This work demonstrates how mixed effects random forests enable accurate predictions of depression severity using multimodal physiological and digital activity data collected from an 8-week study involving 31 patients with major depressive disorder. We show that mixed effects random forests outperform standard random forests and personal average baselines when predicting clinical Hamilton Depression Rating Scale scores (HDRS17). Compared to the latter baseline, accuracy is significantly improved for each patient by an average of 0.199-0.276 in terms of mean absolute error (p 0.05). This is noteworthy as these simple baselines frequently outperform machine learning methods in mental health prediction tasks. We suggest that this improved performance results from the ability of the mixed effects random forest to personalise model parameters to individuals in the dataset. However, we f ind that these improvements pertain exclusively to scenarios where labelled patient data are available to the model at training time. Investigating methods that improve accuracy when generalising to new patients is left as important future work.
This exploratory study examined the effects of varying g-forces, including feelings of weightlessness, on an individual’s physiology during parabolic flight. Specifically, we collected heart rate, accelerometer, and skin conductance measurements from 16 flyers aboard a parabolic flight using wearable, wireless sensors. The biosignals were then correlated to participant reports of nausea, anxiety, and excitement during periods of altered g-forces. Using linear mixed-effects models, we found that (1) heart rate was positively correlated to individuals’ self-reported highest/lowest periods of both anxiety and excitement, and (2) bilateral skin conductance asymmetry was positively correlated to individuals’ self-reported highest/lowest periods of nausea.
Abstract—Depression is the major cause of years lived in disability world-wide; however, its diagnosis and tracking methods still rely mainly on assessing self-reported depressive symptoms, methods that originated more than fifty years ago. These methods, which usually involve filling out surveys or engaging in face-to-face interviews, provide limited accuracy and reliability and are costly to track and scale. In this paper, we develop and test the efficacy of machine learning techniques applied to objective data captured passively and continuously from E4 wearable wristbands and from sensors in an Android phone for predicting the Hamilton Depression Rating Scale (HDRS). Input data include electrodermal activity (EDA), sleep behavior, motion, phone-based communication, location changes, and phone usage patterns. We introduce our feature generation and transformation process, imputing missing clinical scores from self-reported measures, and predicting depression severity from continuous sensor measurements. While HDRS ranges between 0 and 52, we were able to impute it with 2.8 RMSE and predict it with 4.5 RMSE which are low relative errors. Analyzing the features and their relation to depressive symptoms, we found that poor mental health was accompanied by more irregular sleep, less motion, fewer incoming messages, less variability in location patterns, and higher asymmetry of EDA between the right and the left wrists.
Abstract: Ambulatory skin conductance (SC) signals often need to be analyzed independently for different user activities. As an ambulatory SC sensor is usually combined with an accelerometer, we examined its measurements to identify if a user is sitting, walking and running. We present our method for estimating the activities and how SC signals are distributed across daytime and sleep contexts.
Abstract. In recent years Wireless Sensor Networks (WSNs) have been deployed in wide range of applications from the health and environment monitoring to building and industrial control. However, the pace of prevalence of WSN is slower than anticipated by the research community due to several reasons including required embedded systems expertise for developing and deploying WSNs; use of proprietary protocols; and limits in scalability and reliability. In this paper we propose PyFUNS (Pythonbased Framework for Ubiquitous Networked Sensors) to address these
challenges. PyFUNS handles low level and networking functionalities, using the services provided by Contiki, and leaves to the user only the task of application development in the form of Python scripts. This approach reduces required expertise in embedded systems to develop WSN based applications. PyFUNS also uses 6LoWPAN and CoAP standard protocols to enable interoperability and ease of integration with other systems, pursuing the Internet of Things vision. Through a real implementation
of PyFUNS in two constrained platforms we proved its feasibility in mote devices, as well as its performance in terms of control delay, energy consumption and network traffic in several network topologies. As it is possible with PyFUNS to easily compare performance of different deployments of distributed application, PyFUNS can be used to identify optimal design of distributed application.
Abstract: Recently, wearable devices have allowed for long term, ambulatory measurement of electrodermal activity (EDA). Despite the fact that ambulatory recording can be noisy, and recording artifacts can easily be mistaken for a
physiological response during analysis, to date there is no automatic method for detecting artifacts. This paper describes the development of a machine learning algorithm for automatically detecting EDA artifacts, and provides an empirical evaluation of classification performance. We have encoded our results into a freely available web-based tool for artifact and peak detection.
Abstract: To filter noise or detect features within physiological signals, it is often effective to encode expert knowledge into a model such as a machine learning classifier. However, training such a model can require much effort on the part of
the researcher; this often takes the form of manually labeling portions of signal needed to represent the concept being trained. Active learning is a technique for reducing human effort by developing a classifier that can intelligently select the most relevant data samples and ask for labels for only those samples, in an iterative process. In this paper we demonstrate that active learning can reduce the labeling effort required of researchers by as much as 84% for our application, while
offering equivalent or even slightly improved machine learning performance.
In this paper, we build upon the Internet of Things (IoT) paradigm, with aim of delivering networked solutions that enable to connect not only single sensors, but also whole wireless sensor networks (WSN) to the Internet in a secure, simple and efficient way, and describe the design and implementation of a smart-home management
system. The system is composed of a lightweight tool with an intuitive user interface for commissioning of IPenabled WSN with constrained capabilities. The solution includes a visual programming interface with a common framework for discovering smart home services on the constrained WSN, and a code analysis and translation engine to generate python code. This engine analyses the application rules defined with the graphical user interface and translates them into distributed application scripts. The system also includes modules to plan the optimization of the deployment, and deploy and start the generated code. A prototype of the system, with the visual programming solution and code generation module developed is presented in this paper.
Abstract: In this paper, we propose a “Neighbour Disjoint Multipath (NDM)” scheme that increases resilience against node or link failures in a wireless sensor network (WSN). Our algorithm chooses the shortest path between a sensor and the
sink as the primary path, thus ensuring the algorithm is energy efficient under normal circumstances. In selecting the backup paths, we utilise the disjoint property to ensure that i) when there are k paths between source and sink, no set of k node failures can result in total communication break between them,
and ii) by having (k − 1) spatially separated backup paths w.r.t. the primary path, the probability of simultaneous failure of the primary and backup paths is reduced in case of localised poor channel quality or node failures. Our algorithm not only ensures the node disjointedness characteristics of the constructed paths, but also tries to minimise the impact of localised node or link failures where a localised portion of the network may be unusable. We analyse the motivation behind our idea clearly, and discuss the algorithm in detail. We also compare the NDM scheme with other common multipath techniques such as node-disjoint and edge-disjoint approaches, and point out its effectiveness through
Abstract: Building Automation usually involves a large number of systems that should cooperate in
order to improve e.g. the user comfort, security, but also decrease the overall energy consumption. One aim of the EU funded SCUBA project is to improve the coordination among devices and systems installed in a building. This paper deals with the extension of a framework dedicated to Building Automation and built on top of the LINC resource-based middleware. Two particular tools developed in SCUBA and that make use of the middleware are also presented together with the encapsulation of the ontology-based Building Automation System model named BASOnt.
Abstract: This paper provides an overview of the architecture for self-organizing, co-operative and robust Building Automation Systems (BAS) proposed by the EC funded FP7 SCUBA1 project. We describe the current situation in monitoring and control systems and outline the typical stakeholders involved in the case of building automation systems. We derive seven typical use cases which will be demonstrated and evaluated on
pilot sites. From these use cases the project designed an architecture relying on six main modules that realize the design, commissioning and operation of self-organizing, co-operative,robust BAS.
Abstract: IPv6 will make it possible to provide Internet connectivity to any device. In the same line, Web technologies will make managing, communicating and visualizing any information
provided by these devices attractive to the end users and application developers. Most of the new devices connected to this Web of Things (WoT) will be embedded and wirelessly connected.
However, current Web technologies, developed with powerful devices in mind, will not be suited for this kind of environment. In order to make the WoT a reality for low power embedded networks, specialised protocols that consider the energy, memory and processing constraints of these devices must be designed. The IETF recently created the CoRE group whose first goal has been developing a RESTful application layer protocol for communications within embedded wireless networks referred to
as Constrained Application Protocol (CoAP). The year 2011 has seen a big push with regards to research in this area, indicating a growing interest in the community towards RESTful interactions
in low power wireless embedded networks. This paper surveys current research efforts on the Constrained Application Protocol for low power embedded networks.
Abstract: The commissioning of sensors and actuators within a building is often carried out by an operator who manually gathers and later inserts configuration data into the building management system. Data sets collected can easily reach hundreds which makes this manual process a slow complex
operation which in turn is prone to errors. In this paper, we present a client-server architecture for a simple web based commissioning application, based on the Constrained Application Protocol (CoAP), that allows an operator to easily discover and browse through newly installed devices in order to perform commissioning configurations onsite.
Abstract: Quality of Service (QoS) monitoring of end-user services is an integral and indispensable part of service management. However in large, heterogeneous and complex networks where there are many services, many types of end-user devices, and huge numbers of subscribers, it is not trivial to monitor QoS and estimate the status of Service Level Agreements (SLAs). Furthermore, the overwhelming majority of end-terminals do not provide precise information about QoS which aggravates the difficulty of keeping track of SLAs. In this paper, we describe a solution that combines a number of techniques in a novel and unique way to overcome the complexity and difficulty of QoS monitoring. Our solution uses a model driven approach to service modeling, data mining techniques on small sample sets of terminal QoS reports (from “smarter” end-user devices), and network level key performance indicators (N-KPIs) from probes to address this problem. Service modeling techniques empowered with a modeling engine and a purpose-built language hide the complexity of SLA status monitoring. The data mining technique uses its own engine and learnt data models to estimate QoS values based on N-KPIs, and feeds the estimated values to the modeling engine to calculate SLAs. We describe our solution, the prototype and experimental results in the paper.
Abstract: Deployment and upgrade of a mobile network have always been challenging tasks. Very often they require human intervention because telecom networks are complex systems composed of different nodes that need to be compatible in order to communicate and provide network services. Therefore in current telecommunication systems a network expert must check all the requirements and compatibilities of the network prior to activation of a new service. Automation of the assessment of network compatibility is one of the key enablers for Autonomic Management of telecom networks. In this paper we describe a new method for automatic end-to-end assessment of compatibility between network features in a telecom network. The method enables fast, easy and accurate decision making regarding the planning of new feature deployment or the upgrade of already existing features. We built a prototype that demonstrates the described method. It shows that our method is not bound to any type of telecom network and could be used to automate deployment or upgrade of a multiple-domain network.
Abstract—The time synchronization problem needs to be considered in a distributed system. In Wireless Sensor Networks (WSNs) this issue must be solved with limited computational, communication and energy resources. Many synchronization protocols exist for WSNs. However, in most cases these protocols are independent entities with specific packets, communication scheme and network hierarchy. This solution is not energy efficient. Because it is very rare for synchronization not to be necessary in WSNs, we advocate integrating the synchronization service into the routing layer.We have implemented this approach in a new synchronization protocol called Routing Integrated Synchronization Service (RISS). Our tests show that RISS is very time and energy efficient and also is characterized by a small overhead. We have compared its performance experimentally to that of the FTSP synchronization protocol and it has proved to offer better time precision than the latter protocol.
Abstract: The hop distance strategy in Wireless Sensor Networks (WSNs) has a major impact on energy consumption of each sensor mote. Long-hop routing minimizes reception cost. However, a substantial power demand is incurred for long distance transmission. Since the transceiver is the major source of power consumption in the node, optimizing the routing for hop length can extend significantly the lifetime of the network. This paper explores when multi-hop routing is more energy efficient than direct transmission to the sink and the conditions for which the two-hop strategy is optimal. Experimental evidence is provided in to support of these conclusions. The tests showed that the superiority of the multi-hop scheme depends on the source-sink distance
and reception cost. They also demonstrated that the two-hop strategy is most energy efficient when the relay is at the midpoint of the total transmission radius. Our results may be used in existing routing protocols to select optimal relays or to determine whether it is better to send packets directly to the base station or through intermediate nodes.
Abstract: A common time reference across nodes is required in most Wireless Sensor Networks (WSNs) applications. It is needed, for example, to time-stamp sensor samples and for long-term duty cycling of nodes. Also many routing protocols require that nodes communicate according to some predefined schedule for reasons of energy eficiency. However, independent distribution of the time information, without considering the routing algorithm schedule or network topology may lead to a failure of the synchronisation protocol. This was confirmed empirically, and was shown to result in loss of connectivity. This can be avoided by integrating the synchronisation service into the network layer with a so-called cross-layer approach. This approach introduces interactions between the layers of a conventional layered network stack, so that the routing layer may share information with other layers. We explore whether energy eficiency can be enhanced through the use of cross-layer optimisations and present two novel cross-layer routing algorithms. The first protocol, designed for hierarchical, cluster based networks and called CLEAR (Cross Layer Eficient Architecture for Routing), uses the routing algorithm to distribute time information which can be used for eficient duty cycling of nodes. The second method – called RISS (Routing Integrated Synchronization Service) – integrates time synchronization into the network layer and is designed to work well in flat, non-hierarchical network topologies.
We implemented and tested the performance of these solutions in simulations and also deployed these routing techniques on sensor nodes using TinyOS. We compared the average power consumption of the nodes and the precision of time synchronization with the corresponding parameters of a number of existing algorithms. All proposed schemes extend the network lifetime and due to their lightweight architecture they are very eficient on WSN nodes with constrained resources. Hence it is recommended that a cross-layer approach should be a feature of any routing algorithm for WSNs.