Poor sleep quality is common in individuals with chronic obstructive pulmonary disease (COPD) and is linked to higher risk of exacerbation and hospitalisation. While clinical studies typically assess sleep quality using the subjective Pittsburgh Sleep Quality Index (PSQI), this study proposes an objective method based on continuous respiratory signals from the chest-worn RESpeck device. Features were extracted from respiratory rate, breath regularity, PSQI-aligned signal metrics, and attractor reconstruction of respiratory dynamics, alongside a consolidated feature set combining all domains. Five machine learning models, including CatBoost, were trained on overnight data (21:00-10:00) to predict the self-reported score (0-5) to Question 7 of the COPD Assessment Test (Q7CAT), which evaluates sleep quality. The CatBoost model with the consolidated feature set achieved the best performance, with a mean absolute error of$0.75 \pm 0.20$and a quadratic weighted kappa of$0.60 \pm 0.15$, using 160 nights of data from 12 COPD patients (SMILE dataset). These results demonstrate that respiratory signal–derived features can objectively predict sleep quality, offering a potential alternative to subjective questionnaires and supporting early intervention in COPD management.

Chronic obstructive pulmonary disease (COPD) is characterised by persistent respiratory symptoms and activity limitations. While tools like the COPD Assessment Test (CAT) enable self-reported monitoring, they lack physiological objectivity and temporal resolution. This study investigates the use of phase-space attractor reconstruction, a nonlinear timeseries method, for symptom tracking using respiratory signals from a chest-worn accelerometer. Data from 12 COPD patients over four to six weeks were segmented using a CNN-BiGRU-based activity classifier to isolate stationary periods. Attractor reconstructions were computed at 60-second intervals, and 112 features spanning geometric, spectral, and topological domains were extracted. Several features showed noteworthy correlations with total and item-level CAT scores, supporting their potential as objective markers of symptom burden. These results highlight the feasibility of attractor-based analysis for non-invasive, continuous COPD monitoring and personalised disease management.

Chronic obstructive pulmonary disease (COPD) is characterised by persistent airflow limitation and fluctuating symptoms that often go undetected in between hospital visits. This paper investigates the use of attractor-based phase-space reconstruction, a non-linear method for transforming time-series data, to characterise respiratory dynamics from chest-worn RESpeck accelerometer. Respiratory signal data were collected over two to four weeks from 50 participants (18 COPD, 32 controls) in free-living conditions. Two-dimensional attractors were derived from 60-second stationary respiratory windows, and 27 features spanning geometric, spectral, recurrence, and complexity domains were extracted. Several features showed large effect sizes and enabled COPD classification with 84.4% accuracy. Temporal analysis revealed heightened diurnal variability in COPD, particularly during night-to-morning transitions. A case study of a COPD subject demonstrated that attractor-derived features captured gradual pre-exacerbation changes not evident from respiratory rate alone. These findings highlight the potential of attractor-derived features as objective, high-resolution digital biomarkers of respiratory dysfunction, validating their use in passive, continuous monitoring for personalised COPD management.

The respiratory rate and the respiratory flow/effort are important vital signs monitored during the care of patients in the hospital and during remote management of subjects with chronic lung diseases, such as asthma, chronic obstructive pulmonary disease (COPD), and congenital airways malformation. The Respeck is a wireless device worn as a plaster on the chest which monitors continuously respiratory signals, such as the respiratory rate (breaths/minute), and the respiratory flow/effort (amplitude over time). The regularity of the respiratory signal, or rather the nature and frequency of occurrence of irregular breathing episodes can be used as one of the characteristics of the subjects' clinical status. This paper develops computational methods for classifying episodes of rapid-shallow breathing based on approximating the tidal volume using features such as: the area under the respiratory signal curve, peak respiratory flow and the respiratory rate. Results are presented for a data-driven approach to the analysis of a Respeck dataset of respiratory flow signals gathered from 137 asthmatic subjects.

The COPD Assessment Test (CAT) consists of 8 questions, each one scored by patients on a scale of 0 to 5 points, and the aggregate score provides a self-assessment of the impact of their symptoms on their well-being. In the nature of questionnaires, the numeric CAT score is a subjective assessment by COPD patients on the effect of symptoms on their personal condition. This paper evaluates the pulmonary response of COPD patients wearing the Respeck device, during exertion when performing pulmonary rehabilitation (PR) exercises at home, as a measure of their wellbeing. Machine learning models were developed to use features derived from the Respeck respiratory rate and physical activity data during the PR exercises as objective measures to associate with the CAT score on the same day and to predict the CAT score for the next day. Results are presented for COPD patients in terms of standard error metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and clinically-relevant minimum clinically important difference (MCID).

This paper describes an unobtrusive sleep-wake monitor based on the wireless Respeck sensor worn as a patch on the chest, coupled with machine learning-based classification methods for the automatic classification of Sleep-Wake states in the Respeck sensor data. The \Sleep-Wake monitor was evaluated on labelled data collected by a cohort of 20 healthy volunteers and achieved results for accuracy, sensitivity and specificity of greater than 90% using Linear SVM, and non-linear SVM-RBF models. Metrics such as frequency of Wake-After-Sleep-Onset (WASO) episodes were calculated using Respeck data to mark the quality of sleep in the cohort.

This paper describes an unobtrusive cough monitor based on the wireless Respeck sensor worn as a patch on the chest, in tandem with a deep learning-based cough classification method for the automatic detection of instances of coughs in the Respeck sensor data. The cough monitor was evaluated on an unseen, 2.5h-long, Respeck dataset which mimicked real-life settings and achieved an accuracy of greater than 82% using a one-dimensional convolutional neural network. Results are presented on testing the Respeck cough monitor in the wild on asthma and COPD patients which provided insights validated by independent publications.

According to the Lancet report on global burden of disease published in October 2020, air pollution is amongst the five highest risk factors for global health, reducing life expectancy on average by 20 months. This paper describes a data-driven method for establishing causal relationships between two time-series data streams derived from wearable sensors: personal exposure to airborne particulate matter (PM) of aerodynamic sizes less than 2.5 μm(PM2.5) gathered from the Airspeck monitor, and continuous respiratory rate (breaths/minute) measured by the wireless Respeck monitor worn as a plaster on the chest. Results are presented for a cohort of asthmatic adolescents using the PCMCI method on the short-term causal relationship between PM2.5exposure and respiratory rate for time lags in the first 60 minutes at minute-level intervals, and for time lags between 2 to 8 hours at 10-minute time intervals. For the first time a personalised exposure-response relationship between PM2.5exposure and respiratory rate has been demonstrated for short-term effects in asthmatic adolescents during their every day lives.

The Automatic Urban and Rural Network (AURN) [1] is a set of high quality reference monitoring sites for recording air quality in the United Kingdom. They are costly to install and expensive to run, and are therefore limited in numbers. The data from these networks are used to inform regulatory compliance with the Ambient Air Quality Directives [2]. There is also a requirement to monitor air pollution at sufficiently high spatial and temporal resolutions around people to estimate personal exposure to particulates, and gases such as Nitrogen Dioxide and Ozone for better understanding their health impacts. Such high resolution measurements can also be used for validating the air quality models' estimates of variability over space and time due to complex interactions. Networks of air-quality monitors using inexpensive sensors offer a cost-effective alternative approach for recording trends in air quality at a higher spatial resolution, albeit not as accurately as the reference monitoring sites. This paper describes the design, implementation, and deployment of a family of air quality monitors: stationary (AirSpeck-S) monitors for measuring ambient air quality, and mobile wearable AirSpeck-P for monitoring personal exposure to air borne particulates (PM10, PM2.5and PM1), and the gases - Nitrogen Dioxide and Ozone. Results are presented for characterising the ambient air quality in public spaces gathered from people wearing the AirSpeck-P monitors who are out and about in two cities as pedestrians (Edinburgh, Scotland) and as car passengers (Delhi, India). The paper demonstrates the viability of using inexpensive static and mobile AirSpeck monitors for mapping trends in particulate concentrations in urban spaces. Results are presented for comparisons of the mobile personal exposure data from pedestrians with static AirSpeck-S monitors along the same route, and the characterization of urban spaces based on levels of particulate concentration using the AirSpeck-P monitor.

The Retuerta is one of the oldest breed of horses in Europe, which roams wild in the Donana National Park, Andalusia, Spain. Thirty-two of these horses were marked with wireless sensors to gather spatio-temporal data on their behaviour over a period of several months. This paper describes our experiences of tracking and monitoring these wild horses attached with body-worn sensors and operating in a harsh and challenging environment. Analysis of this data for the first two months has revealed rare insights into the horses' social behaviour, such as the group dynamics (group sizes and memberships), dispersal and home ranges which are of interest to both animal ethologists and practitioners managing the ecology of their wild habitats. The paper introduces the Virtual Beacon - Time Division Multiple Access (VB-TDMA) protocol for orchestrating the data collection, and describes the choices that were made for addressing the many technical challenges for an extended deployment, such as in the design of the sensor platform, wireless data collection and battery lifetime issues. Our experiences point to the virtue of simplicity in design of wireless sensor networks to support core functionalities for achieving good average case performances.

This paper presents a hybrid method based on a feed-forward neural network (FNN) embedded in a hidden Markov model (HMM), for detecting phases in a gait cycle, based on data from inertial sensors attached to the lower body. The method was validated against the ground truth obtained concurrently from a Vicon optical motion capture system for five volunteers. The method was characterised using metrics such as sensitivity and specificity for sensor placements, and gait analysis. The results demonstrate that the proposed method is accurate within 23 milliseconds with the added advantages of mobility afforded by wireless sensors and the flexibility of the classification method.

This paper addresses the use of on-body sensors for tracking and monitoring Retuerta wild horses (one of the oldest horse breeds in Europe) over a 12-month period in the Doñana National Park, Andalusia, Spain. The challenges are outlined and the design of the Prospeckz-5 platform, which is used in both the sensor node on the horses and in the base stations for gathering data, is described. Results are presented for tests on the performance of the antenna and the solar cell array charger ahead of deployment on the horses in Spain.

Despite recent advances in 3-D motion capture, the problem of simultaneously tracking human posture and position in an unconstrained environment remains open. Optical systems provide both types of information, but are confined to a restricted area of capture. Inertial sensing alleviates this restriction, but at the expense of capturing only relative (postural) and not absolute (positional) information. In this paper, we propose an algorithm combining the relative merits of these systems to track both position and posture in challenging environments. Offline, we combine an optical (Kinect) and an inertial sensing (Orient-4) platform to learn a mapping from posture variations to translations, which we encode as a translation manifold. Online, the optical source is removed, and the learned mapping is used to infer positions using the postures computed by the inertial sensors. We first evaluate our approach in simulation, on motion sequences with ground-truth positions for error estimation. Then, the method is deployed on physical sensing platforms to track human subjects. The proposed algorithm is shown to yield a lower average cumulative error than comparable position tracking methods, such as double integration of accelerometer data, on both simulated and real sensory data, and in a variety of motions and capture settings.

This paper describes an integrated hybrid simulation environment in which physical electronic devices interact in real-time with a discrete-event simulator and a 3D visualisation engine, where the communication between the real devices and the virtual world is mediated by CoAP. The resulting simulation framework is a powerful tool for designers of Internet of Things (IoT) applications to assess design decisions ahead of deployment, based on realistic data from sensors and typical movement of people within built spaces. A motivating example is used to illustrate the capabilities of hybrid simulations based on a multi-residence housing facility intended for elderly people, each wearing an on-body speck with one or more sensors, to monitor their condition such as breathing, heart-rate, and activity, and which transmits this information via a mesh network of base-stations to a central hub. The results demonstrate that design decisions can be made on the choice of routing protocols based on real-time transmission of data from people, which captures their typical movement in a built environment and based on actual data transmitted by on-body devices.

Research in the field of embodied music cognition has shown the importance of coupled processes of body activity (action) and multimodal representations of these actions (perception) in how music is processed. Technologies in the field of human–computer interaction (HCI) provide excellent means to intervene into, and extend, these coupled action-perception processes. In this article this model is applied to a concrete HCI application, called the “Conducting Master.” The application facilitates multiple users to interact in real time with the system in order to explore and learn how musical meter can be articulated into body movements (i.e., meter-mimicking gestures). Techniques are provided to model and automatically recognize these gestures in order to provide multimodal feedback streams back to the users. These techniques are based on template-based methods that allow approaching meter-mimicking gestures explicitly from a spatiotemporal account. To conclude, some concrete setups are presented in which the functionality of the Conducting Master was evaluated.

An unsupervised learning algorithm is presented for segmentation and evaluation of motion data from the on-body Orient wireless motion capture system for mobile gait analysis. The algorithm is model-free and operates on the latent space of the motion, by first aggregating all the sensor data into a single vector, and then modeling them on a low-dimensional manifold to perform segmentation. The proposed approach is contrasted to a basic, model-based algorithm, which operates directly on the joint angles computed by the Orient sensor devices. The latent space algorithm is shown to be capable of retrieving qualitative features of the motion even in the face of noisy or incomplete sensor readings.

This paper explores the use of Orient specks as on-body network of wireless inertial-magnetic sensors to capture the parameters of the human gait for mobile clinical gait analysis. A range of kinematic and temporal parameters were measured for normal humans using Orient specks and compared to values obtained from a commercial Vicon optical motion capture system. There was a good correlation between the joint angle data from the two systems, most notably in the sagittal plane. Hip flexion graphs showed the highest correlation value of 0.973, with knee flexion at 0.855 and pelvic rotation at 0.943, followed by pelvic obliquity at 0.689 and ankle flexion at 0.626. We conclude that the Orient specks have the potential for obtaining gait parameters outside the laboratory environment by measuring temporal parameters, and detecting the shape and trend of kinematic parameters of the patients when they are out and about during their everyday lives.

Accelerometer-based respiratory monitoring is a recent area of research based on the observation of small rotations at the chest wall due to breathing. Previous studies of this technique have begun to address some sources of interference e.g. subject movements, but have not investigated operation during speech production when breathing patterns are known to be substantially different to normal respiration. We demonstrate measurement of speech breathing with a wireless tri-axial accelerometer in a synchronously captured dataset, including annotated audio and electro-magnetic articulograph data. We find agreement between peaks in the accelerometer-derived rotation signal and manually annotated breath timings, and correlation between peak rotations and the duration of audible in breaths. In speech breathing the rotation rate signal does not appear to be a good proxy for airflow rate as previously suggested, and instead seems to better reflect the role of specific muscles around the accelerometer location. We conclude that the method can be usable during speech breathing, but that this difference should be considered. The method has some advantages for speech breathing research due to its unobtrusive nature.

Simultaneous monitoring of respiratory function and activity level would be of benefit in the monitoring of chronic conditions such as chronic obstructive pulmonary disorder (COPD), but is ill-addressed by existing methods. Current solutions for monitoring respiratory function are obtrusive and not suitable for pervasive monitoring in the home, while existing activity monitors, not equipped to measure parameters of respiration, do not differentiate between causes of sedentary behaviour. Previous work has validated a method for obtaining angular rates of breathing motion of the chest wall using a tri-axial accelerometer against nasal pressure. We have used this method to perform respiratory monitoring during periods of low activity while simultaneously monitoring activity using a single wireless device. We observe that the optimal placement for respiratory monitoring does not preclude successful activity monitoring. We propose an activity monitoring algorithm based on direct estimation of motion energy observed by the device. We show favourable comparison against three commercial activity monitors validated against indirect calorimetry during a programme of exercise activities in healthy subjects.

There is a strong medical need for continuous, unobstrusive respiratory monitoring, and many shortcomings to existing methods. Previous work shows that MEMS accelerometers worn on the torso can measure inclination changes due to breathing, from which a respiratory rate can be obtained. There has been limited validation of these methods. The problem of practical continuous monitoring, in which patient movement disrupts the measurements and the axis of interest changes, has also not been addressed. We demonstrate a method based on tri-axial accelerometer data from a wireless sensor device, which tracks the axis of rotation and obtains angular rates of breathing motion. The resulting rates are validated against gyroscope measurements and show high correlation to flow rate measurements using a nasal cannula. We use a movement detection method to classify periods in which the patient is static and breathing signals can be observed accurately. Within these periods we obtain a close match to cannula measurements, for both the flow rate waveform and derived respiratory rates, over multi-hour datasets obtained from wireless sensor devices on hospital patients. We discuss future directions for improvement and potential methods for estimating absolute airflow rate and tidal volume.

Motion capture using wireless inertial measurement units (IMUs) has many advantages over other techniques. Achieving accurate tracking with IMUs presents a processing challenge, especially for real time tracking. Centralised approaches are bandwidth-intensive and prone to error from packet loss. Methods based solely on local knowledge have poor dynamic accuracy, due to ambiguities introduced by linear acceleration. First we analyse the effect of linear acceleration on orientation accuracy. We then present an efficient distributed method which uses a model of the subject's body structure to estimate and correct for linear acceleration. We validate the behaviour of this method on data from combined optical/inertial capture experiments, and show improved gravity vector estimation and a corresponding increase in orientation accuracy. We estimate the runtime, memory, communication and power requirements of our method, and show that it is a practical software modification to an existing system. The proposed solution is the first to use collaboration between wireless IMUs to improve accuracy.

This paper details the design and implementation of a low power transceiver for free space optical communication (FSO) in wireless sensor networks. The design and implementation of the transceiver using commercial electronic and optical is described. The properties of the FSO communication link are examined and deployment issues arising from using FSO communications in sensor networks are discussed. The use of the FSO communication channel to reduce latency and improve performance in a class of wireless sensor network applications is analysed. The FSO communication channel can reduce latency introduced by low power duty cycled MAC protocols by an order of magnitude using a receiver with a current consumption of less than 100 uA.

This video presents a new method for mobile unobtrusive interaction with bipedal robots using the Orient on-body, fully wireless motion capture system. During the learning phase for the robot, data for motion such as waving the hands, standing on a leg, performing sit-ups and squats is captured from a human operator strapped with the Orient specks. Key features are extracted from the captured motion data using unsupervised learning algorithms. During subsequent interactions with the robot, the motion of the human operator, speckled with Orients, is classified on-line and the robot selects to play the closest motion. This approach is particularly useful in situations where the robot operates a well defined vocabulary of motion, and the advantages are the speed with which new robot motion behaviour can be programmed in a matter of minutes (compared to a heuristics-based approach), and the mobility (compared to a camera-based method) that this mode of interaction affords. Related papers have compared the performances of three unsupervised learning algorithms: c-means, k-means and expectation maximisation (EM) for the four motion scenarios described above, and for walking. The video is also available at http://www.specknet.org/about/edinburgh/DKArvindMBarto sik_RO-MAN2009.mpeg.

This paper presents, to the best of our knowledge, the first instance of real-time human-robot interaction using motion capture (mocap) data obtained from fully wireless, on-body sensor networks. During the learning phase, data for motion such as waving of the hands, standing on a leg, performing sit-ups and squats is captured from a human strapped with the orient motion capture specks. Key features are extracted from the captured motion data using unsupervised learning algorithms. During subsequent interactions with the robot, the motion of the operator, speckled with orients, is classified and the robot selects to play the closest motion. This approach is particularly useful in situations where the robot operates a well defined vocabulary of motion, and the advantages are the real-time interaction and the rapidity (in a matter of minutes) in programming new behaviour compared to a heuristics-based approach. This paper compares the performances of three unsupervised learning algorithms: c-means, k-means and expectation maximisation (EM) for the four motion scenarios. Nine best candidates for the three learning algorithms for each of the four motion scenarios were selected in the Webots robot simulator and then transferred to the real robot. Metrics were defined for each motion scenario and their performances compared for the three learning algorithms. In all the cases the motions were able to be imitated; c-means was the best, followed closely by the k-means algorithms, and the reasons have been analysed.

This project presents a new display concept, which brings together haptics, augmented and mixed reality and tangible computing within the context of an intuitive conceptual design environment. The project extends the paradigm of augmented prototyping by allowing modelling of virtual geometry on the physical prototype, which can be touched by means of a haptic device. Wireless tracking of the physical prototype is achieved in three different ways by attaching to it a 'Speck', a tracker and Nintendo Wii Remote and it provides continuous tangible interaction. The physical prototype becomes a tangible interface augmented with mixed reality and with a novel 3D haptic design system.

The Research Consortium in Speckled Computing is a multidisciplinary grouping of computer scientists, electronic engineers, physicists and electrochemists with the aim of realising wireless networks of miniature programmable speck devices. Each speck is designed to be autonomous with its own rechargeable battery and combines the capabilities to sense, process data and communicate wirelessly. This review paper outlines the unique multidisciplinary approach to the design of networks of specks and how the design of energy-efficient specks should take into account the intimate interactions between the different sub-systems which make up the specks, such as the energy source, wireless communication, physical layer design and the Medium Access Control (MAC) and routing protocols.

This paper presents the results of experiments to compare the performance of energy-efficient SpeckMAC Medium Access Control (MAC) algorithms, SpeckMAC-D and SpeckMAC-B, with the Berkeley MAC for miniature resource-constrained mobile ad-hoc networks, such as Specknets. These MAC algorithms conserve energy by reducing the amount of idle-listening time during communications by periodic sampling of the radio channel. Previous work had compared the three algorithms on the ProSpeckz hardware in terms of the lifetime of the batteries powering the platform: SpeckMAC protocols achieved longer lifetimes compared to B-MAC, especially for batteries with smaller capacities. This paper presents further analysis of the performance of the MAC algorithms through experiments, on the Perspeckz-64, a test-bed of 64 ProSpeckz nodes, for power consumption of unicast and broadcast traffic, the delivery ratio. and one-hop and multi-hop latencies. The results demonstrated that the SpeckMAC algorithms outperformed B-MAC for all the metrics considered.

This paper presents a distributed algorithm to direct evacuees to exits through arbitrarily complex building layouts in emergency situations. The algorithm finds the safest paths for evacuees taking into account predictions of the relative movements of hazards, such as fires, and evacuees. The algorithm is demonstrated on a 64 node wireless sensor network test platform and in simulation. The results of simulations are shown to demonstrate the navigation paths found by the algorithm.

Speckled computing (Arvind, D.K. and Wong, K.J., Proc. IEEE Int. Symp. on Consumer Electronics, p.219-23, 2004) is an emerging technology in which data is sensed in minute (eventually one cubic millimetre) semiconductor grains called specks. Information is extracted in situ from each speck and is exchanged and processed in a collaborative fashion in a wireless network of thousands of specks, called a specknet. Specks are not assumed to be static, and therefore estimating and maintaining the logical location of the mobile specks in a network is essential for a number of speckled computing and sensor network applications. A novel lightweight distributed algorithm is introduced for this purpose and simulation results are presented to determine the goodness of the algorithm for different parameters, The algorithm was also successfully ported to a hardware prototype of the speck called the ProSpeckz. The problems and issues of porting the algorithm onto such a resource-constrained hardware platform are discussed. Finally, the paper concludes with plans to improve the algorithm.

The design and design automation of embedded systems is rapidly emerging as a research area in its own right. It draws from several traditional areas of study such as system specification, modeling and analysis; computer architecture and micro-architecture; as well as compilers and operating systems. However, the embedded domain adds some interesting twists in terms of tighter problem constraints that demand a fresh look at even these traditional areas. In addition, there are several emerging EDA areas such as design reuse and integration of systems on a chip that are critical to the study of embedded systems. These aspects are not typically covered by computer engineering and EDA curricula. This panel addresses the challenges associated with the educational issues in embedded systems design and design automation. The panelists will examine issues in including embedded systems in university curricula, as well as in setting up research programs that are crucial for the education of graduate students.