Clemens Lombriser

Activity recognition from on-body sensors: accuracy-power trade-off by dynamic sensor selection

Activity recognition from an on-body sensor network enables context-aware applications in wearable computing. A guaranteed classification accuracy is desirable while optimizing power consumption to ensure the systems wearability. In this paper, we investigate the benefits of dynamic sensor selection in order to use efficiently available energy while achieving a desired activity recognition accuracy. For this purpose we introduce and characterize an activity recognition method with an underlying run-time sensor selection scheme. The system relies on a meta-classifier that fuses the information of classifiers operating on individual sensors. Sensors are selected according to their contribution to classification accuracy as assessed during system training. We test this system by recognizing manipulative activities of assembly-line workers in a car production environment. Results show that the systems lifetime can be significantly extended while keeping high recognition accuracies. We discuss how this approach can be implemented in a dynamic sensor network by using the context-recognition framework Titan that we are developing for dynamic and heterogeneous sensor networks.

Smart Sensing and Context

Invited Paper.- CenceMe - Injecting Sensing Presence into Social Networking Applications.- Spatial and Motion Context.- Mapping by Seeing - Wearable Vision-Based Dead-Reckoning, and Closing the Loop.- The Design of a Pressure Sensing Floor for Movement-Based Human Computer Interaction.- Sensing Motion Using Spectral and Spatial Analysis of WLAN RSSI.- Inferring and Distributing Spatial Context.- Context Sensitive Adaptive Authentication.- Human Behavior as Context.- A Sensor Placement Approach for the Monitoring of Indoor Scenes.- Recognition of User Activity Sequences Using Distributed Event Detection.- Behavior Detection Based on Touched Objects with Dynamic Threshold Determination Model.- Towards Mood Based Mobile Services and Applications.- Recognising Activities of Daily Life Using Hierarchical Plans.- Context Frameworks and Platforms.- GlobeCon - A Scalable Framework for Context Aware Computing.- ESCAPE - An Adaptive Framework for Managing and Providing Context Information in Emergency Situations.- Capturing Context Requirements.- Deployment Experience Toward Core Abstractions for Context Aware Applications.- Sensing Technologies and Case Studies.- Ambient Energy Scavenging for Sensor-Equipped RFID Tags in the Cold Chain.- Escalation: Complex Event Detection in Wireless Sensor Networks.- Multi-sensor Cross Correlation for Alarm Generation in a Deployed Sensor Network.

Titan: A Tiny Task Network for Dynamically Reconfigurable Heterogeneous Sensor Networks

Context recognition, such as gesture or activity recognition, is a key mechanism that enables ubiquitous computing systems to proactively support users. It becomes challenging in unconstrained environments such as those encountered in daily living, where it has to deal with heterogeneous networks, changing sensor availability, communication capabilities, and available processing power.

Distributed Activity Recognition with Fuzzy-Enabled Wireless Sensor Networks

Wireless sensor nodes can act as distributed detectors for recognizing activities online, with the final goal of assisting the users in their working environment. We propose an activity recognition architecture based on fuzzy logic, through which multiple nodes collaborate to produce a reliable recognition result from unreliable sensor data. As an extension to the regular fuzzy inference, we incorporate temporal order knowledge of the sequences of operations involved in the activities. The performance evaluation is based on experimental data from a car assembly trial. The system achieves an overall recognition performance of 0.81 recall and 0.79 precision with regular fuzzy inference, and 0.85 recall and 0.85 precision when considering temporal order knowledge. We also present early experiences with implementing the recognition system on sensor nodes. The results show that the algorithms can run online, with execution times in the order of 40ms, for the whole recognition chain, and memory overhead in the order of 1.5kB RAM.

Recognition of user activity sequences using distributed event detection

We describe and evaluate a distributed architecture for the online recognition of user activity sequences. In a lower layer, simple heterogeneous atomic activities were recognised on multiple on-body and environmental sensor-detector nodes. The atomic activities were grouped in detection events, depending on the detector location. In a second layer, the recognition of composite activities was performed by an integrator. The approach minimises network communication by local activity aggregation at the detector nodes and transforms the temporal activity sequence into a spatial representation for simplified composite recognition. Metrics for a general description of the architecture are presented. We evaluated the architecture in a worker assembly scenario using 12 sensor-detector nodes. An overall recall and precision of 77% and 79% was achieved for 11 different composite activities. The architecture can be scaled in the number of sensor-detectors, activity events and sequences while being adequately quantified by the presented metrics.

Tandem: a context-aware method for spontaneous clustering of dynamic wireless sensor nodes

Wireless sensor nodes attached to everyday objects and worn by people are able to collaborate and actively assist users in their activities. We propose a method through which wireless sensor nodes organize spontaneously into clusters based on a common context. Provided that the confidence of sharing a common context varies in time, the algorithm takes into account a window-based history of believes. We approximate the behaviour of the algorithm using a Markov chain model and we analyse theoretically the cluster stability. We compare the theoretical approximation with simulations, by making use of experimental results reported from field tests. We show the tradeoff between the time history necessary to achieve a certain stability and the responsiveness of the clustering algorithm.

An Educational and Research Kit for Activity and Context Recognition from On-body Sensors

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.

A power-efficient wireless sensor network for continuously monitoring seismic vibrations

We present a novel power-efficient wireless sensor network for continuously monitoring and analyzing seismic vibrations with sensor nodes and forwarding the retrieved information with low-cost relay nodes to backend applications. The applied vibration sensing algorithms are derived from the DIN 4150–3 standard. All nodes in the network are battery-powered and equipped with an IEEE 802.15.4 compatible radio transceiver. The nodes communicate with each other by executing a novel power-efficient protocol stack, which provides all network functions required by the vibration-sensing application and uses a publish/subscribe messaging protocol for communicating between the network nodes and the backend applications. Results obtained in certification and field tests show that the proposed vibration-sensing solution is standard-compliant, and that the wireless vibration sensor network (WVSN) exhibits excellent performance in terms of packet delivery rate, latency, and power efficiency.

Bringing quality of context into wearable human activity recognition systems

Quality of Context (QoC) in context-aware computing improves reasoning and decision making. Activity recognition in wearable computing enables context-aware assistance. Wearable systems must include QoC to participate in context processing frameworks common in large ambient intelligence environments. However, QoC is not specifically defined in that domain. QoC models allowing activity recognition system reconfiguration to achieve a desired context quality are also missing. Here we identify the recognized dimensions of QoC and the performance metrics in activity recognition systems. We discuss how the latter maps on the former and provide provide guidelines to include QoC in activity recognition systems. On the basis of gesture recognition in a car manufacturing case study, we illustrate the signification of QoC and we present modeling abstractions to reconfigure an activity recognition system to achieve a desired QoC.

Modeling Service-Oriented Context Processing in Dynamic Body Area Networks

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.