ieee instrumentation \u0026 measurement magazine smart walker solutions for physical rehabilitation

$: IEEE Instrumentation \u0026 Measurement Magazine Smart Walker Solutions for Physical Rehabilitation$
October 2015 IEEE Instrumentation & Measurement Magazine 21

Smart Walker Solutions for Physical Rehabilitation

Octavian Postolache, José Miguel Dias Pereira, Vítor Viegas, Luisa Pedro, Pedro Silva Girão, Raul Oliveira, and Gabriela Postolache

1094-6969/15/$25.00©2015IEEE

I n the last decade, the clinical reasoning in physical ther-apy has been to develop systems for physiotherapists to make clinical decisions rapidly, effectively and efficiently,

in response to the increasingly complex needs of health and rehabilitation units [1]–[3]. Some studies show the importance of walking aids during rehabilitation from some diseases, and after surgery for arthroplasty in the elderly population [4], and in elderly patients with balance disorders, muscle weak-ness [5] or in people with diabetes mellitus [6]. Walkers are important devices that aid the rehabilitation process. The use of a walker is recommended for gait changes and imbalance due to various factors, such as surgery of the lower limbs or neurodegenerative changes, especially in the early recovery period [7].

Aging, Walkers, and PhysiotherapyTo improve the quality of life of people affected by motor limitations as they age, a challenging task is to develop new architectures like instrumented smart walkers using soft-ware that assist users increase their balance, and diminish falling [8]–[10], which are major causes of morbidity for the el-derly [11]. At the same time, the use of smart equipment with communication capabilities and software associated with elec-tronic health records [12] will allow objective evaluation of physical rehabilitation programs. Various research groups and industries are developing new devices that help users stay bal-anced by giving a wide base of support.

Types of walkers include the standard walker with no wheels, the two wheeled walker, and the four wheeled walker. The choice of one of these types of walkers is related to the user ’s gait limitations and stability requirements. Standard and two wheeled walkers are generally used as primary walking aids. General use during rehabilitation proves to improve confidence and restore or maintain motor ability at the highest possible level. A study in [13] showed that walker assistance does not interfere with rehabilitation

outcome and in some cases may decrease the rehabilitation period. The study was based on subjective observations of walker users.

Physical Rehabilitation Assessment NetworkTo permit continuous monitoring for the physical rehabilita-tion field, walking aids need to have sensors to measure the parameters related to walking, data acquision and process-ing, communication units, and network capabilities for remote assessment of the users. A physical rehabilitation network ar-chitecture, presented in Fig 1., would include a set of smart walker nodes with wireless communication protocols. We con-sidered Wi-Fi, Bluetooth, Bluetooth LE and IEEE802.15.4 for use because of the required communication ranges, data flow, communication security concerns and the need for autonomy.

The Physical Rehabilitation Network we developed uses Wi-Fi mobile nodes and transmits data from home or clinic-based clients to physiotherapists using an Internet connection. As part of client-server architecture, the data coming from cli-ents is stored in a remote database installed in a Physiotherapy Server where it can be accessed by different web or mobile ap-plications associated with remote physiotherapy to evaluate the rehabilitation sessions in an objective manner. To moni-tor the walker’s activity during gait rehabilitation training, the walker’s sensors could measure forces, acceleration, and lower limb motion patterns. Considering the health condition of the users the measurement of cardio-respiratory parame-ters such as heart rate, and saturation of peripheral oxygen (SPO2) may also help the physiotherapist to optimize the reha-bilitation sessions. Research work related to cardio-respiratory assessment through smart walker is reported in [14].

Instrumented Walkers - Smart WalkersThe measurement of quantities such as forces, accelera-tions, velocities, heart rate, and oxygen saturation during

This work was supported by the Instituto de Telecomunicações and Fundação para a Ciência e Tecnologia and the project “EHR-Physio: Electronic Health Records-Needs, Requirements and Barriers in Physiotherapy”, PTD/DTP-DFS/1661/2012.

22 IEEE Instrumentation & Measurement Magazine October 2015

a training session is per-formed through the use of sensors and signal condi-tioning circuits embedded on the walker level. An-a l o g p r o c e s s i n g a n d signaling circuits may pro-vide visual information to physiotherapists re-garding the user’s balance and risk of falling. How-ever, advanced features such as gait rehabilitation metrics calculations, gait pattern recognition, power management, and selec-tive communication with the physiotherapy clients requires added acquisi-tion, digital processing and communication modules that transform the instru-mented walkers into smart walkers. Smart walker prototypes associated with the Physical Rehabilitation As-sessment Network are presented in Fig. 2. The prototypes are characterized by one or two motion sensors (rad, rad1, rad2) that measure the force and acceleration during the gait reha-bilitation session (standard smart walker, 2W smart walker) or during common daily activity (4W smart walker).

Smart Walker SensorsThe sensors embedded in the walker were chosen to extract the relevant information related to the walker’s use during the physiotherapy sessions such as applied forces, accelerations and motion patterns.

Force versus Pressure Sensors The relationship between force and pressure is given by:

(1)

where F represents the applied force, S the surface area and p represents the pressure. The force distribution over a surface is not uniform and the pressure is not constant. There are some common applications where the measurement of pressure distribution over a support surface of a mechanical structure is essential to avoid excessive loading that cause their col-lapse. In the case of biomedical applications, for example, gait parameters from patients undergoing physiotherapy treat-ments can be extracted by having the patient use smart insoles equipped with force or pressure sensors to measure ground reaction forces distribution during locomotion with or with-out aiding walking devices [15]. For walker solutions that can be used for rehabilitation purposes, interface pressure sensing technologies based on load cells, pressure-mapping systems,

and force sensing resistors (FSR) will be the most promising sensing technologies in this field.

Force and Pressure Sensing TechnologiesLoad cells sensing is based in the deformation of a material when it is submitted to a mechanical effort. For small defor-mations, the relationship between stress and applied load is given by:

(2)

where s represents the normal stress, F represents the load and A represents the cross sectional area. Assuming an elas-tic behavior of the material, the normal stress and the axial or transversal strain are given by:

(3)

where eL and eT represent the longitudinal and transversal strains, respectively, v represents the Poisson ratio of the mate-rial, and Y represents its Young modulus.

Usually, resistive strain gauges are used to sense the defor-mation and the signal conditioning circuit that is associated with these kinds of sensors is based on Wheatstone bridges. Since the resistance variations are very low, sometimes on the order of a few tens of mΩ, full configuration of the bridge cir-cuit with four sensors is used to compensate errors caused by temperature variation and to increase measurement sensitiv-ity. The load cells work as force sensors since, by themselves, they can only give information about the total force applied to its bearing structure and cannot give information about the

Fig.1. Physical Rehabilitation Assessment Network based on Wi-Fi smart walkers.


distribution of pressures on a surface. However, it is possible to use multiple load cells to measure forces over multiple in-terface contact regions.

As an example, this capability can be used to measure the unbalance degree of the forces applied on the legs of mobility aiding device [16]. Assuming that Fig. 3 represents the geomet-rical positioning of the forces applied on each leg of a walker device, the centroid coordinates of the set of the forces is given by:

(4)

where xk and yk represent the coordinates of the point on which force Fk is applied.

Thus, the centroid coordinates of the forces that are applied on the walker legs that give essential information about its cor-rect use are given by:

(5)

Another possible sensing solution that can be used for smart walker devices is based on FSR [17]. These sensors include a flexible, thick film material, with piezoresistive characteristics, that is used as a force sensing element. The FSR exhibits in-verse law dependence between the applied force on its sensing area and the resistance between its terminals, where the con-ductance variation is almost linear. The conductance changes of the sensor are proportional to the overall force applied over its sensing area. The sensors’ sensing area is available in a large variety of shapes, and some commercial sensors can be cut according to application requirements. These sensors can be-come customizable embedded sensing solutions, and the main advantages of their use are that they are low-cost, have high sensitivity, are lightweight, have low thickness, and can easily fit to the force interface surface under measurement. However,

relative to load cells, this sensing solution exhibits a lower re-liability and robustness and requires additional care related to mechanical issues and calibration, particularly when multiple units are used in the same measurement system. Self-cali-bration techniques, based on the three wires technique [16] can be used to compensate offset and gain errors, improving measurement system accuracy. For example, on-line self-cali-bration of the set of FSRs attached to the walker legs, can easily be performed by the walker user when there is no load applied on the walker legs (offset error compensation), the lifted con-dition, and when the user’s full weight is uniformly applied on the walker legs (gain error compensation).

Another pressure sensing solution that is commercially available for gait analysis, and that can be potentially used to perform a joint analysis of gait when aided by a walker, is based on pressure measurement arrays [17]. It is important to note that this type of solution is very expensive and, above all, its application is limited to the platform sensing area, since the gait parameters of the person usually deviate from the ones that are exhibited in normal walking environments [18].

Fig. 2. Smart walker prototypes: (a) 4-wheeled smart walker, (b) 2-wheeled smart walker, (c) 2 wheeled smart walker detail, (d) standard smart walker.

Fig. 3. Geometrical positioning of application forces on walker legs (P1- walker legs positioning, L- distance between front and rear legs, W12- distance between front legs, W34- distance between rear legs).


It is a matrix of FSRs whose position maps the distribution of pressure over the global sensing surface. Usually, each of the sensing elements included in this pressure sensing solution is designated a sensel. The spatial resolution of this pressure sensing solution is defined by the number of sensels, and the sensing scanning area can be performed with speeds of sev-eral hundred of Hz that allow obtaining an accurate spatial and temporal distribution of the applied pressures during gait.

Inertial Measurement Units Inertial measurement units (IMU) are electronic devices that contain a 3-axis accelerometer, a 3-axis gyroscope, and as an option, a 3-axis magnetometer. IMUs with an accelerome-ter and gyroscope (only) have six degrees of freedom (6DoF), while those with a magnetometer have nine degrees of free-dom (9DoF). IMUs are widely used in devices that require knowledge of their position, for example an Unmanned Aer-ial Vehicles (UAV), robotic arms, and tools used to study body motion.

IMUs can be installed on walkers to detect motion and to measure parameters like orientation and displacement. This information can be used to detect situations of imminent dan-ger (such as free falls), to know if the walker is being well used (or not), and to see how the user’s gait is progressing over time.

Motion DetectionMotion can be detected using only the information provided by the accelerometer. First, we combine the axial accelerations to get the amplitude of the acceleration vector:

(6)

and then we extract the variance of the acceleration consider-ing a moving time window with some overlap:

(7)

where N represents the number of samples contained in the time window, a(n) represents the nth acceleration sample, and

represents the mean value of all acceleration samples. Exam-ple values for the quantities are: sampling rate = 200 S/s, time window = 0.5 s and overlap = 0.25 s.

General-purpose IMUs are sensitive to gravitational ac-celeration, meaning that a constant value of 1 g (distributed by the xyz components) appears on the acquired samples. Fortunately, (7) cancels this constant component, allowing the IMU to be placed aboard the walker anywhere with any orientation.

The value returned by (7) can be compared against a given threshold to decide if the walker, and its user, is moving or not. Hysteresis can be added to make the decision more insensitive to noise. Motion detection allows the computation of other in-teresting variables such as the total number of movements (steps) over one day, or the percentage of time with motion activity.

OrientationTo accurately know the orientation in space we need to collect information from all three sensors, the accelerometer, gyro-scope, and magnetometer. If the IMU is mounted aboard the walker and aligned according the coordinate system shown in Fig. 4, the challenge is to compute the nautical angles from the sensors data.

Pitch and Roll Extraction from AccelerometerImagine that the IMU is resting with a pitch angle P and that the y-axis is pointing to you as represented in on the left side of Fig. 5. The projection of gravitational acceleration on the x-axis leads to:

(8)

while the projection on the yz plane is:

(9)

Now, maintaining the pitch angle, rotate the IMU so that the x-axis points to you with a roll angle R as represented in on the right side of Fig. 5. The projection of ayz on the yz-axes leads to:

(10)

where atan2 computes the arctangent function in all the four quadrants.

Equations (8) and (10) serve to extract pitch and roll an-gles, respectively. The pitch angle is limited between -90 and +90 degrees, while the roll angle can take the all range of val-ues from -180 to +180 degrees. The equations are valid as long as the IMU is not moving, i.e., as long as gravity is the only ac-celeration present.

Fig. 4. Walker with coordinate system.


Yaw Extraction from MagnetometerThe gravitational acceleration is not sufficient to extract the yaw angle. Additional information has to be provided by the magnetometer. If the magnetometer is sitting on the Earth’s horizontal plane (normal to the earth’s gravity vector), then the roll and pitch angles are zero and the yaw angle is simply given by:

(11)

where mXX and mYY represent the horizontal components of the magnetic field sensed by the magnetometer. If the IMU is tilted, then the pitch and roll angles are not equal to zero (see Fig. 6), and the magnetic sensor measurements (mx, my and mz) need to be projected on the earth’s horizontal plane to obtain mXX and mYY so that (11) can be applied. This procedure is called tilt compensation [20], [21] and is performed as follows:

If the IMU is lifted by the pitch angle P, then the projections of the magnetic field on Earth’s xz-axes are respectively:

(12)

(13)

If the IMU is also turned by the roll angle R, then the projection of the magnetic field on earth’s y-axis is:

(14)

The values given by (12) and (14) are tilt-compensated and can be used safely in equation (11) to extract the yaw angle.

Sensor FusionEquations (8), (10) and (11) are affected have two main prob-lems: first, they assume that the IMU is not moving, and second, they are affected high-frequency noise usually pres-ent in common-use accelerometers. For these reasons, it is advisable to improve them with information provided by the gyroscope. The gyroscope provides samples of angular veloc-ity, which can be integrated to give nautical angles:

(15)

where q represents the roll, pitch or yaw angle, w represents the angular velocity, T represents the sampling period, and n rep-resents the nth sample.

The angles provided by (15) are quite accurate while the IMU is moving but tend to drift when the IMU stops moving (as opposite to the accelerometer). The simplest (but effective) way to fuse information from both sources is to pass the val-ues of the accelerometer plus the magnetometer through a low-pass filter, then pass the values of the gyroscope through a high-pass filter, and finally sum both. The two filters are usu-ally designed with the same cutoff frequency to be mutually complementary:

(16)

where y represents the output of the filter, x1 represents the input from the accelerometer plus the magnetometer, and x2 represents the input from the gyroscope. If we discretize the fil-ter applying backward differences, i.e., by making , we get:

(17)

Fig. 6. Tilted IMU over the horizontal plane.

Fig. 5. Projection of the gravitational acceleration after rotating the IMU over the y-axis and the x-axis. P and R represent the pitch and roll angles, respectively.


which corresponds to the following differences equation:

(18)

where the coefficient belongs to the interval [0; 1[, and represents the cutoff frequency of both filters in rad/s. It should be noted the term between square brackets represents the value given by (15).

Another alternative is to use a Kalman filter to find the statistically optimal estimate of nautical angles. It is a self-correcting filter that adapts its coefficients to minimize the prediction error. However, in many applications (walkers in-cluded), the performance of Kalman filters is similar to that of the complementary filters rarely justifying its use. More infor-mation about Kalman filters can be found in [19] and [20].

DisplacementDisplacement is computed by double-integrating acceleration. The challenge here is that the gravitational acceleration (which is constant) has to be eliminated so that the integration does not diverge. The key is to accurately determine the orientation of the IMU and then subtract the g vector (which points to the earth’s center) from the accelerometer readings.

Given an orientation RPY, the projection of the g vector on the axes of the IMU is given by:

(19)

(20)

(21)

The vector of dynamic acceleration (exclusively related with movement) is obtained by subtracting these values from the accelerometer readings:

(22)

The double-integration of (in all its components) works rea-sonably well as long as the orientation estimate is accurate and the integration time is not long. Displacement information can be used to characterize the walker’s movement in terms of ele-vation and traveled distance.

Motion Sensors To extract information about the walking gait pattern during the rehabilitation program, the use of infrared, ultrasound or microwave radar motion sensors may be considered. In-frared sensing technology assures orthogonal and tangential motion detection but is almost blind to radial motions. It pres-ents large detection comparing radar-sensing technology and it is useful for presence detection. However, the sensors pro-vide no information about velocity, direction of motion, or its range.

In the case of ultrasound sensing technology, accurate range information is provided in close range (<1m). However,

this technology is highly dependent on the environment con-ditions (e.g., noise, fast change in temperature).

Using microwave radar technology for motion sensing presents the advantages of accurate radial motion detection and penetration in non-metallic materials. It can be used to identify the direction of motion and perform range measure-ments. Thus, radar technology has significant advantages over other techniques, and different research work reports the use of radar for low and very low amplitude motions with applica-tion in cardio-respiratory assessment [21], [22].

To characterize a large amplitude motion such as human gait, the microwave Doppler radar represents an appropriate solution as is reported in [21], where the pulse-Doppler range control radar (RCR), which has a carrier frequency of 5.8 GHz, a pulse repetition frequency f 10 MHz and a duty cycle of 40%, is used to characterize the entire body’s motion. The short time Fourier transform (STFT) spectrogram signature associated with slow gait is presented in Fig. 7.

Fig. 7 illustrates the spectrogram elements associated with lower limb motion. However, to evaluate the particu-lar evolution of gait rehabilitation starting from a whole body spectrogram signature represents a challenging task. Deploy-ing sensors for particular sensing of the lower limb motion represents a solution that proves to be appropriate. In [8] and [9], the authors used one microwave Doppler radar mounted on a two-wheeled walker and oriented to sense lower limb mo-tion. The signals acquired for regular, normal, and abnormal walking gait are presented in Fig. 8.

One of the microwave Doppler radar sensors that was used for gait assessment of walker users during a rehabilitation pe-riod is Frequency Modulated Doppler radar (FMCW-rad) [22]. The block diagram of the motion sensor is in Fig. 9. The FMCW-rad employs a transmission antenna and a receiving antenna connected to a low noise amplifier and a set of two mixers M1 and M2. The intermediate frequency signals I and Q are dependent on lower limbs’ reflected waves during the walker use.

Fig. 7. Whole body radar spectrogram signature.


The microwave transmitter connected to T antenna in-cludes an FSK/FMCW-capable K-Band VCO-transceiver controlled through a tuning voltage (Vtune) that provides 24 GHz-24.250 GHz transmit frequency according to the ap-plied Vtune voltage. Several tests were carried out for different Vtune waves including triangular, sawtooth, or a dc voltage (Vtune 5 Vdc).

The used radar sensors present a transmitted power that not exceed 100mW the received radiation by the users during the monitored training being very reduced comparing the mo-bile phone usage. Considering the distance between the TX antenna and the body moving parts of 25cm, this lowers the power density of the sensor by free space attenuation by 48dB, which conducts to levels of received power on the microwatts order level (e.g., 1.6uW). The relation between transmitted (PTX) and received power (PRX) known as the radar equation:

(18)

with G the radar gain, λ the wavelength of transmitted signal, and D the distance between the radar sensor and in-motion body part underlines the fact that the received power is di-rectly proportional with radar cross section on an object (s). The values of s for the whole exposed human body being equivalent with less than 0.5 m2 at 24 GHz comparing with the values obtained for a vehicle arriving that are included in the 1 m2 to 5 m2 according to the angle arrivals. Thus, the cross section associated with body parts in motion during the re-habilitation process will affect the sensitivity of the sensor representing an important challenge in order to obtain reliable results starting from low-level signals that may be amplified using linear programmable gain amplifier schemes.

Taking into account the system autonomy, the radar sensors are characterized by a current supply up to 50 mA in continu-ous operation mode, allowing their use when autonomy for the smart walker is required. At the same time, a power con-sumption optimization algorithm may be implemented taking into account the information delivered by lower consumption sensors embedded in the walker such as the IMUs. These in-formation may be used to automatically connect or disconnect the radar sensor in order to save power and to provide the data delivered by the radar only the lower limb gait measurement procedure is initiated.

The information delivered by the microwave Doppler ra-dar sensors through the intermediate frequency signals I and Q is acquired and processed by microcontroller based plat-form mounted on the walker level. Bluetooth [8] or Wi Fi [23]communication protocols are used to transmit the information to the client level expressed by an Android OS or Windows 8 Tablet for advanced processing and data synchronization with Physiotherapy Server through an Internet connection (Fig. 1).

Triggering Proactive Actions The sensors measurements from FSRs, IMUs, and radar mo-tion sensors trigger proactive actions of warnings to prevent potential falls by a user. Phone calls to attendants when ori-entation has changed, or generation of warning signals to the user when a fall may be imminent is included in the goal of helping the user operate the walker safely.

Force Sensing ResistorsUsing the FSR sensing data, it is possible to correlate walker legs forces with walker handgrips forces, and also to detect if the walker user applies an excessive weight on the walker. For example, a warning condition can result if an excessive force unbalance, between the front and rear legs of the walker, is detected. Thus, it is possible to display, locally or remotely, warnings to signalize dangerous operating conditions of a walker in order to avoid potential falls and injuries of the

Fig. 8. The time evolution of direct intermediate frequency signals delivered by two radars mounted in a walker.

Fig. 9. FMCW block diagram (M1,M2 mixers, microwave coupler, LNA – low noise amplifier, TX ant – transmit antenna, RX-ant – receive antenna, I(t) direct frequency intermediate signal, Q(t) – quadrature frequency intermediate signal).


walker user. As local indicators of warning conditions, solu-tions based on LED or small vibration motors embedded in the walker’s hand grips can be used.

Inertial Measurement Units IMUs can be used to detect motion, to sense orientation and to measure displacements. Motion detection allows the com-putation of other interesting variables such as the number of steps the walker makes over the day, or the percentage of time with motion activity. Orientation is essential to compute dis-placements and can be used to detect situations of imminent danger (such as free falls). Excessive tilt values can trigger a phone call from a remote assistant to check the condition of the walker user and clarify the cause of abnormal readings. Fi-nally, displacement computations can be used to characterize the walker’s step in terms of elevation and traveled distance.

Radar Motion SensorsThe microwave Doppler radar sensors are used to detect motion patterns and to extract metrics that may be used in di-agnosis of abnormal gain. Based on single or multiple radar sensors the gait assessment procedure may be implemented on the level of smart walker the detected motion records being used for objective rehabilitation. The microwave radar sensors may be used to provide information also for incorrect usage

of the walker that is a main cause for the fall events. Thus, the detection of anomalous pattern associated with inappro-priate walker usage can be used to generate warning signals to the walker user but also to the accompanying person or physiotherapists.

Distributed Computing and APPs The general architecture implemented as part of physiother-apy session assessment and objective evaluation of physical rehabilitation that includes distributed computing of the sen-sors data is presented in Fig. 10. The primary processing of the acquired data delivered by the sensors occurs on the level of a microcontroller platform smart walker i. The mobile appli-cation developed for a compatible Android OS is used by the physiotherapist to read the data provided by one of the micro-controller platforms (e.g., Microchip PIC16F73 [8].) Each new user of a smart walker is registered, and their data become part of the patient’s Electronic Health Record. The smart ob-ject is added to the training session, and the physiotherapist will connect via a Bluetooth compatible device to record each training session the values delivered by the force, acceleration, and motion measurement channels. A set of real-time monitor-ing signals in a particular case of a 2-wheeled smart walker is presented on the physiotherapist monitoring display, as Fig. 11 shows.

Fig. 10. Physical Rehabilitation Assessment System architecture.

Fig. 11. Force and acceleration values during physiotherapy visualized on the Physiotherapist APPS GUI, left and right forces values


Based on the information provided by the sensors, individ-ual training sessions can be selected by the physiotherapist to extract information about the evolution of physical capabili-ties of the trained walker’s user.

Conclusions Smart walkers represent an important research field as health-care needs increase with aging. Implementing physical rehabilitation assessment networks based on smart walker nodes that measure force, acceleration and motion allows re-mote monitoring of physical rehabilitation sessions using Internet connectivity and mobile applications. The applica-tions may also allow sensor data storage and analysis based on client–server architecture. The results that were obtained showed clearly that the proposed technologies can give im-portant contributions to improve the quality of life of walker users by monitoring unbalance and instability conditions that can result in falls and harmful injuries.

AcknowledgmentThe authors give special acknowledgment to MSc. Ruben Costa who collaborated with the project team on mobile appli-cations development.

References[1] I. Edwards, M. Jones, J. Carr, A. Braunack-Mayer and G. M.

Jensen. “Clinical reasoning strategies in physical therapy,” Phys.

Ther., vol. 84, pp. 312-330, 2004.

[2] P. M. Masley, C.-L. Havrilko, M. R. Mahnensmith, M. Aubert and

D. U. Jette, “Physical therapist practice in the acute care setting: a

qualitative study,” Phys. Ther., vol. 91, pp. 906-919, 2011.

[3] B. A. Smith, C. J. Fields, and N. Fernandez, “Physical therapists

make accurate and appropriate discharge recommendations for

patients who are acutely ill,” Phys. Ther., vol. 90, pp. 693-703, 2010.

[4] J. A. Singh and D. G. Lewallen, “Predictors of activity limitation

and dependence on walking aids after primary total hip

arthroplasty,” J. Am. Geriatrics Soc., vol. 58, pp. 2387–2393, 2010.

[5] E. Cetin, J. Muzembo, V. Pardessus, F. Puisieux, and A. Thevenon,

“Impact of different types of walking aids on the physiological

energy cost during gait for elderly individuals with several

pathologies and dependent on a technical aid for walking,”

Annals of Physical and Rehabilitation Medicine, vol. 53, pp. 399-405,

2010.

[6] T. R. de Mettelinge, D. Cambier, P. Calders, N. Van Den Noortgate,

and K.Delbaere, “Understanding the relationship between type

2 diabetes mellitus and falls in older adults: a prospective cohort

study,” PLOS ONE, vol. 3, no. 6, e67055, Jun. 2013.

[7] S. Stowe, J. Hopes, and G. Mulley. “Gerotechnology series: 2.

walking aids,” European Geriatric Medicine, vol. 1, pp. 122-127,

2010.

[8] O. Postolache, P. Girão, J. M. Dias Pereira, and G. Postolache;

“Smart walker for pervasive healthcare,” in Proc. Int. Conf. on

Sensing Technology - ICST, Palmerston North, New Zealand, pp.

1-5, Dec. 2011.

[9] G. Postolache, P. M. Girão, and O. Postolache, “Applying

smartphone apps to drive greater patient engagement and

personalized physiotherapy,” in Proc. IEEE Int. Workshop on

Medical Meas. and Applications 2014, Lisbon, Portugal, pp. 1-6, Jun.

2014.

[10] V. S. Probst, “Mechanisms of improvement in exercise capacity

using a rollator in patients with COPD,” Chest, vol.126, no.4,

pp.1102-1107, 2004.

[11] J. M. Housdorf, D. A. Rios, and H. K. Edelber, ”Gait variability

and fall risk in community living older adults: an 11-year

prospective study,” Archives Phys. Med. Rehab., vol 82, no. 8, pp.

1050-1056, 2001.

[12] J. Linder, J. Ma, D. Bates, B. Middleton, and R. Stafford,

“Electronic health record use and the quality of ambulatory care

in US,” [Online]. Available: http://archinte.jamanetwork.com/

article.aspx?articleid=412773&resultClick=1.

[13] L. Vogt, K. Lucki, M. Bach, and W. Banzer, “Rollator use and

functional outcome of geriatric rehabilitation,” J. Rehabilitation

Res. Dev., vol. 47, no. 2, pp. 151-156, 2010.

[14] A. D. C. Chan and J. R. Green, “Smart rollator prototype,” in

Proc. IEEE International Workshop on Medical Measurements and

Applications, 2008. pp. 97-100, 2008.

[15] S. M. Bradley and C. R. Hernandez, “Geriatric assistive devices,”

Am. Fam. Physician, vol. 84, no. 4, pp. 405-411, 2011.

[16] J. M. Dias Pereira, O. Postolache, V. Viegas, and P. S. Girão, “A

low cost measurement system to extract kinematic parameters

from walker devices,” in Proc. Int. Instrum.Meas. Tech. Conf. 2015,

I2MTC 2015, Pisa, Italy, May 2015.

[17] “Force sensing resistors: integration guide and evaluation parts

catalog,” Interlink Electronics, (Sep. 2014), [Online]. Available:

http://www.interlinkelectronics.com.

[18] “WalkwayTM gait analysis system”, Tekscan, (Jun 2015),

[Online]. Available at: https://www.tekscan.com/sites/default/

files/resources/MDL-Walkway-System-Datasheet.pdf.

[19] K. Lauszus, (2012) “A practical approach to Kalman filter and how

to implement it,” [Online]. Available: http://blog.tkjelectronics.

dk/2012/09/a-practical-approach-to-kalman-filter-and-how-to-

implement-it/.

[20] “Kalman filter vs. complementary filter,” Robottini, [Online].

Available: http://robottini.altervista.org/kalman-filter-

vscomplementary-filter.

[21] O. Postolache, P. S. Girão, G. Postolache, and G. Mendes.

“Cardio-respiratory and daily activity monitor based on FMCW

Doppler radar embedded in a wheelchair,” in Proc.33rd Annual

Int. Conf. IEEE Eng. in Medicine and Biology Soc., pp. 1917-1920,

2011.

[22] O. Boric-Lubecke, V. M. Lubecke, I. Mostafanezhad, B.-K. Park, W.

Massagram, and B. Jokanovic,” Doppler radar architectures and

signal processing for heart rate extraction,” Mikrotalasna Revija,

pp. 12-17, Dec. 2009.

[23] T. Yardibi, P. Cudihy, S. Genc, C. Bufi, M. Skubic, M. Rantz, L. Liu,

and C. Phillips II, “Gait characterization via pulse-doppler radar,”

in Proc. Smart Environments to Enhance Healthcare, pp. 662-667,

2011.

[24] O. Postolache, M. D. Pereira, V. Viegas, and P. Girão,”Gait

rehabilitation based on microwave Doppler radars embedded

in walkers”, in Proc. IEEE Int. Workshop on Medical Meas. and

Applications, MeMeA 2015, Torino, Italy, pp. 1-6, 2015.


Octavian Adrian Postolache ([email protected]) (M’99, SM’2006) graduated in Electrical Engineering at the Gh. Asachi Technical University of Iasi, Romania, in 1992 and he received the Ph.D. degree in 1999 from the same university. He is Senior Researcher of Instituto de Teleco-municações and, since 2012, he joined ISCTE-IUL Lisbon where he is currently assistant professor and IT-IUL mem-ber of scientific council. His fields of interests are smart sensors for biomedical and environmental applications, per-vasive sensing and computing, wireless sensor networks, and computational intelligence applied in instrumentation. Dr. Postolache is author and co-author of nine patents, seven books, fourteen book chapters, 65 papers in international journals with peer review, and more than 200 papers in pro-ceedings of international conferences.

J. M. Dias Pereira (M’00–SM’04) received his degrees in Elec-trical Engineering from the Instituto Superior Técnico (IST) of the Technical University of Lisbon (UTL). For almost eight years, he worked for Portugal Telecom in digital switching and transmission systems. In 1992, he returned to teaching as an Invited Professor in Escola Superior de Tecnologia of Instituto Politécnico de Setúbal where he is a Principal Co-ordinator Professor. His main research areas include smart sensing, data sensor fusion, A/D and D/A conversion tech-niques, instrumentation and transducers for industrial applications, and advanced data processing techniques for biomedical and environmental applications.

Vítor Viegas received his degree in Electrical & Computer En-gineering from the Instituto Superior Técnico/Universidade de Lisboa in 1999, followed by the M.Sc. and Ph.D. degrees in 2003 and 2012, respectively. He works as Assistant Professor at the Polytechnic Institute of Setúbal (www.ips.pt), and as a Researcher at the Instituto de Telecomunicações (www.it.pt). His main research activities concern smart transducers, in-dustrial informatics and biomedical instrumentation.

Luísa Maria Reis Pedro obtained the M.Sc. and Ph.D. degrees from the Higher Institute of Applied Psychology and Univer-sity of Porto Faculty: Faculty of Psychology and Educational Sciences in 2003 and 2007. She is an assistant professor at the Institute Polytechnic of Lisbon – College of Health Technol-ogy of Lisbon. Her research interests involve intervention

programmer for the promotion of autonomy, quality of life, and satisfaction with life in individuals with multiple scle-rosis. She has published about 30 articles in national and international proceedings and scientific journals with ref-eree, and she has made about 70 oral and poster presentations in national and international conferences.

Pedro Silva Girão (M’00–SM’01) is a Full Professor of the Department of Electrical Engineering, Instituto Superior Téc-nico, University of Lisbon, and a Senior Researcher, the Head of the Instrumentation and Measurements Group, and the Coordinator of the Basic Sciences and Enabling Technologies of the Instituto de Telecomunicações. His main research in-terests include instrumentation, transducers, measurement techniques, and digital data processing, particularly for bio-medical and environmental applications. Metrology, quality, and electromagnetic compatibility are also areas of regular activity. Dr. Girão is a Senior Member of the IEEE.

Raul Oliveira received his M.Sc. in Child Development and Ph.D. in the Motricity Sciences. He received both degrees from the Faculdade de Motricidade Humana, Universiadde de Lisboa in 1999 and 2010 where he is currently an assistant professor. He is a researcher for the Laboratory of Motor Be-havior and an expert reviewer for the Swiss National Science Foundation (SNSF) as reviewer of important national and international scientific journals. His research interests are re-lated to neuromuscular adaptations to training, sports, and physical therapy. He is the author or co-author of nine books and articles published in national and international physio-therapy science journals and he is member of the project team “Electronic Health Record for Physiotherapy.”

Gabriela Postolache is researcher at Unidade de Fisiolo-gia e Clinica Translacional, Instituto de Medicina Molecular, Universidade de Lisboa. She obtained a Ph.D. degree in Life Sciences in 2006. Her research interests include elec-trophysiology, biomedical instruments, non-invasive and unobtrusive measurements of body signals. She was in-volved in research projects related with autonomic nervous system assessment and monitoring, cardiac arrhythmia, elec-tronic health records for wheelchair users, electronic health records for physiotherapy and she is author and co-author of several patents.

ieee instrumentation \u0026 measurement magazine smart walker solutions for physical rehabilitation

Documents