Search

Due to the low cost and good privacy protection of binary sensors, there are many applications of binary sensors. To enhance the spatial awareness of binary sensors, researchers utilize modulators to modulate views of sensors into visible and invisible regions so that the monitoring space is segmented into small cells identified by signatures. When a warm object moves in these cells, its location or moving trajectory can be acquired more accurately with modulators than without modulators. Accordingly, the maximum number of cells (MNC) in a deployment determines partially the maximal spatial awareness of the binary sensor system. In this paper, we provide a theoretical study of the MNC, given the number of sensors and the number of modulators. We also find the sufficient and necessary conditions to achieve the MNC so that we provide the reasons why deployment cannot obtain the MNC. These conditions can guide researchers to design the MNC deployments. Furthermore, we provide a method to calculate the number of cells when a deployment sometime cannot obtain the MNC. Our experiments provide deep insights into the influences of those conditions on the MNC and the MNC on the spatial awareness of binary sensor systems.

Continual learning is an emerging research challenge in human activity recognition (HAR). As an increasing number of HAR applications are deployed in real-world environments, it is important and essential to extend the activity model to adapt to the change in people’s activity routine. Otherwise, HAR applications can become obsolete and fail to deliver activity-aware services. The existing research in HAR has focused on detecting abnormal sensor events or new activities, however, extending the activity model is currently under-explored. To directly tackle this challenge, we build on the recent advance in the area of lifelong machine learning and design a continual activity recognition system, called HAR-GAN , to grow the activity model over time. HAR-GAN does not require a prior knowledge on what new activity classes might be and it does not require to store historical data by leveraging the use of Generative Adversarial Networks (GAN) to generate sensor data on the previously learned activities. We have evaluated HAR-GAN on four third-party, public datasets collected on binary sensors and accelerometers. Our extensive empirical results demonstrate the effectiveness of HAR-GAN in continual activity recognition and shed insight on the future challenges.

International audience; This paper studies the identification of infinite impulse response (IIR) systems with binary-valued measurements. The key to our approach is to recast the initial identification problem into an identification problem in the set-membership framework. This enables us to identify the system using an identification algorithm devoted to IIR systems in such a framework. Although the proposed algorithm is designed for noise free data, an analysis is provided in the presence of noise, and illustrative examples are provided to demonstrate the effectiveness of the algorithm.

Producción Científica, In recent years, Artificial Intelligence Technologies (AIT) have been developed to improve the quality of life of the elderly and their safety in the home. This work focuses on developing a system capable of recognising the most usual activities in the daily life of an elderly person in real-time to enable a specialist to monitor the habits of this person, such as taking medication or eating the correct meals of the day. To this end, a prediction model has been developed based on recurrent neural networks, specifically on bidirectional LSTM networks, to obtain in real-time the activity being carried out by the individuals in their homes, based on the information provided by a set of different sensors installed at each person’s home. The prediction model developed in this paper provides a 95.42% accuracy rate, improving the results of similar models currently in use. In order to obtain a reliable model with a high accuracy rate, a series of processing and filtering processes have been carried out on the data, such as a method based on a sliding window or a stacking and re-ordering algorithm, that are subsequently used to train the neural network, obtained from the public database CASAS., Ministerio de Ciencia, Innovación y Universidades (project RTI2018-096652-B-I00), Junta de Castilla y León - Fondo Europeo de Desarrollo Regional (project VA233P18)

Unobtrusive activity recognition is known to be the most preferred solution for monitoring daily activities of elderly people. In this paper, Scanpath Trend Analysis (STA) is employed for unobtrusive activity recognition of elderly people living alone. Binary sensor data are used and each activity is considered as a sequence of sensor points for this purpose. The real-world long-term fully annotated Aruba open dataset collected by binary sensors is used for the verification of accuracy and the efficacy of the proposed approach. With the STA, the F1-score of 0.758 is obtained, and furthermore, by adding some extra semantic information through an activity transition matrix, it is possible to have F1-score as 0.863. This F1-score is superior to all the related works that use binary sensor data for activity prediction, while computationally the approach presented is advantageous since long periods of training process can be avoided.

In this paper we propose a novel method for location recognition and prediction in smart homes based on semi-supervised learning. We use data collected from low-resolution depth video cameras installed in four apartments with older adults over 70 years of age, and collected during a period of one to seven weeks. The location of the person in the depth images is detected by a person detection algorithm adapted from YOLO (You Only Look Once). The locations extracted from the videos are then clustered using K-means clustering. Sequence prediction algorithms are used to predict the next cluster (location) based on the previous clusters (locations). The accuracy of predicting the next location is up to 91%, a significant improvement compared to the case where binary sensors are placed in the apartment based on human intuition. The paper presents an analysis on the effect of the memory length (i.e. the number of previous clusters used to predict the next one), and on the amount of recorded data required to converge.

This article is concerned with the secure particle filtering problem for a class of discrete-time nonlinear cyber-physical systems with binary sensors in the presence of non-Gaussian noises and multiple malicious attacks. The multiple attacks launched by the adversaries, which take place in a random manner, include the denial-of-service attacks, the deception attacks, and the flipping attacks. Three sequences of Bernoulli-distributed random variables with known probability distributions are employed to describe the characteristics of the random occurrence of the multiple attacks. The raw or corrupted measurements are transmitted to sensors, whose outputs are binary according to engineering practice. A modified likelihood function is constructed to compensate for the influence of the randomly occurring multiple attacks by introducing the random occurrence probability information into the design process. Subsequently, a secure particle filter is proposed based on the constructed likelihood function. Finally, a moving target tracking application is elaborated to verify the viability of the proposed secure particle filtering algorithm.

Ambient Assisted Living (AAL) is becoming crucial to help governments face the consequences of the emerging ageing population. It aims to motivate independent living of older adults at their place of residence by monitoring their activities in an unobtrusive way. However, challenges are still faced to develop a practical AAL system. One of those challenges is detecting failures in non-intrusive sensors in the presence of the non-deterministic human behaviour. This paper proposes sensor failure detection and isolation system in the AAL environments equipped with event-driven, ambient binary sensors. Association Rule mining is used to extract fault-free correlations between sensors during the nominal behaviour of the resident. Pruning is then applied to obtain a non-redundant set of rules that captures the strongest correlations between sensors. The pruned rules are then monitored in real-time to update the health status of each sensor according to the satisfaction and/or unsatisfaction of rules. A sensor is flagged as faulty when its health status falls below a certain threshold. The results show that detection and isolation of sensors using the proposed method could be achieved using unlabelled datasets and without prior knowledge of the sensors&rsquo, topology.

Aging is a global challenge facing our society in the next few decades. Ambient assisted living (AAL) is a promising technology that helps people stay active, socially connected, and independent into older age. Even though ambient binary sensors, such as passive infrared (PIR) motion sensors, offer a low cost, easy to deploy, and less intrusive solution to constructing a smart environment, the limited ability of coping with multiple residents hinders the wide adoption of the AAL technology. In this work, we present three multiresident tracking algorithms, nearest neighbor with sensor graph (NN-SG), global nearest neighbor with sensor graph (GNN-SG), and multiresident tracking with sensor vectorization (sMRT), to solve the data association problem between the ambient sensor events and residents in the smart environment. We also introduce new performance metrics to evaluate the success of alternative approaches to multiresident tracking in smart homes. We evaluate all the algorithms with a recent smart home dataset recorded in real-life settings. Among the three algorithms, NN-SG and GNN-SG require sensor location and floor plan of the environment to derive the sensor graph, while sMRT does not require such information and relies solely on the unannotated sensor data. As an initiative of the unsupervised resident tracking solution, sMRT prompts additional research opportunities in multiresident tracking to improve the adoption of AAL technology in our daily life.

In this paper, source localization by a network of primitive binary sensors under various imperfections are studied. Detailed analysis and mathematical modeling of imperfect binary sensors are presented. Imperfections include sensor failures of two types, drifting, uncertainty, and heterogeneity in binary sensor trigger thresholds, presence of noise, and nonradial symmetry of sensing ranges. Theoretical results, including asymptotical convergence, are established, in particular in the presence of substantial outliers due to sensor failure and large noise. Efficient numerical algorithms are proposed and simulated supporting the theoretical analysis.

Many elderly persons prefer to stay alone in a single-resident house for seeking an independent life and reducing the cost of health care. However, the independent life cannot be maintained if the resident develops dementia. Thus, an early detection of dementia is essential for the elderly to extend their independent lifetime. Early symptoms of dementia can be noticed in everyday activities such as front-door events. For example, forgetting something when the person leaves the house might be an early symptom of dementia. In this paper, we introduce a novel front-door events [exit, enter, visitor, other, and brief-return-and-exit (BRE)] classification scheme that validated by using open data sets ( $n =14$ ) collected from 14 testbeds by anonymous wireless binary sensors (passive infrared sensors and magnetic sensors). BRE events occur when four consecutive events (exit-enter-exit-enter) happen in certain time intervals ( $t_{1}$ , $t_{2}$ , and $t_{3})$ , and some of them may be the forget events. Each testbed had one older adult (aged 73 and over) during the experimental period ( $\mu = 547.6 \pm 370.4$ days). The algorithm automatically classifies the resident’s front-door events. Experimental results show the events of total exits, daily exits, out-time per exit, as well as the significance of the $\text{t}_{\mathrm {i}}$ parameters for the number of classified BRE events. Since part of the BRE events may be the forget events, the proposed algorithm could be a useful tool for the forget event detection.

We present a comprehensive study of state-of-the-art algorithms for the prediction of sensor events and activities of daily living in smart homes. Data have been collected from eight smart homes with real users and 13-17 binary sensors each – including motion, magnetic, and power sensors. We apply two probabilistic methods, namely Sequence Prediction via Enhanced Episode Discovery and Active LeZi, as well as Long Short-Term Memory Recurrent Neural Network, in order to predict the next sensor event in a sequence. We compare these with respect to the required number of preceding sensor events to predict the next, the necessary amount of data to achieve good accuracy and convergence, as well as varying the number of sensors in the dataset. The best-performing method is further improved by including information on the time of occurrence to predict the next sensor event only, and in addition to predict both the next sensor event and the mean time of occurrence in the same model. Subsequently, we apply transfer learning across apartments to investigate its applicability, advantages, and limitations for this setup. Our best implementation achieved an accuracy of 77-87% for predicting the next sensor event, and an accuracy of 73-83% when predicting both the next sensor event and the mean time elapsed to the next sensor event. Finally, we investigate the performance of predicting daily living activities derived from the sensor events. We can predict activities with an accuracy of 61-90%, depending on the apartment. This work was supported by the Norwegian Research Council through the SAMANSVAR programme under Grant 247620/O70.

Low cost robots, such as vacuum cleaners or lawn mowers, employ simplistic and often random navigation policies. Although a large number of sophisticated localization and planning approaches exist, they require additional sensors like LIDAR sensors, cameras or time of flight sensors. In this work, we propose a global localization method biologically inspired by simple insects, such as the ant Cataglyphis that is able to return from distant locations to its nest in the desert without any or with limited perceptual cues. Like in Cataglyphis, the underlying idea of our localization approach is to first compute a pose estimate from pro-prioceptual sensors only, using land navigation, and thereafter refine the estimate through a systematic search in a particle filter that integrates the rare visual feedback. In simulation experiments in multiple environments, we demonstrated that this bioinspired principle can be used to compute accurate pose estimates from binary visual cues only. Such intelligent localization strategies can improve the performance of any robot with limited sensing capabilities such as household robots or toys., Accepted to BIOSIGNALS 2019

In this paper, we present work in progress on activity recognition and prediction in real homes using either binary sensor data or depth video data. We present our field trial and set-up for collecting and storing the data, our methods, and our current results. We compare the accuracy of predicting the next binary sensor event using probabilistic methods and Long Short-Term Memory (LSTM) networks, include the time information to improve prediction accuracy, as well as predict both the next sensor event and its time of occurrence using one LSTM model. We investigate transfer learning between apartments and show that it is possible to pre-train the model with data from other apartments and achieve good accuracy in a new apartment straight away. In addition, we present preliminary results from activity recognition using low resolution depth video data from seven apartments, and classify four activities – no movement, standing up, sitting down, and TV interaction – by using a relatively simple processing method where we apply an Infinite Impulse Response (IIR) filter to extract movements from the frames prior to feeding them to a convolutional LSTM network for the classification. Financed by the Norwegian Research Council under the SAMANSVAR programme (247620/O70).

In spite of the fact that many sensors in use today are binary (i.e. produce only values of 0 and 1), and that useful context-aware applications are built exclusively on top of them, there is currently no development approach specifically targeted to binary sensors. Dealing with notions of state and state combinators, central to binary sensors, is tedious and error-prone in current approaches. For instance, developing such applications in a general programming language requires writing code to process events, maintain state and perform state transitions on events, manage timers and/or event histories. In another paper, we introduced a domain specific language (DSL) called Allen, specifically targeted to binary sensors. Allen natively expresses states and state combinations, and detects contexts on line, on incoming streams of binary events. Expressing state combinations in Allen is natural and intuitive due to a key ingredient: semi-causal operators. That paper focused on the concept of the language and its main operators, but did not address its implementation challenges. Indeed, online evaluation of expressions containing semi-causal operators is difficult, because semi-causal sub-expressions may block waiting for future events, thus generating unknown values, besides 0 and 1. These unknown values may or may not propagate to the containing expressions, depending on the current value of the other arguments. This paper presents a compiler and runtime for the Allen language, and shows how they implement its state combining operators, based on reducing complex expressions to a core subset of operators, which are implemented natively. We define several assisted living applications both in Allen and in a general scripting language. We show that the former are much more concise in Allen, achieve more effective code reuse, and ease the checking of some domain properties.

Many real-world applications, which are focused on addressing the needs of a human, require information pertaining to the activities being performed. The UCAmI Cup is an event held within the context of the International Conference on Ubiquitous Computing and Ambient Intelligence, where delegates are given the opportunity to use their tools and techniques to analyse a previously unseen human activity recognition dataset and to compare their results with others working in the same domain. In this paper, the human activity recognition dataset used relates to activities of daily living generated in the UJAmI Smart Lab, University of Jaén. The dataset chosen for the first edition of the UCAmI Cup represents 246 activities performed over a period of ten days carried out by a single inhabitant. The dataset includes four data sources: (i) event streams from 30 binary sensors, (ii) intelligent floor location data, (iii) proximity data between a smart watch worn by the inhabitant and 15 Bluetooth Low Energy beacons and (iv) acceleration of the smart watch. In this first edition of the UCAmI Cup, 26 participants from 10 different countries contacted the organizers to obtain the dataset.‬‬‬‬‬

Activities of Daily Living (ADLs), or a person’s routine activities of self-care, are important factors influencing the feasibility of home health care or aging in place for many individuals. Automated, sensor-based recognition of such activities affords home stay, greater independence and privacy, and improved quality of life to individuals who would require stay in a supervised or medical facility. This paper describes a data-driven framework for the design and deployment of such an automated system for activity recognition using simple, unobtrusive, and privacy-friendly binary sensors. It presents the results of an experimental study, with both numerical and qualitative observations, of this framework on a publicly available real dataset.

Although there have been many studies aimed at the field of Human Activity Recognition, the relationship between what we do and where we do it has been little explored in this field. The objective of this paper is to propose an approach based on machine learning to address the challenge of the 1st UCAmI cup, which is the recognition of 24 activities of daily living using a dataset that allows to explore the aforementioned relationship, since it contains data collected from four data sources: binary sensors, an intelligent floor, proximity and acceleration sensors. The methodology for data mining projects CRISP-DM was followed in this work. To perform synchronization and classification tasks a java desktop application was developed. As a result, the accuracy achieved in the classification of the 24 activities using 10-fold-cross-validation on the training dataset was 92.1%, but an accuracy of 60.1% was obtained on the test dataset. The low accuracy of the classification might be caused by the class imbalance of the training dataset; therefore, more labeled data are necessary for training the algorithm. Although we could not obtain an optimal result, it is possible to iterate in the methodology to look for a way to improve the obtained results.

This short paper presents the activity recognition results obtained from the CAR-CSIC team for the UCAmI’18 Cup. We propose a multi-event naive Bayes classifier for estimating 24 different activities in real-time. We use all the sensorial information provided for the competition, i.e., binary sensors fixed to everyday objects, proximity BLE-based tags, location-aware smart floor sensing and the wrist’s acceleration. The results using training data-sets of 7 days show accuracies (true positives) about 68%; however for the three extra data-sets of the competition we were able to reach a 60.5% accuracy.

The challenge of recognizing different personal activities while living in an apartment is of great interest for the AAL community. Many different approaches have been presented trying to achieve good accuracies in activity recognition, combined with different heuristics, windowing and segmentation methods. In this paper we want to revisit the basic methodology proposed by a naive Bayes implementation with emphasis on multi-type event-driven location-aware activity recognition. Our method combines multiple events generated by binary sensors fixed to everyday objects, a capacitive smart floor, the received signal strength (RSS) from BLE beacons to a smart-watch and the sensed acceleration of the actor’s wrist. Our new method does not use any segmentation phase, it interprets the received events as soon as they are measured and activity estimations are generated in real-time without any post-processing or time-reversal re-estimation. An activity prediction model is used in order to guess the more-likely next activity to occur. The evaluation results show an improved performance when adding new sensor type events to the activity engine estimator. Classification results achieve accuracies of about 68%, which is a good figure taking into account the high number of different activities to classify (24).

We present preliminary results on sensor data prediction in a smart home environment with a limited number of binary sensors. The data has been collected from a real home with one resident over a period of 17 weeks. We apply Recurrent Neural Network with Long Short-Term Memory to a text sequence derived from the sensors’ events to predict the next event in a sequence. We compare our system’s characteristics and results to a baseline method and to similar work in the area. Our implementation achieved a peak accuracy of 69% for a set with 13 sensors in total - motion, magnetic and power sensors - and 75% for five motion sensors. Financed by the Norwegian Research Council under the SAMANSVAR programme (247620/O70).

In this thesis, we deal with the identification of systems and the extraction of patterns from data. In the context of system identification, we focus precisely on the identification of hybrid systems and the identification of linear systems using binary sensors. Two very popular classes of hybrid systems are switched linear systems and piecewise affine systems. First, we give an overview of the different approaches available in the literature for the identification of these two classes. Then, we propose a new real-time identification algorithm for switched linear systems, it's based on an Outer Bounding Ellipsoid (OBE) type algorithm suitable for system identification with bounded noise. We then present several extensions of the algorithm either for the identification of piecewise affine systems, the identification of switched linear systems described by an output error model and the identification of MIMO switched linear systems. After this, we address the problem of the identification of linear systems using binary sensors by introducing an original point of view. We formulate the identification problem as a classification problem. This formulation allows the use of supervised learning algorithms such as Support Vector Machines (SVMs) for the identification of discrete time systems and the identification of continuous-time systems using binary sensors. In the context of pattern extraction, we first present an overview of the different pattern extraction algorithms and clustering techniques available in the literature. Next, we propose an algorithm for extracting patterns from data based on clustering techniques., Les travaux de cette thèse portent sur l'identification des systèmes et l'extraction de motifs à partir de données. Dans le cadre de l'identification des systèmes, nous nous intéressons plus précisément à l'identification des systèmes dynamiques hybrides et l'identification des systèmes dynamiques linéaires ayant une sortie binaire. Deux classes très populaires des systèmes hybrides sont les systèmes linéaires à commutations et les systèmes affines par morceaux. Nous faisons tout d'abord un état de l'art sur les méthodes d'identification de ces deux classes. Nous proposons ensuite un algorithme basé sur une méthode d'identification de type OBE "Outer Bounding Ellipsoid" pour l'identification en temps réel des systèmes à commutations soumis à un bruit borné. Nous présentons ensuite plusieurs extensions de l'algorithme soit pour l'identification des systèmes affines par morceaux, l'identification des systèmes à commutations décrits par un modèle du type erreur de sortie et l'identification des systèmes MIMO à commutations. Nous abordons ensuite le problème d'identification des systèmes linéaires ayant une sortie binaire en introduisant un point de vue original consiste à formuler le problème d'identification comme un problème de classification. Ceci permet de proposer deux algorithmes d'identification basés sur l'utilisation des SVMs. Le premier algorithme est dédié à l'identification des systèmes à temps discret et le deuxième algorithme est dédié à l'identification des systèmes à temps continu. Dans le cadre de l'extraction de motifs, nous présentons dans un premier temps un état de l'art sur les algorithmes d'extraction de motifs et sur les techniques de la classification non supervisée. Ensuite, nous proposons un algorithme d'extraction de motifs à partir des données basé sur des techniques de classification non supervisée.

This paper considers the problem of localising a signal source using a team of mobile agents that can only detect the presence or absence of the signal. A background false detection rate and missed detection probability are incorporated into the assumptions. An estimation algorithm is proposed that discretizes the search environment into cells, and uses Bayesian techniques to approximate the posterior probability of each cell containing the source. Analytical results are presented for a range of specific cases, and simulations are used to investigate more complex scenarios.