Search

At the times of obesity epidemic, innovative ways to nudge food shoppers toward healthier choices are needed. Key ideas from the conceptual metaphor processing, food symbolism and contextual priming literatures are synthetized to argue for the idea that visual, olfactory and gustatory cues can activate shoppers’ well-being goal and guide their in-store movements and healthy food purchasing behaviors. Two field experiments, separated by an 8-month interval, are conducted in the same grocery store. The results show that priming the well-being goal in real store can push consumers to move toward spots where healthier options are available – materializing also into greater sales of yogurts, vegetables and heart healthy-labeled products. Counterintuitively, the well-being goal activation possibly first facilitate approaching indulgent foods. Yet, the actual sales of palatable foods such as ice cream and minced beef meat may in fact diminish. This research spurs numerable academic and practical implications.

Sustainable work aims at improving working conditions to allow workers to effectively extend their working life. In this context, occupational safety and well-being are major concerns, especially in labor-intensive fields, such as construction-related work. Internet of Things and wearable sensors provide for unobtrusive technology that could enhance safety using human activity recognition techniques, and has the potential of improving work conditions and health. However, the research community lacks commonly used standard datasets that provide for realistic and variating activities from multiple users. In this article, our contributions are threefold. First, we present VTT-ConIoT, a new publicly available dataset for the evaluation of HAR from inertial sensors in professional construction settings. The dataset, which contains data from 13 users and 16 different activities, is collected from three different wearable sensor locations.Second, we provide a benchmark baseline for human activity recognition that shows a classification accuracy of up to 89% for a six class setup and up to 78% for a sixteen class more granular one. Finally, we show an analysis of the representativity and usefulness of the dataset by comparing it with data collected in a pilot study made in a real construction environment with real workers.

Sustainable work aims at improving working conditions to allow workers to effectively extend their working life. In this context, occupational safety and well-being are major concerns, especially in labor-intensive fields, such as construction-related work. Internet of Things and wearable sensors provide for unobtrusive technology that could enhance safety using human activity recognition techniques, and has the potential of improving work conditions and health. However, the research community lacks commonly used standard datasets that provide for realistic and variating activities from multiple users. In this article, our contributions are threefold. First, we present VTT-ConIoT, a new publicly available dataset for the evaluation of HAR from inertial sensors in professional construction settings. The dataset, which contains data from 13 users and 16 different activities, is collected from three different wearable sensor locations.Second, we provide a benchmark baseline for human activity recognition that shows a classification accuracy of up to 89% for a six class setup and up to 78% for a sixteen class more granular one. Finally, we show an analysis of the representativity and usefulness of the dataset by comparing it with data collected in a pilot study made in a real construction environment with real workers.

Terrorism is an international security challenge. The early detection of threats (e.g., explosives or firearms) could provide a valuable contribution to the ability to prevent, protect and respond to terrorism. This paper presents a system for the management of a plurality of sensors to improve the threat-detection capabilities without disrupting the flow of passengers. The system improves the prevention capabilities of soft targets (such as airports, undergrounds and railway stations) with a high number of daily commuters. The system architecture consists of three main components. The first component is 2D video tracking and re-identification (Re-ID), which allows the labelling and tracking of commuters in a small area. Thereby, it supports the fusion of sensors at different locations. The Re-ID has a smart training strategy with anonymized snippets to increase flexibility for new environments. The second component is 3D video tracking with a stereo camera, which gives a more accurate location measurement than 2D video. Location prediction is used to compensate for latency in the control of active elements in the threat detection sensor. Recurrent neural networks for location prediction were trained by using real 3D tracking data from a railway station. The performance is evaluated with a ground-truth based on Ultra-Wide Band (UWB) radio positioning and a coordinate conversion method was created to compensate for identified inaccuracies. The third component is Command & Control (C&C), which consists of three submodules: message broker, data-fusion and security client. The message broker is a publish-subscribe middleware layer to enable flexible integration of the various sensors and components. The data-fusion combines outputs of multiple sensors. In case of a suspect person, the security client triggers an alarm and a comprehensive report is sent to the security guards.

The impending exhaustion of non-renewable natural resources emphasizes the importance of sustainable materials. One environmentally friendlier alternative to plastics are biocomposite materials, which are composed of recycled or virgin polymer and a natural fiber material. In order to be to be truly sustainable, the material has to be efficiently recycled, a task which demands a means for their sorting. This paper outlines a method for distinguishing different plastic and biocomposite samples from one another based on hyperspectral imaging. The developed regression model correctly classified 96 % of the samples in the dataset. In the case of biocomposite samples, the accompanying polymer was quite accurately recognized.

Demand for efficient ways to represent vast amount of video data has grown rapidly in recent years. The advances in positioning services have led to new possibilities in combining location information to video content. In this paper we present an automatic video editing system for geotagged mobile videos. In our solution the system creates automatically a video summary from a set of unedited video clips. Location information and timestamps are used to group video clips with the same context properties. The groups are used to create a video summary where subshots from same context group are represented as scenes. The novelty in our solution lies in combining geotags with low level content analysis tools in video abstraction. We have evaluated the created video summaries with a group of users and the system usability for service creation by building a semi-automatic web-based video editing service. The evaluations prove that our concept is useful as it improves coherence and enjoyability of the automatic video summaries.

Intelligent computer applications need to adapt their behaviour to contexts and users, but conventional classifier adaptation methods require long data collection and/or training times. Therefore classifier adaptation is often performed as follows: at design time application developers define typical usage contexts and provide reasoning models for each of these contexts, and then at runtime an appropriate model is selected from available ones. Typically, definition of usage contexts and reasoning models heavily relies on domain knowledge. However, in practice many applications are used in so diverse situations that no developer can predict them all and collect for each situation adequate training and test databases. Such applications have to adapt to a new user or unknown context at runtime just from interaction with the user, preferably in fairlylightweightways, that is, requiring limited user effort to collect training data and limited time of performing the adaptation. This paper analyses adaptation trends in several emerging domains and outlines promising ideas, proposed for making multimodal classifiers user-specific and context-specific without significant user efforts, detailed domain knowledge, and/or complete retraining of the classifiers. Based on this analysis, this paper identifies important application characteristics and presents guidelines to consider these characteristics in adaptation design.

Automatic classification of everyday activities can be used for promotion of health-enhancing physical activities and a healthier lifestyle. In this paper, methods used for classification of everyday activities like walking, running, and cycling are described. The aim of the study was to find out how to recognize activities, which sensors are useful and what kind of signal processing and classification is required. A large and realistic data library of sensor data was collected. Sixteen test persons took part in the data collection, resulting in approximately 31 h of annotated, 35-channel data recorded in an everyday environment. The test persons carried a set of wearable sensors while performing several activities during the 2-h measurement session. Classification results of three classifiers are shown: custom decision tree, automatically generated decision tree, and artificial neural network. The classification accuracies using leave-one-subject-out cross validation range from 58 to 97% for custom decision tree classifier, from 56 to 97% for automatically generated decision tree, and from 22 to 96% for artificial neural network. Total classification accuracy is 82 % for custom decision tree classifier, 86% for automatically generated decision tree, and 82% for artificial neural network.

Demand for efficient ways to represent vast amount of video data has grown rapidly in recent years. The advances in positioning services have led to new possibilities of combining location information to video content. In this paper we present an automatic video editing system for geotagged mobile videos. In our solution the system creates automatically a video summary from a set of unedited video clips. Geotags are used to group video clips with the same context properties. The groups are used to create a video summary where the videos from the same group are represented as scenes. The novelty in our solution lies in the combining of geotags with low level content analysis tools in video abstraction. Evaluations of the system prove the concept useful as it improves coherence and enjoyability of the automatic video summaries.

This paper presents a solution for creating light-weighted content and context-aware mobile multimedia services. The main application domain is in user created multimedia content and experience sharing. Our approach is to have a platform supporting state-of-the art content management functionalities in order to enable easy creation of specialized multimedia services for various target groups and purposes. We have developed a mobile multimedia content creation platform with integrated context metadata support. To verify the overall functionality of our platform we have defined a multimedia content service template and created a set of exemplary services using Web-based technologies such as JavaScript, media RSS feed and Java. (11 refs.)

This work presents a software framework for real time multimodal affect recognition. The framework supports categorical emotional models and simultaneous classification of emotional states along different dimensions. The framework also allows to incorporate diverse approaches to multimodal fusion, proposed by the current state of the art, as well as to adapt to context-dependency of expressing emotions and to different application requirements. The results of using the framework in audio-video based emotion recognition of an audience of different shows (this is a useful information because emotions of co-located people affect each other) confirm the capability of the framework to provide desired functionalities conveniently and demonstrate that use of contextual information increases recognition accuracy. (21 refs.)

State-of-the-art automatic analysis tools for personal audio con-tent management are discussed in this paper. Our main target is to create a system, which has several co-operating management tools for audio database and which improve the results of each other. Bayesian networks based audio classification algorithm provides classification into four main audio classes (silence, speech, music, and noise) and serves as a first step for other subsequent analysis tools. For speech analysis we propose an improved Bayesian information criterion based speaker segmen-tation and clustering algorithm applying also a combined gender and emotion detection algorithm utilizing prosodic features. For the other main classes it is often hard to device any general and well functional pre-categorization that would fit the unforesee-able types of user recorded data. For compensating the absence of analysis tools for these classes we propose the use of efficient audio similarity measure and query-by-example algorithm with database clustering capabilities. The experimental results show that the combined use of the algorithms is feasible in practice.

This paper describes a demonstrated concept implementation that combines sound source localization and face detection from video stream for supporting human behavior analysis. System monitors space containing multiple persons using microphone array and video camera. The aim is to detect which person in the scene is producing the sound that is received by the microphones. For this task the microphone array localizes the sound in the environment. Simultaneously face detection is performed to the video signal produced by the monitoring video camera. If face is detected from the bearing of the sound the system may decide that the sound is produced by the person who's face is detected. Preliminary results indicate that the fusion may give useful information for human behavior analysis for space containing multiple persons.

In this paper, we propose a fuzzy logic-based application layer controlling algorithm for wireless multimedia communication to adjust the bit-rate of transmitted video according to cross-layer feedback information. To further improve the fast adaptation to possible bit rate adjustment proposed by the cross-layer controller, a general enhancement to the rate control algorithm of the video encoder based on the motion activity of the video sequence is also introduced. The proposed cross-layer controlling algorithm for wireless communication has been integrated into the cross-layer simulation platform and the results achieved when using the algorithm with H.264/AVC show that it provides a good and fast adaptability into changing wireless channel conditions and an enhancement in video quality is achieved compared to non-adapted situation.

The aim of this study was to examine whether using soft biometrics, i.e. easily measurable personal characteristics, such as weight and fat percentage, can improve the performance of biometrics in verification type applications. Fusing fingerprint biometrics with soft biometrics, in this case body weight measurements, decreased the total error rate (TER) from 3.9% to 1.5% in an experiment with 62 test subjects. This result shows that simple physiological measurements can be used to support biometric recognition. Furthermore, soft biometrics are unobtrusive, there is no risk of identity theft, the perception of the big-brother effect is small, the equipment needed is low-cost, and the methods are easy to understand. Soft biometrics alone are not suitable for security related applications, but they can be used for improving the performance of traditional biometrics. A potentially feasible use for soft biometrics may be found in non-security, convenience type cases, such as domestic applications.

Video management research has been neglecting the increased attractiveness of using camera-equipped mobile phones for the production of short home video clips. But specific capabilities of modern phones — especially the availability of rich context data — open up new approaches to traditional video management problems, such as the notorious lack of annotated metadata for home video content. In this chapter, we present MobiCon, a mobile, context-aware home video production tool. MobiCon allows users to capture video clips with their camera phones, to semi-automatically create MPEG-7-conformant annotations by exploiting available context data at capture time, to upload both clips and annotations to the users’ video collections, and to share these clips with friends using OMA DRM. Thereby, MobiCon enables mobile users to effortlessly create richly annotated home video clips with their camera phones, paving the way to a more effective organization of their home video collections.

Multimedia applications such as video conference, digital video broadcasting (DVB), and streaming video and audio have been gaining popularity during the last years, growing also among mobile users. Quality of Service requirements for multimedia over wireless raises huge challenges not only concerning the physical bandwidth, but also the network design and services. This has to be addressed by future 4G systems. To this end, joint source and channel coding (JSCC) has provided promising results; however, the proposed mechanisms typically do not provide IP network support. In this paper we present an architecture for JSCC systems designed for an IP network with end-to-end overall optimisation. All necessary mechanisms for transmitting the joint optimisation control information over wired and wireless network using cross-layer design are also described. Finally, partial simulation results obtained with this system design and H.264/AVC video input are presented.

The purpose of this paper is to study the performance of the partial checksum mechanism of UDP-Lite transport protocol compared to the standard UDP protocol in wireless environment in terms of achieved video quality. Video quality is evaluated in PSNR using simulation results achieved with a simulation platform developed in the framework of IST PHOENIX project. The platform offers the necessary components for modelling all relevant ISO/OSI layers for wireless IP video streaming, including partial checksum functionality on transport and data link layers. The achieved gain in video quality in terms of PSNR for IEEE 802.11a simulation model validates that the video streaming benefits of using UDP-Lite transport protocol instead of using regular UDP in wireless channels. The enhancement was 2 - 3 dB in PSNR for AWGN channel and for fading channels the difference was smaller, being 1dB or less. The major conclusion of our study is that the best improvement in video quality using UDP-Lite is achieved in conditions where bit errors appear in very short bursts. Also a minor enhancement using UDP-Lite is achieved in transmission conditions where signal-to-noise ratio in the channel is very low.

MediaTeam Oulu and VTT Technical Research Centre of Finland participated jointly in semantic feature extraction, manual search and interactive search tasks of TRECVID 2003. We participated to the semantic feature extraction by submitting results to 15 out of the 17 defined semantic categories. Our approach utilized spatio-temporal visual features based on correlations of quantized gradient edges and color values together with several physical features from the audio signal. Most recent version of our Video Browsing and Retrieval System (VIRE) contains an interactive cluster-temporal browser of video shots exploiting three semantic levels of similarity: visual, conceptual and lexical. The informativeness of the browser was enhanced by incorporating automatic speech transcription texts into the visual views based on shot key frames. The experimental results for interactive search task were obtained by conducting a user experiment of eight people with two system configurations: browsing by (I) visual features only (visual and conceptual browsing was allowed, no browsing with ASR text) or (II) visual features and ASR text (all semantic browsing levels were available and ASR-text content was visible). The interactive results using ASR-based features were better than the results using only visual features. This indicates the importance of successful integration of both visual and textual features for video browsing. In contrast to previous version of VIRE which performed early feature fusion by training unsupervised self-organizing maps, newest version capitalises on late fusion of features queries, which was evaluated in manual search task. This paper gives an overview of the developed system and summarises the results.

In TREC 2002 Video Track MediaTeam Oulu and VTT Technical Research Centre of Finland participated jointly in semantic feature extraction, manual search and interactive search tasks. In the semantic feature extraction task, we sent results for semantic categories of cityscape, landscape, people, speech and instrumental sound. Spatio-temporal correlation of oriented gradient occurrences was used with example shots to detect shots containing people, cityscape or landscape. The audio signal features consisted of various statistical measurements and were used to detect shots containing speech or instrumental sound. Our videobrowsing and retrieval system, VIRE, was used for manual and interactive search tasks. Our system offers two techniques for video retrieval: 1. Multi-modal indexing based on self-organizing feature maps with semantic filtering. 2. An interactive navigating tool that combines two inter-shot properties, temporalcoherency and metric similarities, into a view where database shots are presented in a lattice structure. We tested our interactive navigating tool with eight persons to obtain results for 25 pre-defined search topics. In this paper we give an overview of the approaches and a summary of the results.

Modern buildings are expected to fulfill energy efficiency regulations while providing a healthy and comfortable living environment for their occupants. Demand-controlled heating, ventilation, and air conditioning can improve energy efficiency and well-being, especially if changes in indoor environmental factors can be forecast reliably enough. As a first step, this study afforded a baseline for CO2 forecasting evaluation by providing for the public use a comprehensive dataset covering a full year. Secondly, we studied the applicability of four machine learning methods, Ridge regression, Decision Tree, Random Forest, and Multilayer Perceptron, for modeling the future concentration of CO2 in indoors. We evaluated their prediction accuracy within different forecasting and history window time frames and the impact of multiple sensor modalities. All models performed better than the baseline method of predicting the last observed value, and the Decision Tree was found to be almost as accurate as the computationally heavier Random Forest model. When the future forecasting window was longer than a minute, the optimal sensor modalities included occupant activity data from passive infrared sensors in addition to CO2 concentration. Our findings suggest that machine learning can be applied in multimodal time-series data to find a simple, accurate, and resource-efficient forecasting model for proactive control of indoor environments.