Wearable inertial measurement unit (IMU)-based human activity recognition (HAR) is a key enabler of mobile and pervasive computing applications. To address the scarcity of labeled data, self-supervised contrastive learning has been adopted. It leverages data augmentations to generate diverse views of sensor data, using them as positive pairs in a pretext task to learn representations from unlabeled data. However, conventional signal transformation-based data augmentations (STDAs) are poorly grounded in the realistic variability that affects IMU readings, e.g., diverse motion patterns and device positioning, often resulting in unstable downstream task performance. In this work, we investigate the effectiveness of ensuring physical plausibility in data augmentations within contrastive learning. Specifically, we compare physically plausible data augmentations (PPDAs), realized via physics simulation to incorporate domain knowledge, against legacy STDAs. Evaluated on three public datasets covering daily activities and fitness workouts (8 to 34 classes), models pretrained with PPDAs consistently outperform STDA-pretrained models and fully supervised baselines, improving the downstream classification macro-F1 score by 3 to 14 percentage points. Qualitative analyses reveal that PPDAs shape the learned representation space into a more physically grounded topology. Furthermore, PPDAs achieve fully supervised-level performance with less than 10% of labeled data. We also demonstrate the robustness of physically grounded augmentations, showing they remain effective with lower-fidelity simulation data and generalize across different encoder architectures and contrastive learning frameworks. These findings suggest that incorporating physical domain knowledge to ensure physical plausibility enables the generation of semantically consistent positive pairs, helping models learn representations more relevant to HAR.

Wearable mobile devices demand seamless connectivity with low power consumption for improved quality of user experience. Capacitively coupled communication that is based on sensing a low frequency modulated electric field (several MHz), is a promising alternative to conventional electromagnetic wave radio links. However, achieving reliable over-the-air as well as air-to-body communication through electric field remains a research challenge as the presence of the human body influences the performance of an e-field communication channel. Therefore, in this work, we first benchmark a system for air-to-air communication and then characterise the effect of orientation and distance between the transmitter and receiver nodes when the receiver node is worn on the body. The presented work is an adaptation of a system previously developed for a standalone air-to-air communication scenario [16]. As an application, we integrate the receiver node into commercial overhead headphones to create a wireless audio system, recovering and demodulating a binary frequency shift keying (BFSK) digital audio signal. The signal is transmitted through the varying electric field resulting from a 4.45 MHz alternating voltage applied to the transmitter electrode. We present the bit error rate (BER) performance of the channel for various on-body electrode configurations and transmitter-receiver orientations. The system is capable of transmitting 8-bit mono audio at a sample rate of 16 kHz (128 kbps) with a BER of less than 10−6.7at 50 cm, and less than 10−4at 75 cm,

Wearable and home IoT applications require low-power connectivity to maximize battery life. Sensing a modulated electric field, referred to as capacitively coupled communication, is a promising alternative to communication via electromagnetic waves. However, achieving a reliable over-the-air communication through electric field remains a research challenge. We describe an efficient electric field based over-the-air communication system using frequency shift keying (FSK) which achieves arm’s length communication range with throughput suitable for real-time sensor streaming or even audio streaming. This is achieved through a highly sensitive electric potential receiver allowing communication when weakly coupled, such as between two devices far from any other object or human body. The digital FSK receiver uses an undersampling technique to reduce the complexity of the implementation. Another contribution of this work is the use of low-power ARM microcontroller to perform the modulation/demodulation, with a few additional off-the-shelf digital and analog components. We present a detailed performance analysis of the system when varying the transmit voltage and communication distance in five different scenarios/environments. The performance is analyzed and compared in terms of bit error rate (BER) and throughput. The system is capable of providing reliable communication link up to 1.2m with a user throughput of 75 kbps in the best case scenario, when placed over tarmac road. It is also shown that the system achieves a user throughput of at least 80 kbps for a distance of 70cm in the worst case scenario, when both the devices are hanged in air through strings.

Template matching methods can benefit from multi-cores architecture in order to parallelise and accelerate the matching of multiple templates. We present WLCSSCuda: a GPU accelerated implementation of the Warping Longest Common Subsequence (WLCSS) pattern recognition algorithm. We evaluate our method on 4 NVIDIA GPUs and 4 multi-cores CPUs. We observe a 67-times speedup for the GPU implementation in the best case against the multithreaded CPU implementation.

Smartphone-based identification of the mode of transportation of the user is important for context-aware services. We investigate the feasibility of recognizing the 8 most common modes of locomotion and transportation from the sound recorded by a smartphone carried by the user. We propose a convolutional neural network based recognition pipeline, which operates on the short-time Fourier transform (STFT) spectrogram of the sound in the log domain. Experiment with the Sussex-Huawei locomotion-transportation (SHL) dataset on 366 hours of data shows promising results where the proposed pipeline can recognize the activities Still, Walk, Run, Bike, Car, Bus, Train and Subway with a global accuracy of 86.6%, which is 23% higher than classical machine learning pipelines. It is shown that sound is particularly useful for distinguishing between various vehicle activities (e.g. Car vs Bus, Train vs Subway). This discriminablity is complementary to the widely used motion sensors, which are poor at distinguish between rail and road transport.

We have completed the collection of one of the richest accurately annotated mobile dataset of modes of transportation and locomotion. To do this, we developed a highly reliable Android application called DataLogger capable of recording multisensor data from multiple synchronized smartphones simultaneously. The application allows real-time data annotation. We explain how we designed the app to achieve high reliability and ease of use. We also present an evaluation of the application in a big-data collection (750 hours, 950 GB of data, 17 different sensor modalities), analysing the data loss (less than 0.4%) and battery consumption (≈ 6% on average per hour). The application is available as open source.

We explain how to obtain a highly versatile and precisely annotated dataset for the multimodal locomotion of mobile users. After presenting the experimental setup, data management challenges and potential applications, we conclude with the best practices for assuring data quality and reducing loss. The dataset currently comprises 7 months of measurements, collected by smartphone's sensors and a body-worn camera, while the 3 participants used 8 different modes of transportation. It comprises 950 GB of sensor data, which corresponds to 750 hours of labelled data. The obtained data will be useful for a wide range of research questions related to activity recognition, and will be made available to the community1.

With growing needs to better understand our environments, the Internet-of-Things (IoT) is gaining importance among information and communication technologies. IoT will enable billions of intelligent devices and networks, such as wireless sensor networks, to be connected and integrated with computer networks. In order to support large scale networks, IETF has defined the routing protocol for low power and lossy networks (RPL) to facilitate the multihop connectivity. In this paper, we provide an in-depth review of current research activities. Specifically, the large scale simulation development and performance evaluation under various objective functions and routing metrics are pioneering works in RPL study. The results are expected to serve as a reference for evaluating the effectiveness of routing solutions in large scale IoT use cases.

We present a way to annotate cystoscopy finding on Google Glass in a reproducible and hands free manner for use by surgeons during operations in the sterile environment inspired by the current practice of hand-drawn sketches. We developed three data entry variants based on speech and head movements. We assessed the feasibility, benefits and drawbacks of the system with 8 surgeons and Foundation Doctors having up to 30 years' cystoscopy experience at a UK hospital in laboratory trials. We report data entry speed and error rate of input modalities and contrast it with the participants' feedback on their perception of usability, acceptance, and suitability for deployment. The results are supportive of new data entry technologies and point out directions for future improvement of eyewear computers. The findings can be generalised to other endoscopic procedures (e.g. OGD/laryngoscopy) and could be included within hospital IT in the future.

The machine recognition of user trajectories and activities is fundamental to devise context-aware applications for support and monitoring in daily life. So far, tracking and activity recognition were mostly considered as orthogonal problems, which limits the richness of possible context inference. In this work, we introduce the novel unified computational and representational framework S-SMART that simultaneously models the environment state (semantic mapping), localizes the user within this map (tracking), and recognizes interactions with the environment (activity recognition). Thus, S-SMART identifies which activities the user executes where (e.g., turning a handle next to a window ), and reflects the outcome of these actions by updating the world model (e.g., the window is now open ). This in turn conditions the future possibility of executing actions at specific places (e.g., closing the window is likely to be the next action at this location). S-SMART works in a self-contained manner and iteratively builds the semantic map from wearable sensors only. This enables the seamless deployment to new environments. We characterize S-SMART in an experimental dataset with people performing hand actions as part of their usual routines at home and in office buildings. The framework combines dead reckoning from a foot-worn motion sensor with template-matching-based action recognition, identifying objects in the environment (windows, doors, water taps, phones, etc.) and tracking their state (open/closed, etc.). In real-life recordings with up to 23 action classes, S-SMART consistently outperforms independent systems for positioning and activity recognition, and constructs accurate semantic maps. This environment representation enables novel applications that build upon information about the arrangement and state of the user’s surroundings. For example, it may be possible to remind elderly people of a window that they left open before leaving the house, or of a plant they did not water yet, using solely wearable sensors.

We describe a qualitative study investigating the acceptability of the Google Glass eyewear computer to people with Parkinson's disease (PD). We held a workshop with 5 PD patients and 2 carers exploring perceptions of Glass. This was followed by 5-day field trials of Glass with 4 PD patients, where participants wore the device during everyday activities at home and in public. We report generally positive responses to Glass as a device to instil confidence and safety for this potentially vulnerable group. We also raise concerns related to the potential for Glass to reaffirm dependency on others and stigmatise wearers.

Activity recognition systems based on body-worn motion sensors suffer from a decrease in performance during the deployment and run-time phases, because of probable changes in the sensors (e.g. displacement or rotatation), which is the case in many real-life scenarios (e.g. mobile phone in a pocket). Existing approaches to achieve robustness tend to sacrifice information (e.g. by rotation-invariant features) or reduce the weight of the anomalous sensors at the classifier fusion stage (adaptive fusion), ignoring data which might still be perfectly meaningful, although different from the training data. We propose to use adaptation to rebuild the classifier models of the sensors which have changed position by a two-step approach: in the first step, we run an anomaly detection algorithm to automatically detect which sensors are delivering unexpected data; subsequently, we trigger a system self-training process, so that the remaining classifiers retrain the “anomalous” sensors. We show the benefit of this approach in a real activity recognition dataset comprising data from 8 sensors to recognize locomotion. The approach achieves similar accuracy compared to the upper baseline, obtained by retraining the anomalous classifiers on the new data.

We present SmartActionSLAM, an Android smartphone application that performs location tracking in home and office environments. It uses the integrated motion sensors of the smartphone and an optional foot-mounted inertial measurement unit to track a person. The application implements an instance of the ActionSLAM algorithmic framework. ActionSLAM combines pedestrian dead reckoning with the observation of activities (in SmartActionSLAM: sitting and standing still) to build and update a local landmark map of the user's environment. This map is used to compensate for error accumulation of dead reckoning in a particle filter framework. We show that it is possible to execute the ActionSLAM algorithm in real-time on a smartphone without platform-specific optimizations. Furthermore, we analyze the localization performance of the application in six constrained and two real-life recordings. When using only the smartphone's internal sensors, tracking was adequate in most constrained setups, but failed in the real-world scenarios because of errors in recognizing irregular leg movements. By including the foot-mounted sensor, mapping with a mean landmark positioning error of90% was achieved in all environments. Smart-ActionSLAM is fully wearable and requires no infrastructure in the environment. The approach is therefore ideally suited for rapid deployment in home and office environments, as for example required in patient monitoring studies.

Activity annotation in videos is necessary to create a training dataset for most of activity recognition systems. This is a very time consuming and repetitive task. Crowdsourcing gains popularity to distribute annotation tasks to a large pool of taggers. We present for the first time an approach to achieve good quality for activity annotation in videos through crowdsourcing on the AmazonMechanical Turk platform (AMT). Taggers must annotate the start, end boundaries and the label of all occurrences of activities in videos. Two strategies to detect non-serious taggers according to temporal annotated results are presented. Individual filtering checks the consistence in the answers of each tagger with the characteristic of dataset to identify and remove nonserious taggers. Collaborative filtering checks the agreement in annotations among taggers. The filtering techniques detect and remove non-serious taggers and finally, the majority voting applied to AMT temporal tags to generate one final AMT activity annotation set. We conduct the experiments to get activity annotation from AMT on a subset of two rich datasets frequently used in activity recognition. The results show that our proposed filtering strategies can increase the accuracy by up to 40%. The final annotation set is of comparable quality of the annotation of experts with high accuracy (76% to 92%).

The vast availability of mobile phones with built-in movement and location sensors enable the collection of detailed information about human movement even indoors. As mobility is a key element of many processes and activities, an interesting class of information to extract is movement patterns that quantify how humans move, interact and group. In this paper we propose methods for detecting two common pedestrian movement patterns, namely individual following relations and group leadership. The proposed methods for identifying following patterns employ machine learning on features derived using similarity analysis on time lagged sequences of WiFi measurements containing either raw signal strength values or derived locations. To detect leadership we combine the individual following relations into directed graphs and detect leadership within groups by graph link analysis. Methods for detecting these movement patterns open up new possibilities in — amongst others — computational social science, reality mining, marketing research and location-based gaming. We provide evaluation results that show error rates down to 7%, improving over state of the art methods with up to eleven percentage points for following patterns and up to twenty percentage points for leadership patterns. Our method is, contrary to state of the art, also applicable in challenging indoor environments, e.g., multi-story buildings. This implies that even quite small samples allow us to detect information such as how events and campaigns in multistory shopping malls may trigger following in small groups, or which group members typically take the lead when triggered by e.g. commercials, or how rescue or police forces act during training exercises.

Previous work on the recognition of human movement patterns has mainly focused on movements of individuals. This paper addresses the joint identification of the indoor movement of multiple persons forming a cohesive whole - specifically a flock - with clustering approaches operating on features derived from multiple sensor modalities of modern smartphones. Automatic detection of flocks has several important applications, including evacuation management and socially aware computing. The novelty of this paper is, firstly, to use data fusion techniques to combine several sensor modalities (WiFi, accelerometer and compass) to improve recognition accuracy over previous unimodal approaches. Secondly, improve the recognition of flocks using hierarchical clustering. We use a dataset comprising 16 subjects forming one to four flocks walking in a building on single and multiple floors. With the best settings, we achieve a F-score accuracy of up to 87 percent an improvement of up to twelve percent points over existing approaches.

Fingerprinting-based localization methods relying on WiFi and GSM information provide sufficient localization accuracy for many mobile phone applications. Most of the existing approaches require a training set consisting of geo-referenced fingerprints to build a reference database. We propose a collaborative, semi-supervised WiFi+GSM fingerprinting method where only a small fraction of all fingerprints needs to be geo-referenced. Our approach enables indexing of areas in the absence of GPS reception as often found in urban spaces and indoors without manual labeling of fingerprints. The method takes advantage of the characteristic that the similarity of two fingerprints correlates to the distance between their corresponding location. By applying multidimensional scaling, a topology estimation is generated and with the help of a small set of geo-referenced fingerprints anchored to physical locations. An evaluation with an urban-scale data set shows that we can locate a mobile device with a median error of 30m. While normally all fingerprints of the training set need to be geo-referenced, with our method, only 8% require geo-referencing. We further show that the localization error decreases as new fingerprints are added and converges to an accuracy comparable to related work.

Indoor localization at minimal deployment effort and with low costs is relevant for many ambient intelligence and mobile computing applications. This paper presents ActionSLAM, a novel approach to Simultaneous Localization And Mapping (SLAM) for pedestrian indoor tracking that makes use of body-mounted sensors. ActionSLAM iteratively builds a map of the environment and localizes the user within this map. A foot-mounted Inertial Measurement Unit (IMU) keeps track of the user's path, while observations of location-related actions (e.g. door-opening or sitting on a chair) are used to compensate for drift error accumulation in a particle filter framework. Location-related actions are recognizable from body-mounted IMUs that are often used in ambient-assisted living scenarios for context awareness. Thus localization relies only on on-body sensing and requires no ambient infrastructure such as Wi-Fi access points or radio beacons. We characterize ActionSLAM on a dataset of 1.69km walking in three rooms and involving 241 location-related actions. For the experimental dataset, the algorithm robustly tracked the subject with mean error of 1.2m. The simultaneously built map reflects the building layout and positions landmarks with a mean error of 0.5m. These results were achieved with a simulated action recognition system consisting of an IMU attached to the wrist of a user and a smartphone in his pocket. We found that employing more complex action recognition is not beneficial for ActionSLAM performance. Our findings are supported by evaluations in synthetic environments through simulation of IMU signals for walks in typical home scenarios.

Template matching methods using Dynamic Time Warping (DTW) have been used recently for online gesture recognition from body-worn motion sensors. However, DTW has been shown sensitive under the strong presence of noise in time series. In sensor readings, labeling temporal boundaries of daily gestures precisely is rarely achievable as they are often intertwined. Moreover, the variation in daily gesture execution always exists. Therefore, here we propose two template matching methods utilizing the Longest Common Subsequence (LCSS) to improve robustness against such noise for online gesture recognition. Segmented LCSS utilizes a sliding window to define the unknown boundaries of gestures in the continuous coming sensor readings and detects efficiently a possibly shorter gesture within it. WarpingLCSS is our novel variant of LCSS to determine occurrences of gestures without segmenting data and performs one order of magnitude faster than the Segmented LCSS. The WarpingLCSS requires low-resource settings to process new arriving samples, thus it is suitable for real-time gesture recognition implemented directly on the small wearable devices. We compare our methods with the existing template matching methods based on Dynamic Time Warping (DTW) on two real-world gesture datasets from arm-worn accelerometer data. The results demonstrate that the LCSS approaches outperform the existing template matching approaches (about 12% in accuracy) in the dataset that suffers from boundary noise and execution variation.

In Pervasive Computing research, substantial work has been directed towards radio-based sensing of human movement patterns. This research has, however, mainly been focused on movements of individuals. This paper addresses the joint identification of the movement indoors of multiple persons forming a cohesive whole - specifically flocks - with clustering approaches operating on three different feature sets derived from WiFi signals which are comparatively analysed. Automatic detection of flocks has several important applications, including social and psychological sensing and emergency research studies. We use a dataset comprising 16 subjects forming one to four flocks walking in a building on single and multiple floors. For the detection of flocks we achieved an average F-measure accuracy of up to 85 percent. We report on the advantages and drawbacks of the three different types of feature sets considering their suitability for use “in the wild” or in well-defined environments.

Wearable gesture recognition enables context aware applications and unobtrusive HCI. It is realized by applying machine learning techniques to data from on-body sensor nodes. We present an gesture recognition system minimizing power while maintaining a run-time application defined performance target through dynamic sensor selection. Compared to the non managed approach optimized for recognition accuracy (95% accuracy), our technique can extend network lifetime by 4 times with accuracy >90% and by 9 times with accuracy >70%. We characterize the approach and outline its applicability to other scenarios.

We present a bilateral vibrotactile feedback system for accurate lateralization of target angles in the complete 360 degree-range. We envision integrating this system into context-aware hearing instruments (HIs) or cochlear implants (CIs) to support users that experience lateralization difficulties. As a foundation for this it is vital to investigate which kind of feedback and vibration patterns are optimal to provide support for lateralization. Our system enables to evaluate and compare different encoding schemes with respect to resolution, reaction time, intuitiveness and user dependency. The system supports bilateral vibrotactile feedback to reflect integration into HIs or CIs worn at both ears and implemented two approaches: Quantized Absolute Heading (QAH) and Continuous Guidance Feedback (CGF). We provide a detailed description of our hardware that was designed to be also applicable for generic vibrotactile feedback applications.

Physical activity, location, as well as a person's psychophysiological and affective state are common dimensions for developing context-aware systems in ubiquitous computing. An important yet missing contextual dimension is the cognitive context that comprises all aspects related to mental information processing, such as perception, memory, knowledge, or learning. In this work we investigate the feasibility of recognising visual memory recall. We use a recognition methodology that combines minimum redundancy maximum relevance feature selection (mRMR) with a support vector machine (SVM) classifier. We validate the methodology in a dual user study with a total of fourteen participants looking at familiar and unfamiliar pictures from four picture categories: abstract, landscapes, faces, and buildings. Using person-independent training, we are able to discriminate between familiar and unfamiliar abstract pictures with a top recognition rate of 84.3% (89.3% recall, 21.0% false positive rate) over all participants. We show that eye movement analysis is a promising approach to infer the cognitive context of a person and discuss the key challenges for the real-world implementation of eye-based cognition-aware systems.

The field of research on activity recognition is relatively young compared to others, like computer vision. In more mature fields, algorithms are usually tested on standardized, reference datasets. This way, algorithms coming from different groups can be tested in a fair manner, which accelerates the process of developing new knowledge. Collecting a reference dataset under realistic settings for activity recognition poses many challenges due to the large amount of sensors and sensor modalities which are needed to provide a sufficiently complete playground. We here report on some lessons learned while collecting such a reference dataset with a heterogeneous setup. We argue for the importance of a few principles to obtain a clean dataset, starting from the sampling and acquisition, down to the synchronization and labeling of the data.

Human activity recognition is a thriving research field. There are lots of studies in different sub-areas of activity recognition proposing different methods. However, unlike other applications, there is lack of established benchmarking problems for activity recognition. Typically, each research group tests and reports the performance of their algorithms on their own datasets using experimental setups specially conceived for that specific purpose. In this work, we introduce a versatile human activity dataset conceived to fill that void. We illustrate its use by presenting comparative results of different classification techniques, and discuss about several metrics that can be used to assess their performance. Being an initial benchmarking, we expect that the possibility to replicate and outperform the presented results will contribute to further advances in state-of-the-art methods.

We present an educational and research kit to support hands-on teaching and experience of real-time activity or gesture recognition from on-body sensors. The kit is comprised of: wireless wearable sensor nodes for motion and ECG sensing; software infrastructure for synchronized data acquisition from multiple sensors, data visualization, signal alignment, and synchronized signal/video exploration; algorithm demonstration and education software with a hidden Markov model-based activity recognition system and a low-latency gesture recognition for a platform game; support hardware for the annotation of user activities from a wireless keypad, and for prototyping of other wireless sensor nodes. All hardware and software is open-source.

In order to develop complexity science based modeling, prediction and simulation methods for large scale socio-technical systems in an Ambient Intelligence (AmI) based smart environment, we propose a scenario based modeling approach. With a case study on AmI technology to support the evacuation from emergency scenarios, i.e. the Life Belt, a wearable computing systems for vibro-tactile directional guidance, we introduce the concept of model scaling from a micro to a macro level. Aligned with the scenario, we present how crowd simulation strategies encoded into a small scale simulation setup can be extended to a mixed-level simulation based on combining model aspects also coming from the large scale model. The experimental results of a real evacuation trail at a local railway station are incorporated to compare the evacuation efficiency for three strategies: (i) Potential Map, (ii) Evacuees familiarity of the exits and (iii) Exits usage optimization. A comparison with the earlier results from small scale simulation suggest that a real large scale simulation results may not be similar to that of small scale simulation due to dynamics of crowd built up and complexity of building structure.

Non-stationary data distributions are a challenge in activity recognition from body worn motion sensors. Classifier models have to be adapted online to maintain a high recognition performance. Typical approaches for online learning are either unsupervised and potentially unstable, or require ground truth information which may be expensive to obtain. As an alternative we propose a teacher signal that can be provided by the user in a minimally obtrusive way. It indicates if the predicted activity for a feature vector is correct or wrong. To exploit this information we propose a novel incremental online learning strategy to adapt a k-nearest-neighbor classifier from instances that are indicated to be correctly or wrongly classified. We characterize our approach on an artificial dataset with abrupt distribution change that simulates a new user of an activity recognition system. The adapted classifier reaches the same accuracy as a classifier trained specifically for the new data distribution. The learning based on the provided correct - error signal also results in a faster learning speed compared to online learning from ground truth. We validate our approach on a real world gesture recognition dataset. The adapted classifiers achieve an accuracy of 78.6% compared to the subject independent baseline of 68.3%.

In this paper, we present a wearable assistant for Parkinson's disease (PD) patients with the freezing of gait (FOG) symptom. This wearable system uses on-body acceleration sensors to measure the patients' movements. It automatically detects FOG by analyzing frequency components inherent in these movements. When FOG is detected, the assistant provides a rhythmic auditory signal that stimulates the patient to resume walking. Ten PD patients tested the system while performing several walking tasks in the laboratory. More than 8 h of data were recorded. Eight patients experienced FOG during the study, and 237 FOG events were identified by professional physiotherapists in a post hoc video analysis. Our wearable assistant was able to provide online assistive feedback for PD patients when they experienced FOG. The system detected FOG events online with a sensitivity of 73.1% and a specificity of 81.6%. The majority of patients indicated that the context-aware automatic cueing was beneficial to them. Finally, we characterize the system performance with respect to the walking style, the sensor placement, and the dominant algorithm parameters.

Opportunistic sensing allows to efficiently collect information about the physical world and the persons behaving in it. This may mainstream human context and activity recognition in wearable and pervasive computing by removing requirements for a specific deployed infrastructure. In this paper we introduce the newly started European research project OPPORTUNITY within which we develop mobile opportunistic activity and context recognition systems. We outline the project's objective, the approach we follow along opportunistic sensing, data processing and interpretation, and autonomous adaptation and evolution to environmental and user changes, and we outline preliminary results.

Context processing in Body Area Networks (BANs) faces unique challenges due to the user and node mobility, the need of real-time adaptation to the dynamic topological and contextual changes, and heterogeneous processing capabilities and energy constraints present on the available devices. This paper proposes a service-oriented framework for the execution of context recognition algorithms. We describe and theoretically analyze the performance of the main framework components, including the sensor network organization, service discovery, service graph construction, service distribution and mapping. The theoretical results are followed by the simulation of the proposed framework as a whole, showing the overall cost of dynamically distributing applications on the network.

In this work we describe the design, implementation and evaluation of a novel eye tracker for context-awareness and mobile HCI applications. In contrast to common systems using video cameras, this compact device relies on Electrooculography (EOG). It consists of goggles with dry electrodes integrated into the frame and a small pocket-worn component with a DSP for real-time EOG signal processing. The device is intended for wearable and standalone use: It can store data locally for long-term recordings or stream processed EOG signals to a remote device over Bluetooth. We describe how eye gestures can be efficiently recognised from EOG signals for HCI purposes. In an experiment conducted with 11 subjects playing a computer game we show that 8 eye gestures of varying complexity can be continuously recognised with equal performance to a state-of-the-art video-based system. Physical activity leads to artefacts in the EOG signal. We describe how these artefacts can be removed using an adaptive filtering scheme and characterise this approach on a 5-subject dataset. In addition to explicit eye movements for HCI, we discuss how the analysis of unconscious eye movements may eventually allow to deduce information on user activity and context not available with current sensing modalities.