and stress levels. EDA can nowadays be monitored using unobtrusive wearable devices, such as smartwatches, and leveraged in personal informatics systems. A still largely uncharted issue concerning EDA is the impact on real applications of potential differences observable on signals measured concurrently on the left and right side of the human body. This phenomenon, called lateralization, originates from the distinct functions that the brain's left and right hemispheres exert on EDA. In this work, we address this issue by examining the impact of EDA lateralization in two classification tasks: a cognitive load recognition task executed in the lab and a sleep monitoring task in a real-world setting. We implement a machine learning pipeline to compare the performance obtained on both classification tasks using EDA data collected from the left and right sides of the body. Our results show that using EDA from the side that is not associated with the specific hemisphere activation leads to a significant decline in performance for the considered classification tasks. This finding highlights that researchers and practitioners relying on EDA data should consider possible EDA lateralization effects when deciding on sensor placement.