RESIMA: A Smart Multisensor Approach For Indoor AAL

B. Andò1, S. Baglio2, C. O. Lombardo3, V. Marletta4, E. Pergolizzi5, A. Pistorio6

Abstract This paper deals with a smart multi-sensor system to provide assistance to weak users in indoor environments. A wireless sensor network is used to monitor users position and to detect their inertial status and the environment itself. Advanced computational paradigms process information coming from the sensors to build awareness of the user-environment interaction and user-environment contextualiza-tion. Main novelty introduced by the proposed approach is a real time monitoring of user status exploiting a high resolution localization feature. Moreover, the possibility to reconstruct the user-environment relationship allows for the implementation of suit-able actions to assist users and to manage alert situations. Activities described in this work are carried out under the RESIMA project - POR- FESR Sicilia 2007-2013.

1 Introduction

Sensor networks have been deeply investigated for applications in both research and industrial contexts. Environmental monitoring, logistics, Ambient Assistive Liv-

1 Bruno Andò

DIEEI-University of Catania, V.le A. Doria, 6, Catania, Italy, e-mail: bruno.ando@dieei.unict.it 2 Salvatore Baglio

DIEEI-University of Catania, V.le A. Doria, 6, Catania, Italy, e-mail: salvatore.baglio@dieei.unict.it 3 Cristian O. Lombardo

DIEEI-University of Catania, V.le A. Doria, 6, Catania, Italy, e-mail: cristian.lombardo@gmail.com 4 Vincenzo Marletta

DIEEI-University of Catania, V.le A. Doria, 6, Catania, Italy, e-mail: vincen-

zo.marletta@dieei.unict.it 5 Elisa Pergolizzi

DIEEI-University of Catania, V.le A. Doria, 6, Catania, Italy, e-mail: elisa.pergolizzi@dieei.unict.it 6 Antonio Pistorio

DIEEI-University of Catania, V.le A. Doria, 6, Catania, Italy, e-mail: antonio.pistorio@dieei.unict.it

B. Andò, S. Baglio

2

ing are examples of real contexts where the need for acquiring and elaborating distrib-uted information arises.

Cognitive Sensor Networks [1-3] and Ambient Intelligent [4] are enabling technol-ogies promoting new directions in the field of Ambient Assistive Living, Home Au-tomation and Advanced Monitoring [5-8].

In [9] a CANBUS sensor network for user-environment perception is presented which can be suitably applied in cases where the user must be aware of surroundings. The system provides support for elderly or impaired people exploring an indoor envi-ronment. The aim of the proposed cognitive sensor network is to perceive the user within the environment in order to acquire awareness of User-Environment Interaction (UEI) and to provide the user with useful information for a safe and efficient environ-ment exploitation. The system exploits advanced user localization algorithms and par-adigms implementing the UEI functionality.

The WSN discussed in [10] represents an advance as respect to the architecture proposed in [9]. The new sensor platform adopted allows for a distributed monitoring of the environment including the position and the inertial behavior of the user. The main novelty consists in the possibility to build awareness of the User-Environment Contextualization (UEC) which actually defines the user status related to the environ-ment status. Moreover, the use of a WSN conveniently substitutes the CANBUS net-work.

In this paper a novel WSN for application in AAL context is proposed. Each node of the network is represented by a multisensor architecture aimed at the monitoring of environmental status (environmental node) or user inertial status (user module). More-over, the system implements a Ultrasound based trilateration system for the sake of user tracking.

Despite the developed architecture implements both the UEI and UEC functionali-ties, in the following a special focus on the sensing features implemented into the sys-tem and the functionalities of the UEC tool will be provided.

Joining the UEI and UEC awareness represents a novel approach to support weak users during site exploration. As an example, typical application contexts could be the exploitation of educational/cultural/job environments by visually impaired people.

Actually, information coming from the UEC and UEI tools could be automatically exploited to build an optimal information for the user and a list of suggested actions to be used by a system supervisor in order to manage hazards or alert situations.

The continuous awareness of the user-environment relationship (contextualization and interaction) leads to the possibility to implement a real time management of the user needs and safety and it can be claimed as the main advantage of the proposed as-sistive system. The use of low cost hardware, the architecture flexibility, ease of use and implementations represent other fundamental aspects of the system.

RESIMA: A SMART MULTISENSOR APPROACH FOR INDOOR AAL

3

In the following, the User-Environment Contextualization feature of the system de-veloped is addressed, with particular regards to the system architecture and solutions adopted for both the environment monitoring and the user inertial behavior assess-ment.

2 An Overview Of The System Developed

Figure 1 shows a schematization of the system proposed to accomplish indoor AAL activities. The system architecture consists of a wireless sensor network with multi-parametric sensor nodes, a user module, pre-processing algorithms, UEC and UEI par-adigms, a DSS, and dedicated User Interface for users and the system supervisor.

The Wireless Sensor Network exploits RF modules “Iris XM2110” provided by Crossbow, equipped with a ATMega1281 microcontroller unit and interfacing facili-ties. The Wireless Sensor Network is composed by motes spread through the environ-ment and user modules. Environmental nodes, shown in Fig. 2, are equipped with gas sensor (Figaro LPM2610), smoke detectors (FR209) and temperature sensors (LM35).

Figure 1. A schematization of the cognitive Wireless Sensor Network

B. Andò, S. Baglio

4

The user module, shown in Fig. 3, is in charge of estimating heading and inertial status of users through a compensated compass module HMC6343 with a tri-axial in-tegrated accelerometer. The server node is connected to a PC where a LabVIEWTM GUI is used to monitor user/environment status. Network nodes and user module are equipped with ultrasonic transducers adopted to implement user localization.

Pre-processing algorithms use data collected by the WSN for user localization, user inertial status monitoring and environmental status monitoring.

Figure 2. Schematization of the Environmental Node

Figure 3. Schematization of the User Module

RESIMA: A SMART MULTISENSOR APPROACH FOR INDOOR AAL

5

The user localization is performed through trilateration algorithm exploiting dis-tances between the user and the sensor nodes measured by Ultrasound Sensors. The experimental characterization of the system revealed an uncertainty of 2 cm in the es-timation of the user coordinates along the x and the y axis.

UEI algorithm combines the positions of obstacles and services in the environment and the user position in order to evaluate possible interactions and generate candidate messages for the user with an appropriate degree of priority.

User posture estimation and UEC functionalities are described in Sec. 2.1. Events generated by the UEI and UEC tools are managed by a decision-making system that operates autonomously by sending appropriate messages to the user and suggesting possible actions to the supervisor.

User notifications will be delivered through the message center managing the user interface. User skill is a key factor for the decision-making system in order to make the most appropriate choices.

Routing algorithms are also implemented to drive the user on user-demand or in case of dangerous event.

Figure 4. Data flow representation of user posture algorithm

B. Andò, S. Baglio

6

2.1 The UEC tool

The user posture is estimated by exploiting information provided by the inertial sensor embedded in the user module, following direction schematized in Fig.4.

Algorithm for posture estimation classifies 6 different well defined postures: Erect (ER), Bowed (BO), Sitting (SI), Lie Down Prone (LP), Lie Down Supine (LS), Lie Down Lateral (LL). Algorithms individuates also 4 different undetermined condition: Erect/Bowed (ERBO), Erect/Sitting (ERSI), Sitting/Lie Down (SILD) and Bowed/Lie Down (BOLD), these conditions are necessary to avoid fluctuations between 6 pos-tures.

Several tests have been implemented to assess performances of the user posture tool. Three people, in good health, aged over 30 years were recruited to evaluate algo-rithm in a supervised mode. Users have been chosen with different heights and body styles.

Figure 5. Data flow representation of UEC algorithm.

User – Environment Contextualization

Env. statusUser dynamics User posture

INP

UT

Data Fusion &UEC status identification

OU

TP

UT

UE

C a

lgor

ithm

s

UEC Priority0=no warning

2=Env warning4=user warning

ID_uecEvents

UECVerbal Descriptionfor the supervisor

Safe Warning indicator

Candidates messages for the User

RESIMA: A SMART MULTISENSOR APPROACH FOR INDOOR AAL

7

Table 1. Experimental results of the user posture tool

POSTURE MEASURED SENS. SPEC.

ER BO SI LP LS LL ERBO ERSI SILD BOLD

PO

ST

UR

E E

XP

EC

TE

D

ER 289 2 0 0 0 1 7 1 0 0 0.96 0.98

BO 2 273 0 0 0 0 25 0 0 0 0.91 0.99

SI 17 0 234 0 0 0 0 45 0 4 0.78 0.99

LP 0 0 0 225 0 75 0 0 0 0 0.75 1

LS 0 0 0 0 207 48 0 0 0 45 0.69 0.99

LL 8 0 2 0 1 285 0 3 0 1 0.95 0.92

Each experiment was repeated 100 times for each posture to be recognized. Users

have been requested to behave in a natural way. In particular, erected posture was per-formed in static and walking condition. Obtained experimental results are given in Ta-ble 1 in terms of number false and positive recognitions. Sensitivity and specificity, reported in Table.2, have been estimated by the following relationships:

FNTP

TPySensitivit

+= (1)

FPTN

TNySpecificit

+= (2)

where TP = True Positives, FN = False Negatives TN = True Negatives, FP = False Positives.

All postures are in general well recognized and in particular warning conditions like lie down or bowed postures are correctly identified. Moreover, it must be consid-ered that for practical applications a light statistic approach can help to reduce false positive.

The UEC tool exploits data coming from both environment and user inertial behav-ior monitoring to provide the supervisor with awareness of the user status as respect to the environment status and to generate candidates messages for the user with an as-signed priority. Figure 5 gives a general schematization of the UEC tool data flow where each condition (combination of user status and environment status) is associated to an Identification Number, ID_UEC, and its priority, UEC Priority. Typical user sta-tus considered in above described experiments have been used to simulate the UEC tool operation for different environmental status. UEC outputs in terms of candidate user messages and indications for the supervisor are given in Table 2. Measured per-formances of user posture tool, trilateration algorithms and UEC/UEI tools promise good capability of the developed system to provide a reliable form of assistance to weak users.

B. Andò, S. Baglio

8

Table 2. UEC outputs in terms of user messages and indications for the supervisor

References

1. Luca Bixio, Lorenzo Ciardelli, Marina Ottonello, Carlo S. Regazzoni, "Distributed Cognitive Sensor Network Approach for Surveillance Applications," Advanced Video and Signal Based Surveillance, IEEE Conference on, pp. 232-237, 2009 Sixth IEEE International Conference on Advanced Video and Signal Based Surveillance, 2009

2. Hai Ngoc Pham, Yan Zhang, Paal E. Engelstad, Tor Skeie, Frank Eliassen, “Optimal Coopera-tive Spectrum Sensing in Cognitive Sensor Networks”, IWCMC ’09, June 21-24, 2009, Leipzig, Germany

3. K. Shenai and S. Mukhopadhyay, “Cognitive Sensor Networks”, Proc. 26th International Con-ference On Microelectronics (MIEL 2008), NIŠ, Serbia , 11-14 May, 2008

4. Eric J. Pauwels, Albert A. Salah, Romain Tavenard,. “Sensor Networks for Ambient Intelli-gence”, IEEE 9th. Workshop on Multimedia Signal Processing, 2007. MMSP 2007

5. E. Aarts and J. Encarnac¸ao, “The Emergence of Ambient Intelligence” Berlin, Germany: Springer, 2006.

6. E. Aarts, H. Harwig, and M. Schuurmans, “Ambient intelligence, in The Invisible Future”, J. Denning, Ed. New York, NY, USA: McGraw Hill, 2001.

7. M. Pantic, A. Pentland, A. Nijholt, and T. Huang, “Human computing and machine understand-ing of human behavior: A survey”, Lec. Not. in Artificial Intelligence, 4451, pp. 47.71, 2007.

8. B. Andò, S. Baglio, V. Marletta, N. Pitrone, “A Mixed Inertial & RF-ID Orientation Tool For The Visually Impaired”, IEEE SSD09, Djerba - Tunisia, March 2009.

9. B. Andò, S. Baglio, S. La Malfa, V. Marletta, A Sensing Architecture for Mutual User-Environment Awareness Case of Study: A Mobility Aid for the Visually Impaired, IEEE Sens. Jour., vol.11-3, pp. 634-640, 2011.

10. B. Andò, S. Baglio, S. La Malfa, A. Pistorio, C. Trigona, A Smart Wireless Sensor Network for AAL, IEEE M&N Workshop, 2011.