Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

In recent years there is an increasing attention on the topic of “prolonging independent living”. Several initiatives all over the world have focused on the problem of helping aging population with technology [3, 7, 12] and funding programs have been triggered like the Ambient Assisted Living (AAL), promoted by the European Commission together with the topic “ageing well” within the FP7 research area. Such an attention for the problem has been increased in the starting Horizon 2020 program. The general aim is the one of promoting a healthier society that constitutes a main social and economic challenge. In fact, most elderly people aim at remaining in their homes as long as possible as this is in general conducive of a richer social life and paramount to maintaining established habits. To adhere to this wish is also positive from an economic perspective as the cost of care at home is almost always much less than the cost of residential care. Several issues need to be addressed in order to prolong independent living. One is the early detection of possible deterioration of health so that problems can be remediated as soon as they are arising with a timely involvement of health care providers and family [9]. A second issue consists in providing adaptive support to assist in coping with age-related impairments [2, 4, 11]. Third, ways of supporting preventive medicine must be found as it has been increasingly recognized that preventive medicine can contribute to promote a healthy lifestyle and delay the onset of age-related illnesses.

Observing the current efforts in the AAL projects, and the trends in the “internet of things” [1, 10], an amount of R&D effort exists in issues like the gathering of continuous information at home, the standardization of formats in order to create environments more easily, the extraction of further information from raw data using different techniques to reconstruct context, etc. This paper aims at contributing a particular perspective to the general debate by posing the attention toward the work needed to bring an intelligent environment in continuous use in a real home. It points the attention to the need to an end-to-end perspective in organizing an intelligent environment, keeping into account the services offered to the users of such systems in which the key services are organized and tested. Following this vision, this paper introduces the general ideas of the intelligent environment pursued by the GiraffPlus project, points out its components and services then describes the core choices for the basic level and the work done for achieving more challenging capabilities at the “enhanced level”. The GiraffPlus system has been developed also thanks to an incremental test of the complete system in laboratory, in a pilot site and currently in a complete deployment in 15 real homes in three different European Countries (Italy, Spain and Sweden).

2 The GiraffPlus System Functionalities

Figure 1 offers a conceptual schematization of the GiraffPlus components of the system and allows to identify some key concepts relevant in the project work and useful to contextualize the current paper.

Fig. 1
figure 1

The general information flow inside the GiraffPlus system

Full size image

Given the attention to human actors that use the intelligent environment we first identify the different users of the system services: (a) the primary user is the older adult living at home, mostly alone, that the system is supposed to actively support; (b) the secondary users are a network of people who participate in the support of the older adult from outside his/her home. Such users can be formal caregivers (a doctor, the nurse, etc.), informal caregivers (a son, a group of generic relatives, etc.) or simply friends that want to maintain a contact with the old person.

The GiraffPlus system is basically composed of four parts: (1) a network of sensors deployed in the home that continuously gathers data; (2) a data management infrastructure that guarantees either data gathering in a permanent data store or direct information delivery to some external user in real time. A central role in the infrastructure is played by the middleware software; (3) a telepresence robot (the Giraff) that guarantees communication between people outside the house and the primary user inside the house, enriching such dialogue with the possibility of moving in the home environment and performing visual monitoring through a camera connected to the robot [5]; (4) a personalizable interaction front-end that allows to visualize data from the house, to call the robot from outside the house, and to access some specialized services, like those of reporting and reminding.

A first result has been presented in [6] where the choices for the middleware and the data store have been demonstrated as the key building block for the initial version of the GiraffPlus intelligent environment and AAL services. Then, given these ingredients, a key aspect for creating a useful tool for real people are the services that the complete system can deliver to the different users. Looking again at the figure it is worth highlighting that there are two different action paths: from inside the house to outside, and viceversa. If we consider the from-inside-to-outside path we can say that we are: (a) exporting data for a long term data analysis (storing them first in a data storage service). Notice that the secondary users usually are heterogeneous people having different “social goals” toward the old person, hence a doctor may be interested in physiological data and general information on the daily activities connected to health, an informal caregiver may be interested in a daily summary that says “everything is OK”, “the windows were left open”, etc. (b) exporting data for a real time use (hence, for example, for issuing alarms). In fact, the system can rise alarms and/or send warnings, for instance, in case of falls or in case of abnormal physiological parameters. Here a problem exists of delivering them timely and to the right person. On the other hand, if we explore the from-outside-to-inside path we can say that: (c) the communication through the telepresence robot is the basic media for social communication from outside into the house; additionally (d) having such a general set up, the system offers now a channel from the secondary to the primary user for messages and reminders created through the visualisation front-end and delivered on the robot screen (also other media are possible but here we are focusing on the basic GiraffPlus components). This information exchange is stored by the data storage service for subsequent analysis.

In addition, it is worth underscoring how there are also other services that are possible given the infrastructure in Fig. 1: (e) long term data from the home sensor network are available in the data storage and can be also shown, with some attention, to the primary user; (f) having the home sensor network and a communication media in the home through the robot screen opens up the possibility of reasoning on those data and synthesize different messages (for the moment we assume precompiled messages) triggered by such a reasoning (for example, to react to some information contained in the data).

3 Managing Sensor Data to Synthesize Additional Information

In order to satisfy the requirements of easy installation and acceptance of the system, in GiraffPlus commercially available sensors are used to guarantee a good appearance and form factor, battery-powered, and durability for the long-term monitoring. The environmental sensors used in the GiraffPlus system are provided by Tunstall.Footnote 1 FAST Passive Infrared (PIR) motion detectors, Electrical Usage Sensor, and Universal sensor that can be configured according to different needs (Door Usage, Bed/Chair Occupancy) have been used.

This section shows how the information coming from such kind of sensors deployed in the environment can be processed in order to infer more meaningful input for context analysis. An important information about the user context is its localization in terms of the room in which the user is. Hence a first service introduced here is a simple reasoner for continuous person tracking.

The fundamental studies of target tracking often focus on networks composed of sensor nodes with the most elementary sensing capabilities that provide just binary information about the target, indicating whether it is present or absent in the sensing range of a node [8]. These so-called binary sensor networks constitute the simplest type of sensor networks that can be used for target tracking. Each sensor has the ability of indicating the presence of a user in a room. For instance, when a magnetic sensor puts on the door indicates that the door has changed its status (open/closed), this means that the user is near to the sensor itself. However, each sensor has different features that must be individually treated in order to infer the user’s position.

Each sensor, when is individually leveraged, does not give a lot of information. However, these sensors have different properties which, when exploited together, can reveal a surprising amount of information. Thus, in order to detect the position of the user in the house, each signal provided by the sensor has been first filtered and later processed together with the other signals by the proposed algorithm. In particular, signals coming from the magnetic contacts and from the electrical usage sensors have been processed in order to obtain information on when they change their status i.e., when the users is in front of them. Moreover, the spikes produced by the electrical usage sensor of the personal computer has been filtered by using the median filter. The median filter is a nonlinear digital filtering technique, typically used in image processing in order to reduce the noise. In order to eliminate spikes, the median filter slides the entire signal, entry by entry, replacing each entry with the median of the sample in the time window T.

After the filtering process, the samples coming from the binary sensor network deployed in the environment are processed by the “where is” (WHIZ) algorithm (Algorithm 1).

For each second t in the time window T, WHIZ collects the data coming from each of the N sensors deployed in the environment \(s_{i} ,\forall i \subseteq \{ 1,2, \cdots ,N\}\), then the number of occurence of \(s_{i}\) is weighted with the parameter \(w_{i}\) according to Eq. 1 obtaining the vector \(\bar{S}\). Indeed, the algorithm weights more the information coming from sensors that directly provide the user’s position (such as the force/pressure sensors).

$$\bar{S} = [s_{1} \quad s_{2} \cdots s_{N} ] \times \left[ {\begin{array}{*{20}c} {e_{{\mathbf{1}}} } & 0 & \cdots & 0 \\ 0 & {e_{{\mathbf{2}}} } & \cdots & 0 \\ \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & \cdots & {e_{N} } \\ \end{array} } \right]$$
(1)

where \(e_{i}\) is a ones array of length \(w_{i}\). Latter, \(\bar{S}\) is translated into R, where each \(r_{j}\) represents one of the M room in which the sensor is deployed (\(r_{j} \forall j \subseteq \{ 1,2, \cdots ,M\}\)). Finally, the WHIZ algorithm evaluates the mode (the most frequent value in the data set) of the vector resulting from the concatenation of the T vectors \(R_{t - i} = [r_{{\mathbf{1}}}^{t - i} ,r_{ 2}^{t - i} , \cdots ,r_{\text{M}}^{t - i} ]\), where \(i = 1, \cdots ,T\). The main idea of using the mode is due to the fact that the presence of a user in a given room will affect a greater number of sensors in that room. Therefore we obtain the room occupied by the user \(r^{*}\) applying Eq. 2.

$$r^{*} = mode\left( {R_{t} ,R_{t - 1} , \cdots ,R_{t - W} } \right)$$
(2)

Figure 2 shows the results of the WHIZ algorithm applied on a subset of data coming from one of the GiraffPlus testsites.

Fig. 2
figure 2

The results of the WHIZ room-level localization technique

Full size image

4 User Interaction Services

After having obtained reliable data it is important to endow the system with an effective capability to transform data in services for users. In GiraffPlus we have designed and realised a set of interactive services that aims at going beyond simple visualization capabilities of the raw data. A specific module called the Data Visualization, Personalization and Interaction Service (DVPIS) has been realized to manage interaction with the different actors involved in the AAL scenario. In particular, two different instances of the DVPIS have been created: one devoted to serve secondary users (DVPIS@Office), and another dedicated to the primary users (DVPIS@Home). The basic goals pursued by the two GiraffPlus parts are tailored to the diversity of users: secondary users need to be supported by a flexible and efficient monitoring tool; primary users may take advantage from being aware of the information on their own health condition to foster them better manage their health and lifestyle, but also would benefit from an enriched set of communication features from their assistive social network.

Figure 3 illustrates the general idea of the DVPIS. The module is composed of a back end part that is devoted to both organize the content of the information to be shown to the users and also to offer different services tailored to classes of users. The DVPIS back end is responsible for preparing “personalized information to the users” and to offer support for different types of services like reminders, reports and alarms (and also the proactive suggestions and warnings). The front end part is responsible for presenting the information and services to the different categories of users. The project has developed two modules: (1) the @Office devoted to the secondary users and currently runs on their personal computer (“their office workstation”) and (2) the @Home dedicated to the primary users that runs as an additional service on the Giraff robot.

Fig. 3
figure 3

The data visualization and personalization services

Full size image

In the rest of this section we introduce some of the relevant features that have been included in the more recent release of the DVPIS services.

4.1 The @Office Environment

Figure 4 shows a composition of screens from the @Office service. Specifically, the module is adapted to different type of end-users, hence medical doctors may access different GiraffPlus houses that they while a parent may access exclusively his/her relative. The @Office environment for the single home is subdivided in three sub environments (accessible by the three buttons highlighted with label (1) in the figure): (a) an environment (button “Long Term”) contains different services that interacting with the data storage of GiraffPlus allow to inspecting long term data on all the sensors active in the home; (b) a second environment (button “Real Time”) enable a visualisation of current data from the house; (c) a third, button called “People”, creates a communication service for all the people involved in the assistance with the house.

Fig. 4
figure 4

Some screenshots of the @Office environment

Full size image

The Long Term environment for visualisation (not shown in the figure) allows to access several information. A main distinction concerns the one between environmental data (from the sensors distributed in the house) and physiological data (from a set of tools that allows to measure physical parameters (e.g., blood pressure, glucose level, weight). The interested user can also access a specific environment, called “Context Reasoner”, that creates abstractions with respect to raw data in the attempt to infer additional context features by checking combination of raw sensor values over time (such capability is developed by our Swedish colleagues [13]). Furthermore we created a “Report Capability” that through a dialogue window allows for the generation of aggregated statistics over period of times, offering again a more abstract representation of the average activities in the house. The combination of the different capabilities is aims at giving the user a view of the person at home over time.

A screenshot of the Real Time environment accessible through the button with the same name is shown in Fig. 4 identified with number (2). This service is dedicated to give an immediate feeling of what is going on in a monitored house in the specific moment a user is accessing @Office. For this service we have developed a representation based on a map (screen (2) in the figure) using colour to distinguish sensors that are not active (red) from those that are active (green). Additionally using color blinking for few seconds we highlight the current change of state. Furthermore, a small puppet graphically represents the position of the old person according to the tracking algorithm introduced in Sect. 3.

The third environment, the People, offers a different type of service, namely to facilitate communication among the network of people related to a primary user. Specifically, this environment represents a dialogue space to allow the social networking of people who assist the same primary user. The environment allows the different actors, involved in the care of a primary users, to exchange information and opinions so as to maximize the overall care for the old person at home. The People functionality essentially customise the concept of social network to the case of the assistance. Specifically the environment puts available a message board for leaving communication to other secondary users, allowing to send messages to specific secondary user (peer to peer modality). A different functionality set a peer to peer communication channel from secondary user to the primary. It is possible to send messages and/or set reminders to the primary users at home that will be delivered through the Giraff telepresence robot. This functionality is connected to the @Home service (next section). Figure 4 (3) and (4) shows two aspects of this environment: in particular (3) shows a message board while (4) shows a dialogue box to set a message with a question for the primary user.

4.2 The @Home Environment

We decided to use the telepresence robot as a means to provide further services to the primary users in the attempt of studying the enlargement of its use with respect to the telepresence. This part of our work is also in line with what emerged from the feedback gathered during the iterative evaluation session sperfored within the project.

Figure 5 shows the main services provided to primary users through the @Home:

  • Avatar: this functionality preserves the “traditional telepresence” service that the Giraff robot provides. The Giraff application has been indeed embedded within the @Home module so as to maintain the possibility for secondary users to visit the older user’s apartment through the telepresence robot.

  • Messages: an additional environment has been added to allow the primary user to receive messages from secondary users or reminders and recommendations. Messages and reminders are provided in both textual and spoken form. Specifically we have developed a message listener that collects messages coming from the middleware and gathered them with a specific panel that “mimic” the functionality of a mail client or a messenger on a smartphone. In addition to adapting the font size to the user we have decided to integrate an off-the-shelf text-to-speech translator to give the user the possibility of “reading aloud” the messages and re-reading them again and again (most older adult have sight problems).

  • Personal Data: this environment is intended to allow primary users to visualize personal data (e.g., physiological measures) and, more in general, to endow the system with a shared space between the primary user and the secondary users that could foster a discussion on the health status and habits of the old person. The general aim is to improve his/her awareness and also to encourage responsible behaviors for increasing his/her well being. The idea is that a secondary user (e.g., an Health Professional) calls a primary user via the Giraff robot and then uses this environment to discuss about the health related data to both explain them to the assisted person and possibly deepen the understanding of them through questions to the old person.

Fig. 5
figure 5

A representation of the @Home environment

Full size image

It is worth emphasizing that this module has been produced according to the user requirements elicitation and a first run of evaluation in real homes. We are running now a quite intensive evaluation, both within the test sites and with ad hoc evaluation sessions in the lab, in order to better validate the current achievements and to receive feedback for further improvements before the end of the project.

5 Conclusion

This paper describes features of the GiraffPlus project according to the particular perspective that led the project to deliver its intelligent environment in 15 different homes (equally distributed among Italy, Spain and Sweden). We have now 6 homes running since 12 Months and additional 9 homes running since 5 months. To achieve a robust continuous delivery in real homes a stratified effort was required, first guaranteeing a basic cycle that gathers data and distributes them to data storage and end users, then building better and reliable user-oriented services to better serve diversified users of such data intensive technological solution.

The advantages coming from the use of the GiraffPlus system have been analyzed focusing on the possibility given by the system to collect elderly people’s movements, behavior and physiological measurements. A dedicated methodology for the localization and monitoring of elderly has been studied and integrated. Noninvasive wireless solutions that expose binary information regarding the interaction of the user with environment have been exploited and tested in real test sites where the sensors have been deployed together with the installation of the GiraffPlus system.