A COST-EFFECTIVE INDOOR VIBROTACTILE NAVIGATION

SYSTEM FOR THE BLIND

Marco Altini, Elisabetta Farella, Marco Pirini and Luca Benini

DEIS – Department of Electronics, Computer Sciences and Systems, University of Bologna, Bologna, Italy

Keywords: Localization, Navigation, EEG, Bluetooth, Vibrotactile.

Abstract: This paper describes the development of an indoor vibrotactile navigation system for the visually impaired

people. We aimed at realizing a wearable, low-cost, and effective system able to help blind users in

unknown indoor environments that they might visit occasionally, such as hospitals, airports, museums, etc.

The designed system implements a Bluetooth (BT) localization service, and provides wayfinding cues to the

user by means of a wearable device equipped with five motors. The last part of our work describes early

results obtained by the use of electroencephalographic (EEG) analysis to evaluate the vibrotactile feedback.

1 INTRODUCTION

Navigation, in general, involves providing

directional information to the user during a travel

with respect to the intended route and the desired

destination. People with visual deficits can be

challenged by planning and performing navigation

because of the reduced amount of information that

they can perceive from the surrounding. Regarding

navigation and wayfinding systems, the use of

vibrotactile displays has several advantages over

vocal messages. First of all they are less intrusive.

Although tactile and vibrotactile displays can often

offer a limited amount of information, they are less

invasive and do not risk to distract the blind person

from unexpected events, or more important tasks,

such as orientation. Since there is no standard

system for indoor localization (compared to the well

known Global Positioning System used in outdoor

environments), many technologies can be employed.

Among the possible choices BT is a cost-effective,

widespread and standard technology. In this context,

this work presents an indoor navigation system

based on vibro-tactile feedback and BT localization.

Wayfinding cues are provided to the user by use of

coded vibro-tactile messages, occurring in the case

the user should change its direction of movement.

Employing high sensitivity areas, such as shoulders

and the stern, we can reach perfect stimulus

detection while maintaining power consumption

level low and adequate to the expected lifetime of

the system. The last phase addressed in this work is

an early validation of the system. We first performed

traditional assessment by use of questionnaires on a

group of subjects. However, we are also interested in

determining the effectiveness both of the vibrotactile

actuator and of the particular kind of feedback

provided, without involving the subjective

perception of the user. Therefore, we explored use of

EEG to analyze and quantitatively assess effects of

the vibrotactile stimulation. Even if we describe very

early tests and methodology, the use of EEG

analysis seems promising and can be further used to

compare different kinds of vibrotactile actuators.

The paper is organized as follows. An overview of

the existing solutions for indoor and vibrotactile

navigation is given in Section 2. In Section 3 the

system architecture is explained. Section 4 describes

briefly the localization system while Section 5

concerns the vibrotactile system. Experimental

results are given in Section 6 whereas Section 7

deals with preliminary studies on EEG analysis.

Conclusions can be found in Section 8.

2 RELATED WORK

Several navigation systems have been proposed over

the last years, each of them employs one or more of

the following modality to provide information to the

user: video, audio or tactile. Targeting the

development of an aid for visually impaired people

the focus is narrowed to audio and tactile feedback.

Wayfinding is aimed at helping the blind in reaching

a destination within a building. The investigation of

how to support this task has not yet standard

477

Altini M., Farella E., Pirini M. and Benini L..

A COST-EFFECTIVE INDOOR VIBROTACTILE NAVIGATION SYSTEM FOR THE BLIND.

DOI: 10.5220/0003127504770481

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2011), pages 477-481

ISBN: 978-989-8425-34-8

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

solutions in indoor environments. Technologies for

indoor navigation still require to be tuned to achieve

satisfying results. We focused our effort in

addressing the problem of supporting navigation in

indoor spaces, having as target scenarios non-

familiar public places where visually impaired

people can require a stronger support (e.g. to be

accompanied by someone). Navigation systems are

moving from auditory (Crandall, 2001) to tactile

interfaces for providing information to the user, this

is due to multiple factors: first of all a tactile display

is less intrusive. Secondly, progress in electronics is

making it possible to realize wearable lightweight

and low power systems, well suited for vibrotactile

systems. Many devices have been proposed for

outdoor navigation based on tactile interfaces: most

of them are belts (Tsukada, 2004), others are instead

wrist based interfaces (Bujnowski, 2008). Only a

few systems address the problem of indoor

navigation (Ross, 2004), (Ghiani, 2008) mainly by

means of vocal messages and expensive

architectures. As a matter of fact, one of the issues

that arise in indoor environments is how to localize

the user in a building. High accuracy is required to

help a blind person in wayfinding tasks, thus the

current state of the art employs RFID tags (Ghiani,

2008) or ultrasounds (Ross, 2004). Those methods

require a big amount of tags or base stations in order

to determine the user position, resulting in quite

expensive systems. For this reason, they are often

based on the proximity approach, providing

information to the user only when he is close to a tag

or a base station, implementing indeed obstacle

avoidance instead of navigation. Radio frequency

technology, such as BT, has not been adopted for

navigation purposes because of the low resolution

that usually characterize it. Employing a localization

system based on BT would result in a cost-effective

infrastructure, since many building are equipped

with PCs with BT connectivity. Our novel approach

in this direction showed promising results (0.5

meters of accuracy), as pointed out in (Altini, 2010).

Navigation systems are usually validated in three

ways: questionnaires, time employed for completing

a route, difference between the optimal path and the

one taken by the user (Tsukada, 2004), (Heuten,

2008). Hence, other metrics to compare different

kind of actuators or even to assess quantitatively the

performance given by such systems are still needed.

A small step in this direction was taken in

(Bujnowski, 2008); the authors show the activation

of the sensorimotor area of the brain during

vibrotactile stimulation. They demonstrated that the

tactile stimulation was actually influencing the user

perception, even if it was not possible to quantify or

characterize such influence.

3 SYSTEM OVERVIEW

The navigation process, often called wayfinding, is

usually based on three steps. The first step is user

localization, the position of the user in relation to

some known landmarks in the building needs to be

determined. The second step regards choosing the

correct route. This phase usually requires some kind

of mental elaboration by the user, which is aware of

the current position and of the destination. The third

step is keeping the user on the right track. This is the

most challenging step. To overcome the challenge of

keeping effectively the user on the route towards the

destination selected, we developed the system

supported by the architecture described in Fig. 1.

The system is composed of four main components;

the localization engine, a compass module, the

haptic node (vibrotactile actuator), and the

navigation engine. The navigation engine is the core

of the system, receiving input from all other building

blocks. It runs on a smart processing unit (in our

preliminary setup, a netbook). The navigation engine

coordinates input from the localization system and

the compass module to control the vibrotactile

actuation, in an effective close loop. For a blind

person to be completely autonomous during a visit,

we included a vocal component. The vocal

component takes care of helping the user in deciding

which destination to reach and notifies the user

when he is arrived at the selected destination.

Connecting the magnetometer module to a

microcontroller we can easily provide the user

orientation to the main application running on the

netbook by means of a serial interface. The netbook

is carried by the user and uses the information

provided by the compass module, along with the

RSSI values retrieved from the base station nodes, to

determine the position of the user. Once the user has

been localized the route planner has all the

information that it needs for guiding him. Given the

location and the orientation it can send the proper

command to the vibrotactile actuator. The

vibrotactile actuator is a small board composed of a

microcontroller and five motors (see Fig. 2).

The whole system can supply the blind person with

assistance while walking a route in a building using

vibrations to point out what decision to make. Fig. 3

shows a map of the building in which the system has

been tested. The circles in the picture represent the

positions that the localization system can recognize

HEALTHINF 2011 - International Conference on Health Informatics

478

with high accuracy. Fig. 3 shows also the location of

the base stations that are employed for the

localization of the user. A total of five base stations

are used by our architecture.

Figure 1: System architecture.

Figure 2: The Haptic Node, our vibrotactile actuator.

4 THE LOCALIZATION SYSTEM

One of the main building blocks of the navigation

system is the localization engine. Many different

methods have been proposed for indoor localization

and navigation. Among others, localization

algorithms based on BT technology have the

advantage to work on top of the most widespread

wireless communication standard. Many BT based

localization and positioning systems are based on the

use of RSSI (Received Signal Strength Indicator) to

determine the user location. Unfortunately, the

shortcomings that affect this parameter are manifold,

mainly due to propagation effects. Thus, it is almost

impossible to obtain accurate location services using

standard techniques such as triangulation from three

or more BT base stations. RSSI based localization

systems are typically affected by low accuracy due to

variability of the signal strength in presence of

obstacles between the base stations and the user that

is carrying the system. As a consequence, the

differences in RSSI values often depend also on user

orientation, which determines different degree of

power absorption by user body. Thus, we introduced

a multiple neural networks architecture that can

handle changes in RSSI values due to user

orientation. We demonstrated with experimental

results that first training and then activating neural

networks tailored on the user orientation (determined

by the compass module), high definition accuracy is

achievable, allowing indoor navigation with a cost-

effective BT architecture. The system can provide

position estimate with 0.5 meters of accuracy during

a walk. Details on the system can be found in (Altini,

2010).

Figure 3: Map of the building where the system was

tested.

5 VIBROTACTILE ACTUATION

The use of haptic interfaces to provide information

non-visually has been widely investigated. The

sensitivity of different parts of the body to

vibrotactile stimulation has also been studied in

depth. Vibrotactile displays are based on low-cost

and low-power motors, such as page motors

available in every mobile phone nowadays. Many

studies on how to convey complex information with

this kind of motors have been carried out, basically

tuning parameters such as frequency, amplitude,

rhythm and location on the body. One of the main

goals of tactile displays must be providing

directional information intuitively, in the easiest way

possible. We realized a wearable system, composed

of small sized parts, such as the 312-103 motors by

Precision Microdrivers (see Fig. 2). The frequency

of resonance of these motors is close to 200 Hz,

which is the maximum sensitivity of the human skin.

Our system is a wireless device composed of five

motors, a microcontroller (MSP430) and a BT

module. The MSP430 is a well known low power

microcontroller, in our application it is constantly

kept in Low Power Mode, since the actuator will be

off for the most of the time (e.g. the user reached the

destination and will go back to the entrance of the

building after a few hours). In case the actuator has

to provide vibrotactile feedback activating the

motors the microcontroller is woken up by an

interrupt followed by a code indicating the type of

vibro-message to deliver. Four of the motors are

placed on the shoulders, while the last one on the

chest (see Fig. 4), those areas have been proved

effective for this kind of application in other studies

(Toney, 2003). According to van Erp (van Erp,

A COST-EFFECTIVE INDOOR VIBROTACTILE NAVIGATION SYSTEM FOR THE BLIND

479

2005) the resolution on the torso is about 1 cm,

nevertheless increasing the number of motors would

bring easily to higher percentage of error during the

recognition task. Thus we decided to employ only

five motors and to activate them as follows:

- Go forward: motor on the chest is activated

- Turn left of 90°: both the motors on the left

shoulder are activated simultaneously

- Turn right of 90°: both the motors on the right

shoulder are activate simultaneously

Figure 4: Position of the motors on the body.

Each command is provided by a double vibration

of the motors involved (500 ms long). The indication

of the direction to follow provided with motors

vibrating in the corresponding location on the body

is very intuitive and the user does not need any

previous training to be able to use the device.

6 EXPERIMENTAL RESULTS

All the different parts of the system need to be

tested. The system performance was measured in

three different modalities on sighted users; time

needed to complete a route, deviations between the

correct path and the route employed by the user, and

usability tests. By means of those tests we can

determine how accurate the system is in a navigation

task. Every route within the building in which we

tested the system was composed of three basic paths.

During the tests the system randomly chose five

consecutive paths in order to build a more

complicated route, necessary to effectively test the

system. In this way every route was approximately

75 meters long (the main corridor that can be seen in

Fig. 3 is 18 meters long). During the tests both the

time and the route taken by the users were logged.

Preliminary tests were run on three users, all men,

between 25 and 27 years old. All participants had no

mental or physical impairments. They had no

previous knowledge about the routes. Each of them

received a brief introduction on the system.

The main limitations shown by those tests are

due to the localization system. In fact for navigation

purposes the localization system needs to be reliable

and thus the user has to walk at a slow pace, in our

case 0.5 km/h. Although the speed is quite low it

might be sufficient for blind users, especially in

indoor environments. Finally, an adaptation of the

USE questionnaire (Lund, 2008) was filled in by the

users. The questionnaire is subdivided into three

sections: ease of use, ease of learning, and

satisfaction. It emerged that system is easy to use,

user friendly and its use is effortless. It is easy both

to learn and to remember how to use it. Moreover,

the system is not really flexible due to its prototypal

nature.

7 EEG ANALYSIS

Although many researchers developed different

kinds of vibrotactile systems there is no common

methodology for comparing them, making

challenging to establish which one would fit best a

given application. Typically, the evaluation of

actuation systems is carried out by means of

questionnaires and therefore based on user

subjective perception. For those reasons we tried to

explore a new approach to evaluate the efficacy of

the vibro-tactile feedback by acquiring information

on the user perception in a quantitative and

measurable way. This is pursued by means of EEG

analysis. The basic idea behind our experiment is

based on the study of Event Related Potentials

(ERPs), in particular the P300. An ERP (Neuper,

2006) is an electrical potential recorded from the

nervous system following presentation of a stimulus

and it can be directly associated with perception.

ERPs do not depend only on the processing of the

physical stimulus itself, but are believed to be

caused by higher processes, that might involve

memory, expectation or attention. The signal is

typically measured in the parietal area, and it is

evoked delivering a stimulus in one of the sensory

modalities. The most common procedure is called

oddball paradigm. A target stimulus is presented

amongst more frequent standard background stimuli.

In addition, a distracter stimulus is often used to

make sure that the EEG response to the target

stimulus is not only due to the change from the

background pattern but it is a real cognitive function.

One subject took part in this preliminary study. The

experiment was structured as follows: the subject

(26 years old, male, no impairments) was wearing

the vibrotactile actuator (the haptic node). He was

instructed to pay attention to one typology of

stimulus, the target, counting the number of

stimulations, and to ignore the other stimuli. Each

one of the stimuli used for the experiment (both

target and non-target) was one of the vibrotactile

HEALTHINF 2011 - International Conference on Health Informatics

480

messages defined in section 5. A third stimulus was

introduced as a distraction. Each experiment was

carried out in different modalities, a total of six tests

were run (oddball paradigm with two stimulations,

target on the right (1) and left shoulder (2), with eyes

pen (3-4) and closed (5-6). The inter-stimuli interval

was between 2 and 3 seconds, for each trial the

subject received more than 200 stimulations. 19

channels of earlinked referential EEG data were

recorded positioning the electrodes according to the

standard 10-20 system. Data was filtered and

averaged. Independent Component Analysis was

used for artifact removal (eye blinking). After this

phase the data was ready for ERPs analysis. The

P300 elicited by the haptic node showed low

latency. The result of the experiment is shown in

Fig. 5. The P300 is elicited by the rare non-target

stimulus (marked as “center” in the picture), as well

as by the rare target stimulus, but the P300

associated to the target stimulus, in this case on the

left shoulder, is faster. This can be associated with

the cognitive process that follows the recognition of

the target stimulus.

Figure 5: P300 elicited by the haptic node.

In our point of view those P300 evoked potential can

possibly be used as the “fingerprint” of a given

vibrotactile actuator, and along with other

information such as the typology of motors adopted

and the vibrotactile messages conveyed to the user

can serve as a parameter for characterizing different

kind of vibrotactile actuator systems, making it

possible to compare them and to choose the most

appropriate for a given application.

8 CONCLUSIONS

In this paper we presented an indoor navigation

system for the visually impaired people that use

vibrotactile messages to provide directional

information to the users. The system is low cost and

low power, employing off-the shelf motors, a few

BT base stations, a compass module and common

office devices we could obtain encouraging results

during preliminary studies on non impaired users.

Additional tests on blind users shall be carried out in

the near future. Our system can enhance the

autonomous mobility of individuals with visual

losses within a building, providing directional

information intuitively. The last section of the paper

introduced a methodology to evaluate a vibrotactile

system and to compare it against other solutions, by

means of EEG analysis. In this context we analyzed

the P300 evoked potential after vibrotactile

stimulation in different conditions.

REFERENCES

Crandall, W., Bentzen, B. L., Meyers, L. & Brabyn, J.

“New orientation and accessibility option for persons

with visual impairment: Transportation applications

for remote infrared audible signale”, Clinical and

Experimental Optometri, 2001.

Tsukada, K., and Yasumura, M., “Activebelt: Belt-type

wearable tactile display for directional navigation,” in

Lecture Notes in Computer Science. Springer-Verlag

GmbH, Oct 2004, vol. 3205, pp. 384 – 399.

Bujnowski, A., Drozd, M., Kowalik, R., Wtorek, J, “A

tactile system for informing the blind on direction of a

walk,” Human System Interactions, 2008 Conference

on , vol., no., pp.893-897, 2008.

Ross, D., “Cyber Crumbs for Successful Aging with

Vision Loss,” IEEE Pervasive Computing, vol. 3, no.

2, pp. 30-35, Apr.-June 2004.

Ghiani, G., Leporini, B., Paternò, F., Santoro, C.,

“Exploiting RFIDs and tilt-based interaction for

mobile museum guides accessible to vision-impaired

users,” ICCHP 2008.

Altini, M., Brunelli, D., Farella, E., Benini, L., “Bluetooth

Indoor Localization with Multiple Neural Networks,”

Proceedings of the 5

International Symposium on

Wireless Pervasive Computung, 2010.

Heuten, W., Henze, N., Boll, S., Pielot, M., “Tactile

wayfinder: a non-visual support system for

wayfinding,” NordiCHI '08: Proceedings of the 5th

Nordic conference on Human-computer interaction,

page 172-181. ACM, 2008.

Toney, A., Dunne, L., Thomas, B., Ashdown, S., "A

shoulder pad insert vibrotactile display," iswc, pp.35,

Seventh IEEE International Symposium on Wearable

Computers, 2003.

van Erp, J. B., “Vibrotactile spatial acuity on the torso:

effects of location and timing parameters,” In

Proceedings of the First Joint Eurohaptics Conference

and Symposium on Haptic interfaces For Virtual

Environment and Teleoperator Systems, 2005

Lund, A., “Measuring usability with the USE

questionnaire”, 2008.

Neuper, C., Klimesch, W., “Event-Related dynamics of

brain oscillations,” Volume 159.

A COST-EFFECTIVE INDOOR VIBROTACTILE NAVIGATION SYSTEM FOR THE BLIND

481