Urban forests are a proven human health resource, as they restore the physical and mental health of the human body and mind. They have a positive impact on air quality and thermal conditions by reducing concentrations of gaseous and particulate pollutants and lowering air temperature, respectively. In addition, urban forests provide several important ecosystem services, including those associated with human health. Different human-place concepts, show that a close connection of humans to their natural environment is an important determinant of people's well-being. Urban forests, however, vary based on e.g., tree species composition, structure, age and diameter of trees, canopy cover, number and density of canopy layers, abundance of plant species, dead wood, visibility distance, and light conditions. In any human-centered approach, these physical forest characteristics cannot be considered independently of subjective human perception.Thus, besides answering the question of whether different urban/peri-urban forest and open land structures are associated with different local and human bioclimatic characteristics, another major objective of the project is to collect, digitize, process, model and assess data on human physiological effects gathered during field experiments and walking studies in selected study regions within the city and the urban forest of Augsburg, Germany. Thus, the overarching research question of our study is whether "climatic" forest types are also "human physiological" and "therapeutic" forest types.The thermal properties of the study regions will be modelled with the microclimatic model ENVI-met and validated against field measurements of climate variables like air temperature, relative humidity, and wind conditions. If successful, this will allow the calculation of further bioclimatological thermal indices such as physiological equivalent temperature (PET), predicted mean vote (PMV) or universal thermal climate index (UTCI) and the development of silvicultural scenarios. Data on physiological effects on humans will be collected during monitored thermal walks along predefined routes in and near-by the study regions, where participants will be equipped with wearable electronic devices that collect physiological data, such as heart activity. Stress levels of participants along the routes will be assessed by saliva-cortisol probes. Questionnaires will be used to collect sociodemographic data and data on participants perceived thermal and visual sensations during the walks. Subjective thermal sensations will be compared to objectively derived thermal indices based on the model results and mobile measurements taken simultaneously with the thermal walks.Measurements of bioclimatic parameters, human physiological responses, hormone releases, and the recording of subjective well-being as well as subjective perceptions of environmental variables allow for a comprehensive analysis of positive human-environment relationships in urban forests. Thus, a qualitative and quantitative assessment of recreational effects, differentiated by different forest structures, bioclimatic parameters, and social groups, can be comprehensively presented.

Positive effects of forest walking on well-being have been known for a long time. Researchers aim to understand the physiological and psychological effects of forest walking utilizing statistical analysis of physiological data, such as blood pressure, heart rate, and cortisol levels, and psychological data, such as mood and stress levels, from participants before and after forest walking. Recently wearables have been used to continuously monitor and analyze the effect of the forest when compared to the city conditions. However, using statistical methods to demonstrate the effect of forest walking on individual heart rate variability features has been challenging due to various confounding factors. In this study, we used a more comprehensive approach by incorporating a set of over 80 time-domain, frequency-domain, and nonlinear heart rate variability features, and applying dimensionality reduction and using machine learning classifiers to show the difference between the effects of walking in the forest and city. The findings indicate that forest walking can be a practical way to improve both physiological and psychological well-being and that machine learning can be a useful tool for analyzing the complex relationships between different environment and their effects on physiological outcomes.

Many regular meeting participants know the atmosphere of a meeting in a closed room when the air becomes stifling and uncomfortable. With increased concentration of carbon dioxide (CO2), the cognitive performance of participants can also suffer as a result. Therefore, it makes sense to retrofit such meeting rooms with smarthome technology to improve the air quality. However, existing meeting rooms often do not have built-in ventilation systems. Hence, smart technology can be used to measure air quality and communicate that by ambient notifications such as ambient lighting. Actively ventilating a room is then left to the participants of a meeting. This paper reports on a user study demonstrating that the kind of unobtrusive notification from an ambient air quality display has a significant influence on response times of perceiving the information while being concentrated on the course of the meeting.

Contact with nature can help to reduce stress, enhance stress resilience, promote mental and physical health and has a positive impact on people's mood. Beside urban park and residential green, recreation in urban forests can act as therapeutic means. From a climatological point of view, urban and periurban forests and green spaces provide a number of benefits particularly including air temperature and humidity control as well as air pollution reduction. Due to their compensating thermal effects urban green and urban forests may help to counteract potentially health relevant effects of urban warming. The main objective of our study is to explore the quantification of forest recreation based on measurement campaigns for the combined simultaneous recording of relevant features along routes comprising varying urban structural types (ranging from built up to densely forested areas). Combining data on subjective well-being and objective data on human physiology can help to quantify health effects of varying environments. The study area is the urban forest of Augsburg, in the German Federal State of Bavaria, Southern Germany. Our results substantiate clear cut and statistically significant climatic differences among varying urban environments (i.e. local climate zone categories) and prove the potential positive effects of urban forests/urban green on bioclimatic conditions (e.g. via a reduction in maximum air temperatures during summer). Moreover, the beneficial effects of urban green structures on human physiological parameters (e.g. reductions in heart rate) could be verified.

A human's heart beating can be sensed by sensors and displayed for others to see, hear, feel, and potentially "resonate'' with. Previous work in studying interaction designs with physiological data, such as a heart's pulse rate, have argued that feeding it back to the users may, for example support users' mindfulness and self-awareness during various everyday activities and ultimately support their health and wellbeing. Inspired by Somaesthetics as a discipline, we designed and explored multimodal displays, which enable experiencing heart beats as natural stimuli from oneself and others in social proximity. In this paper, we report on the design process of our design PiHearts and present qualitative results of a field study with 30 pairs of participants. Participants were asked to use PiHearts during watching short movies together and report their perceived experience in three different display conditions while watching movies. We found, for example that participants reported significant effects in experiencing sensory immersion when they received their own heart beats as stimuli compared to the condition without any heart beat display, and that feeling their partner's heart beats resulted in significant effects on social experience. We refer to resonance theory to motivate and discuss the results, highlighting the potential of how digitalization of heart beats as rhythmic natural stimuli may provide resonance in a modern society facing social acceleration.

This paper presents a robotic piano tutor which aims to support and motivate students with gamification, hints and feedback. It uses a screen for displaying the musical score, a MIDI keyboard for monitoring the user’s play and a social robot for providing feedback. Musical pieces are divided into four categories with different degrees of difficulty. An adaptation approach based on reinforcement learning is used to optimize the hints for the individual user.

This paper presents an application that classifies forest's aesthetics using interactive machine learning on mobile devices. Transfer learning is used to be able to build upon deep ANNs (MobileNet) using the limited resources available on smart-phones. We trained and evaluated a model using our application based on a data-set that is plausible to be created by a single user. In order to increase the comprehensibility of our model we explore the potential of incorporating explainable Artificial Intelligence (XAI) into our mobile application. To this end we use deep Taylor decomposition to generate saliency maps that highlight areas of the input that were relevant for the decision of the ANN and conducted a user study to evaluate the usefulness of this approach for end-users.

Elderly people could take benefit of speech assistants since they provide the most natural way to interact with assistive technology. Nevertheless, current speech assistants are mainly based on cloud-systems which are non-functional without a stable internet connection or introduce severe privacy-concerns. In this work we provide an overview of state-of-the-art available components for developing a privacy-by-design, open source speech assistant for seniors, as well as a fully functional implementation. We chose the health related task of getting nutrition information of packaged foodstuff for which we rely on an open data database. Our prototype was used in a workshop with two German seniors to gather further insights into the special requirements of elderly people, which can be taken into account for future speech assistants.

The quality of indoor air exerts influence on the wellbeing of people. However, people rarely notice a constant and creeping deterioration of indoor air. Especially in enclosed places where several people get together, like meeting rooms, school rooms and public transportation, bad air quality might cause a reduction of cognitive performance, increased headache, fatigue and sleepiness. This paper describes and discusses a privacy respecting system, built with low cost IoT components and open-source software, that informs room occupants about bad air quality with ambient lights. Additionally, the status of the windows is indicated so that they are not forgotten to be closed after airing. We conducted a workshop with an implementation of the system with five users and present the results which show the usefulness and desirability of ambient notifications for indoor air quality monitoring.

In the future, an increasing amount of social robots will be found in our domestic environments to support and facilitate everyday life. Especially in the context of assistive, health-related support, more and more robotic products are on the way to the consumer market. While one can observe that many commercial efforts are put into the visual appearance, embodiment, motion and sound of companion robots, this paper focuses on the robot's conversational skills. We investigate how to adapt the robot's linguistic style to the individual user's preferences. This includes two forms of robot persona in the context of information retrieval tasks and games, as well as politeness with regard to recommendations. Therefore, we present an autonomous companion robot, which adapts its spoken language based on explicit human feedback. It provides several functionalities for information retrieval, reminders, communication and entertainment as well as health-related recommendations. Results of the in-situ study with elderly participants indicate that human preferences vary with regard to the robot's employed politeness strategies. Furthermore, the participants preferred assistant persona over companion persona in the information retrieval context.

Underwater sounds provide essential information for marine researchers to study sea mammals. During long-term studies large amounts of sound signals are being recorded using hydrophones. To facilitate the time consuming process of manually evaluating the recorded data, computational systems are often employed. Recent approaches utilize Convolutional Neural Networks (CNNs) to analyze spectrograms extracted from the audio signal. In this paper we explore the potential of relevance analysis to enhance the performance of existing CNN approaches. For this purpose, we present a fusion system that utilizes intermediate outputs of three state of the art CNNs, which are fine tuned to recognize whale sounds in spectrograms. Hereby we use Explainable Artificial Intelligence (XAI) to asses the relevance of each feature within the obtained representations. Based on those relevance values, we create novel masking algorithms to extract significant subsets of respective representations. These subsets are used to train an ensemble of classification systems that are serving as input for the final fusion step. We observe that a classification system can benefit from the inclusion of Relevance-based Feature Masking in terms of improved performance and reduced input dimensionality. The presented work is part of the INTERSPEECH 2019 Computational Paralinguistics Challenge.

Gas sensors using the low-priced MOS technique face several problems that reduce their applicability for classification tasks. In literature, several methods can be found that try to alleviate these problems. In this extended abstract, we focus on the problem that the behavior of different sensors (of same type) results in different data even for the same gas concentrations. Hence, the models cannot be reused for sensors of same type. Additionally, it is hardly possible to record high amounts of this type of data which makes transfer learning techniques on such data sets necessary. We apply a calibration transfer procedure and present the results on a data set for food recognition, which we recorded simultaneously with two identical gas sensor boxes in a controlled environment. We compare the results on MOS-based gas sensors from the MQ series and a more modern sensor using MEMS fabrication. The applied technique shows increased recognition rates on this data set.

Socially assistive robots help the human in everyday tasks. In particular, they can also provide assistance with regard to health and wellbeing, which is an important opportunity to support the independence of the elderly. Since spoken language often carries the essential information, the robot's verbal communication style is of special interest. In order to explore individual user's preferences with regard to the robot's linguistic style we built an autonomous, assistive health companion. It is operated with a custom hardware control panel and offers games, jokes, health-related recommendations, applications for information retrieval and communication. This paper reports on insights from the in-situ user study with regard to the system's usability and participants' feedback.

Non-verbal sounds are an essential communication channel for social robots. However, it requires expert knowledge to create and compose synthesizers, develop melodic structures or record samples which express a robot's internal intentions and emotions. This paper presents an approach for adapting a robot's timbre based on non-expert human comparative feedback in order to personalize the sonic interaction design to an individual user's preferences. An evolution strategy learns parameters of real-time sound synthesis for different intentions and emotions. Ultimately, the strategy aims to improve the perceived goodness of how well a specific melody's sound maps to a specific emotion or intention. In order to demonstrate the feasibility of the approach, we report on a user study with a robot, 6 exemplary melodies and 27 participants. Our study results show that the strategy indeed results in improved and preferred sound designs and that many participants are willing to apply such a process to improve their robots' expressivity.

This work outlines a concept and the necessary building blocks for creating a persuasive and personalized robotic dietitian for everyday health-related support based on existing technology and recent research insights. Key components include natural language generation for the social robot's linguistic style, mobile sensing hardware for tracking nutrition, and machine learning for adaptation.

Designing for wellbeing and a better quality of life is a challenge due to individual differences and contextual influences. How a person feels may depend on inner-body and outer-body circumstances. Moreover, some places may foster while others hinder wellbeing. The objective of this paper is to identify typical patterns across characteristics of the urban environment, the user's self-reported wellbeing as well as physiological measures as indicators of wellbeing. To this end, we conducted a study with 7 participants in an urban environment covering different kinds of climate zones. In addition to multi-sensor data collection, participants were asked to provide location-specific experience samples with the help of off-the-shelf wrist accessories and a smartphone. The paper at hand presents classification results for the collected data. It furthermore analyzes the limitations of the approach and discusses the potential for future therapeutic applications that enhance the value of urban green by making use of sensory technologies.

Social robots become increasingly important in the domain of healthcare and maintenance. Nutrition is not an exception: research robots are used in several experiments to teach people about healthy nutrition. Moreover, robotic products, whose task is to keep track of the user's nutrition, to provide tips, reminders, and to recognize anomalies in health-related behaviors are on the way to market. To convince users of a robot's recommendations, speech is an important interaction modality. However, automatically adapting a robot's spoken advise depending on users' behaviors is still a challenge. We address this issue by building Drink-O-Mender, an interactive installation, which includes a Reeti robot augmented with additional sensing and adaptation abilities. The robot is designed to offer drinks in a social setting. It aims to convince users of consuming healthy drinks while adapting its spoken advices depending on the users' selected beverage choices. The installation is equipped with custom hardware including a smartscale to sense the type and quantity of consumed drinks. We describe the interactive installation in detail and demonstrate feasibility of generating adaptive spoken advices by reporting on insights gained from exhibiting the installation during a public event and observing interactions of 78 users with the robot.

More people could benefit of Machine Learning (ML) as an increasingly important technology and service, if state-of-the-art ML techniques with training capability were accessible on personal devices. To this end, we report details on how to deploy Tensor-Flow on off-the-shelf mobile and embedded devices and retrain current deep neural networks for image recognition on-device. Our motivation is to both grant privacy and allow users to efficiently personalize image classifiers for their own needs and purposes, and thus contribute towards turning ML into a "social good", which benefits the largest number of people in the greatest possible way.

In the past, the performance of machine learning algorithms depended heavily on the representation of the data.Well-designed features therefore played a key role in speech and paralinguistic recognition tasks.Consequently, engineers have put a great deal of work into manually designing large and complex acoustic feature sets.With the emergence of Deep Neural Networks (DNNs), however, it is now possible to automatically infer higher abstractions from simple spectral representations or even learn directly from raw waveforms.This raises the question if (complex) hand-crafted features will still be needed in the future.We take this year's INTERSPEECH Computational Paralinguistic Challenge as an opportunity to approach this issue by means of two corpora -Atypical Affect and Crying.At first, we train a Recurrent Neural Network (RNN) to evaluate the performance of several hand-crafted feature sets of varying complexity.Afterwards, we make the network do the feature engineering all on its own by prefixing a stack of convolutional layers.Our results show that there is no clear winner (yet).This creates room to discuss chances and limits of either approach.

This paper presents preliminary results of the ongoing project TheOdor which explores the potential of electronic noses that make use of commodity gas sensors (MOS, MEMS) for applications in the smarthome, for example, to classify human activities based on the odors generated by activities. We describe the system and its components and report on classification results from first validation experiments.