INTELLIGENT CHAIR SENSOR-ACTUATOR

A Novel Sensor Type for Seated Posture Detection and Correction

Rui Lucena

, Cláudia Quaresma

2,3

, Adelaide Jesus

and Pedro Vieira

Departamento de Física, Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa, Caparica, Portugal

CEFITEC, Departamento de Física, Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa

Quinta da Torre P-2829-516, Caparica, Portugal

Departamento de Saúde, Instituto Politécnico de Beja, P-7800-111, Beja, Portugal

Keywords: Sensing chair, Pressure-distribution sensors, Haptic feedback, Posture guidance, Sitting position.

Abstract: In order to build an intelligent chair capable of posture guidance and correction we propose a new

sensor/actuator pressure cell capable of measuring applied pressure and conformation change, which will

allow posture evaluation, guidance and correction. We developed and applied the pressure cells to the seat

pad of an office chair to test if both the cells and their placement were suitable for pressure map

reconstruction. When tested for 10 different postures, the results showed distinguishable pressure maps for

each posture, making the pressure cells suitable for pressure map reconstruction and posture evaluation.

This paper also presents a briefly description of our vision and goals for the intelligent chair project.

1 INTRODUCTION

The evolution of the society significantly reduced

the demands for physical activity. Changes in

transportation, communications, workplace and

entertainment introduced into our population

sedentary behaviours. While some of these

behaviours can be avoided, some activities force us

to spend long periods of time in a sitting position.

The consequences of slouching and poor posture are

well documented and can lead to a number of

detrimental health issues like anatomical

characteristics changes of the spine, problems with

intervertebral discs and joints, back and neck pain,

headaches, fatigue and others (John Schubbe, 2004).

The long term goal of this project is to build an

intelligent chair that effectively corrects and

prevents bad posture adoption in order to minimize

the health issues previously described.

Our main hypothesis is that by increasing

discomfort when a poor posture is adopted, the user

will be encouraged to change his position. The

conformation changes in the chair will be made

through the use of pressure cells that we developed,

which are also responsible for evaluating the posture

through their integrated pressure sensor.

The second hypothesis is that slight changes in

the chair conformation over long periods of time,

may help to evenly distribute the applied pressure on

contact zones, reducing fatigue and discomfort. This

could help preventing the adoption of incorrect

postures over long periods of time due to the need of

pressure relief on compressed tissues.

In this paper, we introduce the pressure cell

concept and its results in differentiating 10 different

postures using only a seat pad with 4 pressure cells.

We also present a global vision of our approach to

an intelligent chair for postural guidance and

correction.

2 RELATED WORK

Over the years, several research groups have studied

postural sensing and classification systems and

applied them in several different areas, such as

irregular behavior detection, emotional evaluation

and biometric authentication. Others have used

haptic feedback responses to control human-

computer interactions. However, few have applied

them to posture guidance and correction and no

studies were found in pressure relief patterns in

seated position. In this section we describe the most

relevant papers on posture detection and correction

in seated position.

333

Lucena R., Quaresma C., Jesus A. and Vieira P..

INTELLIGENT CHAIR SENSOR-ACTUATOR - A Novel Sensor Type for Seated Posture Detection and Correction.

DOI: 10.5220/0003772003330336

In Proceedings of the International Conference on Biomedical Electronics and Devices (BIODEVICES-2012), pages 333-336

ISBN: 978-989-8425-91-1

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

Tan et al (2001) described the use of a chair as a

haptic interface for human-computer interactions.

This work used two Tekscan® sensor sheets, with

42-by-48 sensing units to monitor the pressure

distribution in both the seat pad and the backrest.

The use of pattern recognition technics to develop a

static posture classification algorithm, such as

Principal Component Analyses (PCA), achieved an

overall classification accuracy of 96% and 79% for

familiar and unfamiliar users, respectively.

Other researchers used the same sensing system

configuration to test other classification algorithms.

Mota and Picard (2003) used Neural Networks to

classify 9 static postures in real time, achieving an

overall accuracy of 87.6%. In addition, Hidden

Markov Models were used to associate seated

postures with affective states. The algorithm

achieved an overall performance of 82.3% with

postures sequences coming from known subjects and

76.5% from unknown subjects. Zhu et al. (2003)

tested several classification algorithms to find which

one suits best static posture classification. Between

k-Nearest Neighbor, PCA, Linear Discriminant

Analysis and Sliced Inverse Regression (SIR), the

authors found PCA and SIR comparable in

performance and both outperformed the other

methods tested.

As suggested by Tan et al (1997, pp.57) “a low-

cost and low-resolution pressure sensing system will

then be developed to facilitate the widespread use of

smart chairs”. Mutlu et al (2007) adopted a near

optimal sensor placement approach to drastically

reduce the number of pressure sensors used. The

algorithm down-sampled the sensor data from the

high resolution sensor sheets used in previous

researches (Tan et al 2001; Mota and Picard 2003;

Zhu et al 2003) and determined the near optimal

placement of 19 one-and-a-half-inch-square FSR

(Force Sensitive Resistors) sensors. The system

achieved 78% accuracy with 19 sensors, and 87%

using data from 31 sensors.

More recently, Zheng and Morrell (2010)

developed a system with only 7 FSR and 6

vibrotactile actuators, specifically designed to

posture guidance through haptic feedback. With a

classification algorithm based on the mean squared

error between the pressure measurements and the

reference pressure for each static posture, an overall

accuracy of 86.4% was achieved when

distinguishing among 10 postures. This study has

also successfully shown the effectiveness of haptic

feedback for coaching motor behavior in the form of

seated posture.

3 CHAIR DEVELOPMENT

3.1 Pressure Cells

Figure 1: An inflated pressure cell with a gauge pressure

sensor attached.

The objectives of this project required an interface

that could measure the applied pressure and also

change the chair´s conformation. Taking in

consideration a low cost and commercial available

solution, we developed the pressure cells. This

pressure cells can sense pressure and serve as

actuators, by inflation and deflation, for posture

guidance and correction.

As seen in figure 1, a pressure cell is composed

of a large sealed thermoplastic polyurethane pocket

with 20x19cm of dimensions with a rubber pipe

attached. On the other end of the rubber pipe there is

a piezoelectric gauge pressure sensor to measure the

internal pressure of the cell. All cells have an equal

residual air volume for a baseline pressure

measurement. The gauge pressure sensors used were

the Honeywell 24PC series rated to 15PSI. In order

to compensate the inherent linear differences and

offsets between the sensors, calibration curves were

taken and corrected in software data processing, so

the sensors better match themselves.

3.2 Cells Placement

The use of a low resolution sensor matrix requires a

strategically sensor placement in order to achieve

good performance results. Two main strategies were

identified on previous literature. A pure

mathematical and statistical approach (Mutlu et al.,

2007) and an anatomical approach (Zheng and

Morrell, 2010) which considered the ischial

tuberosities, the thigh region behind the knee, the

lumbar region of the spine and the shoulder blades

as the most important and distinguishable areas of

the body for detecting postures.

BIODEVICES 2012 - International Conference on Biomedical Electronics and Devices

334

Figure 2: Seat pad stripped of padding foam, showing the

individual pressure cells placement. 1. Back Left (BL), 2.

Back right (BR), 3. Front left (FL), 4. Front right (FR).

We concur with the anatomical approach since

those are the areas where most pressure is applied in

the seated position. It is also the approach that uses

the least number of sensors, which is helpful in

reducing the overall cost of the chair. Therefore, the

seat pad was divided into 2-by-2 similar areas, one

area to each ischial tuberosities, and one area to each

thigh region behind the knee. The cells were placed

beneath the chair’s padding foam to maintain the

anatomical cut of the seat pad. The seat pad division

is illustrated in figure 2.

To ensure that the interesting areas are always on

top of the respective pressure cell, we developed

large cells that cover almost the entire seat pad. This

also brings an extra advantage of minimizing the

gaps between cells, which could prove

uncomfortable for the users.

The backrest is planned to follow the same

criteria as the seat pad.

4 EVALUATION AND RESULTS

To evaluate our pressure cells performance and their

placement [Back Left (BL); Back right (BR), Front

left (FL), Front right (FR)] on the seat pad we resort

to a posture list used in previous related papers (Tan

et al. 2001; Mutlu et al. 2007; Ying and Morrell.

2010). Those positions are:

1. Upright;

2. Slouching;

3. Leaning forward;

4. Leaning back;

5. Leaning left;

6. Leaning right;

7. Left leg crossed over right;

8. Right leg crossed over left;

9. Left leg crossed over right and leaning right;

10. Right leg crossed over left and leaning left.

The user started in position 1. and successively

changed to the next position with an interval of

approximately 5 seconds. The data was acquired in

real time with a sampling rate of 20Hz and the

results are displayed in Figure 3.

The best scenario to characterize a posture is to

have a unique pressure distribution to each posture.

Our system was able to produce a different pressure

distribution for each of the 10 postures evaluated,

using only 4 sensible areas in the seat pad, thus

allowing an easy characterization of each posture.

Also, we noticed that every time the user

changed is position the output of the pressure cells

was divided into two zones: a “transient zone” and a

“stable zone”, showed in figure 4.

On changing posture from leaning right to left

leg crossed over right, the pressure on the back left

cell (BL) rapidly increased as expected. Also, the

user spent 2 to 3 seconds in the “transient zone”

before reaching the “stable zone”. This could be

relevant to a real time monitoring system so that

classification algorithms won’t get confused in the

“transient zone”, or even to develop algorithms that

based on the “transient zone” are able to predict

what will the user’s posture be.

5 CONCLUSIONS AND FUTURE

WORK

As a first step for our intelligente chair approach our

pressure cells were able to produce distinguishable

pressure maps for each of the 10 posture tested. We

therefore conclude that the pressure cells conjugated

with the cell placement of this study are suitable for

our intelligent chair project as a pressure mapping

system.

The next step is to control the air pressure and

volume inside the pressure cells, allowing for

changes in the chair conformation and stiffness.

In other applications, hydrogel pads have been

widely used to reduce discomfort in compressed

areas. In a later stage of this project, we plan to use

hydrogel, instead of air, to fill the pressure cells in

an attempt to further increase the chair´s overall

comfort.

Upon conclusion of the chair fabrication,

algorithms will be developed for posture evaluation

and correction. A temporal analysis of a person

sitting behaviour will be needed, in order to study an

3 4

INTELLIGENT CHAIR SENSOR-ACTUATOR - A Novel Sensor Type for Seated Posture Detection and Correction

335

Figure 3: Real time pressure measurements from each cell for the 10 reference postures previously described.

effective pressure relief pattern. At this point,

clinical studies should be made in order to evaluate

the correction models applied and the benefits to the

target population.

ACKNOWLEDGEMENTS

The authors wish to thank Eng. Pedro Duque of

NGNS-is company for consultation on electronics

development and for helping with the pressure cell

fabrication and implementation on the chair.

This project is supported by FEDER at the light

of QREN – Quadro de Referência Estratégico

Nacional, Portugal 2007/2013.

Figure 4: Pressure cell measurement from the back left

cell during posture change from position 6. to position 7.

REFERENCES

John Schubbe, 2004. Good Posture Helps Reduce Back

Pain. [WWW] spine-health. Available from:

<http://www.spine-

health.com/wellness/ergonomics/good-posture-helps-

reduce-back-pain> [Accessed 24 June 2011].

Tan, H. Z., Lu, I., Pentland, A., 1997. The Chair as a

Novel Haptic User Interface. In: PUI’97, Proceedings

of the Workshop on Perceptual User Interfaces. Banff,

Alberta, Canada, 19-21 October 1997.

Tan, H. Z., Slivovsky, L. A., Pentland, A., 2001. A

Sensing Chair Using Pressure Distribution Sensors.

IEEE /ASME Transactions on Mechatronics, 6(3), pp.

261 – 268.

Mota, S., Picard, R. W., 2003. Automated Posture

Analysis for Detecting Learner's Interest Level. In:

CVPRW '03, Conference on, Computer Vision and

Pattern Recognition Workshop. Madison, Wisconsin,

USA, 16-22 June 2003, IEEE

Zhu, M., Martinez, A. M., Tan, H. Z., 2003. Template-

based Recognition of Static Sitting Postures. In:

CVPRW '03, Conference on, Computer Vision and

Pattern Recognition Workshop. Madison, Wisconsin,

USA, 16-22 June 2003, IEEE

Mutlu, B. et al., 2007. Robust, Low-cost, Non-intrusive

Sensing and Recognition of Seated Postures. In: UIST

'07, Proceedings of the 20th annual ACM symposium

on User interface software and technology. Newport,

RI, USA, 07-10 October 2007, New York: ACM.

Zheng, Y. J., Morrell, J. B., 2010. A vibrotactile feedback

approach to posture guidance. In: IEEE, Haptics

Symposium, 2010. Waltham, MA, USA, 25-26 March

2010, IEEE.

BIODEVICES 2012 - International Conference on Biomedical Electronics and Devices

336