Scholarly article on topic 'Evaluation of different gyroscope sensors for smart wheelchair applications'

Evaluation of different gyroscope sensors for smart wheelchair applications Academic research paper on "Medical engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Procedia Engineering
OECD Field of science
Keywords
{Gyroscope / "iPod Touch 4" / iPhones / resolution / "signal processing" / wheelchairs}

Abstract of research paper on Medical engineering, author of scientific article — Julian J.C. Chua, Franz Konstantin Fuss, Aleksandar Subic

Abstract Inertial sensors have been used extensively in recent years for measuring and monitoring performance in many different sports except in wheelchair sports. Mounting an accelerometer directly on a wheelchair frame and determining performance parameters from linear acceleration measurements can provide valuable insight into wheelchair sports performance. However, the processing required for this purpose is tedious. With improvements in the accuracy and measuring range of MEMS gyroscopes, it is now possible to mount a gyroscope on a wheelchair racing wheel and measure speeds close to 30m/s. This paper evaluates and compares angular velocity measurements from a custom built wireless gyroscope sensor, a commercial inertial sensor and an iPod Touch 4 device. With effective filtering, gyroscope sensors provide a suitable tool for performance analysis in wheelchair sports.

Academic research paper on topic "Evaluation of different gyroscope sensors for smart wheelchair applications"

Available online at www.sciencedirect.com

ScienceDirect

Procedia Engineering 13 (2011) 519-524

5th Asia-Pacific Congress on Sports Technology (APCST)

Evaluation of different gyroscope sensors for smart wheelchair applications

Julian J. C. Chua*, Franz Konstantin Fuss, Aleksandar Subic

School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University, Bundoora, VIC 3083, Australia Received 24 March 2011; revised 30 April 2011; accepted 1 May 2011

Abstract

Inertial sensors have been used extensively in recent years for measuring and monitoring performance in many different sports except in wheelchair sports. Mounting an accelerometer directly on a wheelchair frame and determining performance parameters from linear acceleration measurements can provide valuable insight into wheelchair sports performance. However, the processing required for this purpose is tedious. With improvements in the accuracy and measuring range of MEMS gyroscopes, it is now possible to mount a gyroscope on a wheelchair racing wheel and measure speeds close to 30m/s. This paper evaluates and compares angular velocity measurements from a custom built wireless gyroscope sensor, a commercial inertial sensor and an iPod Touch 4 device. With effective filtering, gyroscope sensors provide a suitable tool for performance analysis in wheelchair sports.

© 2011 Published by Elsevier Ltd. Selection and peer-review under responsibility of RMIT University

Keywords: Gyroscope; iPod Touch 4; iPhones; resolution; signal processing; wheelchairs

1. Introduction

Inertial sensors have been applied in various sports for measuring and monitoring performance. They usually incorporate measurements from accelerometers, gyroscopes and GPS readings. Some examples of sports performance measurements using inertial sensors include measuring kinematic parameters of golf swings [1], characterizing swimming strokes [2, 3], measuring kayaking velocity and comparing symmetry between left and right strokes [4] and measuring rotational speed and rotational axis in bowling [5]. An inertial sensor has also been used to measure accelerations and decelerations of a rugby wheelchair during a coasting down experiment [6]. These measurements indicated a slight offset of the

* Corresponding author. Tel.: +61 3 9925 6123; fax: +61 3 9925 6108. E-mail address: julian.chua@student.rmit.edu.au.

1877-7058 © 2011 Published by Elsevier Ltd. doi:10.1016/j.proeng.2011.05.124

data at zero acceleration. This could be due to the positioning of the inertial sensor not being perfectly flat thus causing the acceleration data to be slightly affected by gravity. An uneven floor surface could also have the same effect. The acceleration data was numerically integrated once to obtain velocity, followed by manual correction of offset error from acceleration and verification based on the velocity-time plot. Finally, noise was removed using a Savitzky-Golay filter. Therefore, although an accelerometer can be used to measure instantaneous velocity and distance of a sports wheelchair, extensive data processing is required to ensure that the results are valid. The combined use of GPS with accelerometers and gyroscopes may be a good option to improve the accuracy of kinematic data [7] but it has been shown that accuracy of GPS readings varies due to weather conditions and satellite positions. GPS readings are also not applicable at present for indoor environments (e.g. wheelchair rugby and basketball).

Moss et al [8] and Fuss [9] developed velocimeters and attached them to racing wheelchairs to measure and log velocity data. Although these methods of measuring velocity are highly accurate, the setup is not easily transferable to other wheelchairs and may require particular customization and calibration.

On the other hand, a gyroscope with appropriate measurement range can be easily mounted onto a wheelchair wheel to measure angular velocity, which when multiplied by the wheel's diameter will derive linear velocity. Thereafter, it only needs to be numerically integrated once to derive distance and differentiated once to obtain linear acceleration.

The aim of this study was to: (i) develop a simple wireless gyroscope sensor (WGS) to measure angular velocity of a wheelchair wheel; (ii) evaluate measurements of the WGS, MinimaxX (an inertial sensor) and the iPod Touch 4 device using a known number of rotations and compare the accuracy of the data; and (iii) evaluate the feasibility of using gyroscope sensors for measuring wheelchair performance.

2. Experimental Investigation

2.1. Wireless gyroscope sensor (WGS)

There are two main components of the wireless gyroscope sensor (WGS): a remote unit which is the sensor itself and a base unit which is connected to a laptop computer (Fig. 1). The remote unit consists of a dual axis gyroscope sensor (Model LPY5150AL, STMicroelectronics, Geneva, Switzerland; Range: +/-6000 °/s) a 3V battery and an XBee RF module (Model: Series 1, Digi International, Minnetonka, Minnesota, USA) that uses the IEEE 802.15.4 networking protocol. The base unit is simply another XBee RF module connected to a breakout board with a USB connection. The configuration is similar to the one described in [10].

The XBee RF module was selected because it allows wireless transmission of data up to 100 m outdoors and 30 m indoors [11]. This covers indoor games such as wheelchair rugby and basketball, and outdoor games such as wheelchair tennis and wheelchair racing events on the track if the base unit is placed in the middle of the track. It also allows multiple RF modules to communicate with each other which makes it possible for two remote units (one placed on each wheel of the wheelchair) to transmit data to the base unit at the same time. In this way, performance parameters of the wheelchair travelling on a straight line or turning can be determined. The XBee RF module acts as a microcontroller and is capable of a 10-bit analog to digital conversion of the gyroscope signal. Therefore, there is no need for an additional microcontroller, which makes the circuit design relatively simple.

The base unit can be connected to a USB port on a laptop computer which acts as a serial port. In this research a custom program was developed in Visual Basic to read and store the data transmitted from the remote unit and received by the base unit.

2.2. Method

The WGS and two other mobile devices with gyroscopes were mounted onto the left wheel of a rugby wheelchair. The two other mobile devices were the MinimaxX (Catapult Innovations Pty Ltd, Melbourne, Australia) and the iPod Touch 4 (Apple Inc., Cupertino CA, USA;). The resolution and measurement range of the gyroscope sensors are shown in Table 1. The wheelchair was secured onto a wheelchair ergometer to ensure that the wheel was turning on a fixed axis. A rope was attached to the pushrim of the wheel to facilitate the turning of the wheel. The sensors were turned on when the wheel was stationary. Then starting at low speed, the wheel was turned in an anti-clockwise direction using the rope. The speed of rotation was increased to approximately two revolutions per second then slowed to a complete stop. The location where the rope was attached was used as a marker to indicate the start and stop position of the wheel and they had to be identical to enable accurate manual counting of the number of rotations. Three measurements were obtained for anti-clockwise turns and another three for clockwise turns.

Fig. 1. Wireless Gyroscope Sensor - left: remote unit; right: base unit connected to laptop Table 1. Specification of gyroscope sensors.

Sensor Measurement range (deg/s) Resolution (deg/s) 100 * Resolution / Measurement range

WGS +/- 6000 deg/s 14.6627 deg/s 0.244 %

iPod Touch 4 +/- 2000 deg/s ~ 0.0655 deg/s 0.003275 %

MinimaxX +/-1000 deg/s 0.5 deg/s 0.05 %

2.3. Data processing

All angular velocity data were sampled at 100 Hz (or close to 100 Hz) and converted either from degrees/sec or radiant/sec to revolutions/second (RPS). Due to hardware limitations, the WGS could only sample the angular velocity data at 97 Hz, so for the purpose of comparison, repeated data points had to be added into the WGS data as fillers (3 data points per second). At this point, all the data were considered as raw data. The raw data were then numerically integrated to obtain total number of revolutions (NR) and a percentage error from the actual NR was calculated. Based on the average percentage error, a correction factor was then multiplied to the raw data to adjust the amplitude such that the derived NR is closer to the actual NR and the percentage error is reduced to a minimal. Finally the noise in all three sensor data was removed in MATLAB® (Mathworks, Inc., Natick, MA, USA) using a 2nd order Savitzky-Golay filter with a window width of 25.

2.4. Statistical analysis

The NR derived from angular velocity, were compared with the actual NR. This was done for the three stages of data processing: 1) raw data, 2) after the correction factor was multiplied to the raw data, and 3)

after filtering was done. Standard deviation was also calculated for the derived NR for all three stages of data processing. Angular velocity (RPS) data from all three sensors were compared against each other in the following pairs: WGS Vs MinimaxX, iPod Vs MinimaxX and WGS Vs iPod. The comparison was done by plotting one sensor data against the other and adding a linear fit. The gradient of the linear fit reveals how close the WGS and iPod 4 measurements are to the MinimaxX measurements. Then based on the linear fits, the residual standard deviation was determined for each pair of data (WGS & MinimaxX, iPod & MinimaxX and WGS & iPod).

3. Results

Both the WGS and iPod raw angular velocity data were slightly lower and required a multiplication factor of 1.035 and 1.04 respectively, while the MinimaxX used a multiplication factor of 0.995. Summaries of the derived NR from the three sensors are shown in Tables 2, 3 & 4. The WGS had an average improvement of 2.593 % after applying the filter, the iPod 3.684% and the MinimaxX 0.471 %. A comparison of standard deviation for derived NR, for the three stages of data processing is shown in Table 5. Improvement in standard deviation was more pronounced for the WGS and iPod but not so much for the MinimaxX. Table 6 shows the linear fit values (gradient of plots) and the residual standard deviation values.

Fig. 2. Example plot of angular velocity with respect to time - Top: Raw data; Bottom: Corrected and filtered data

Julian J. C. Chua et al. / Procedia Engineering 13 (2011) 519-524 Table 2. Comparison of NR derived based on the sensors' raw data

Test no. Actual NR WGS NR % Error iPod Touch 4 NR % Error MinimaxX NR % Error

1 35 34.565 -1.242 34.125 -2.501 35.808 2.309

2 35 33.494 -4.303 33.108 -5.404 34.768 -0.662

3 40 38.536 -3.660 38.035 -4.914 39.912 -0.219

4 37 34.262 -7.401 35.006 -5.390 36.480 -1.404

5 40 40.297 0.742 39.057 -2.359 40.601 1.502

6 40 38.184 -4.540 39.043 -2.392 40.532 1.331

Avg. % Error -3.401 -3.827 0.476

Table 3. Comparison of NR after multiplying the correction factor

Test no. Actual NR WGS NR % Error iPod Touch 4 NR % Error MinimaxX NR % Error

1 35 35.775 2.214 35.490 1.399 35.629 1.797

2 35 34.666 -0.953 34.433 -1.621 34.594 -1.159

3 40 39.885 -0.288 39.556 -1.110 39.713 -0.718

4 37 35.461 -4.160 36.406 -1.606 36.298 -1.897

5 40 41.707 4.268 40.619 1.547 40.398 0.995

6 40 39.521 -1.198 40.605 1.512 40.330 0.824

Avg. % Error -0.020 0.020 -0.026

Table 4. Comparison of NR after applying a Savitzky-Golay filter

Test no. Actual NR WGS NR % Error iPod Touch 4 NR % Error MinimaxX NR % Error

1 35 35.773985 2.211 35.77457 2.213 35.629 1.797

2 35 34.665161 -0.957 34.43276 -1.621 34.594 -1.159

3 40 39.882455 -0.294 39.55589 -1.110 39.714 -0.715

4 37 35.461695 -4.158 36.4244 -1.556 36.306 -1.875

5 40 39.695748 -0.761 40.56845 1.421 40.398 0.995

6 40 39.643042 -0.892 40.605 1.513 40.395 0.988

Avg. % Error -0.808 0.143 0.005

Table 5. Comparison of standard deviation of derived NR data for the three stages of data processing

Stage of data processing. WGS iPod MinimaxX

1. Raw data 1.606 1.527 0.521

2. Correction factor applied 1.020 0.557 0.483

3. Filter applied 0.743 0.597 0.490

Table 6. Comparison of linear fit gradient and residual standard deviation after filtering

Test no. WGS - MinimaxX iPod - MinimaxX WGS - iPod

Linear fit grad. Res. Std Dev Linear fit grad. Res. Std Dev Linear fit grad. Res. Std Dev

1 0.9972 0.0158 0.9972 0.0158 1.0000 0.0000

2 0.9982 0.0103 1.0053 0.0141 0.9928 0.0091

3 0.9979 0.0273 1.0085 0.0121 0.9891 0.0297

4 0.9912 0.0264 1.0098 0.0127 0.9812 0.0299

5 0.9897 0.0088 1.0116 0.0030 0.9783 0.0092

6 0.9930 0.0167 1.0162 0.0201 0.9770 0.0087

Avg. 0.9945 0.0176 1.0081 0.0130 0.9864 0.0144

4. Conclusion

The paper presented a comparative analysis of a new wireless gyroscope sensor (WGS), developed in this research to measure angular velocity of a wheelchair wheel, against alternative commercially available devices that can be used for this purpose, such as MinimaxX (an inertial sensor) and the iPod Touch 4 device. The measurements and analysis results obtained in this research using all three devices lead to the following conclusions.

Applying a custom multiplication factor to the raw angular velocity data of all three gyroscope sensors greatly improved the value of derived NR.

The WGS that was developed for measuring angular velocity on a wheelchair wheel is an inexpensive and simple solution to measure linear velocities and possibly turning speeds of a sports wheelchair. The XBee setup would enable close to real-time monitoring of a wheel's angular velocity measurements. However, the accuracy of the data could still be improved with a higher bit A/D converter. This would increase the resolution of the gyroscope signal and the angular velocity measurement. Nevertheless, it has been proven that simply by applying a correction factor and filter to remove noise, the data obtained was comparable to the actual NR. This applied to all three gyroscope sensors and their final filtered matched closely as shown in the paper (see Table 6). Therefore, using gyroscope sensors for measuring wheelchair kinematic data is highly feasible.

The iPod Touch 4 which is similar to the iPhone 4 device was selected for this study because it is considered an ubiquitous device embedded with powerful sensors including a 3D accelerometer and gyroscope. It is highly accessible to athletes and coaches. With a properly designed App (application), it can be a very useful tool for sports performance monitoring.

References

[1] King K, Yoon SW, Perkins NC, Najafi K. Wireless MEMS inertial sensor system for golf swing dynamics. Sensors and actuators 2008; 141: 619-30.

[2] James DA The application of inertial sensors in elite sports monitoring. In: Moritz EF, Haake S, editors. The Engineering of Sport 6 , Munich: Springer; 2006, p. 289-94.

[3] Ohgi Y, Ichikawa H, Miyaji C. Microcomputer-based acceleration sensor device for swimming stroke monitoring. JSME International Journal 2002, 45(4): 960-6.

[4] Janssen I, Sachlikidis A. Validity and reliability of intra-stroke kayak velocity and acceleration using a GPS-based accelerometer. Sports Biomechanics 2010; 9(1): 47-56.

[5] King K, Perkins NC, Churchill H, McGinnis R, Doss R, Hickland R. Bowling ball dynamics revealed by miniature wireless MEMS inertial measurement unit. Sports Engineering 2011; 13: 95-104.

[6] Chua JJC, Fuss FK, Subic A. Non-linear rolling friction of a tyre-caster system: analysis of a rugby wheelchair. Proc. IMechE Part C: J. Mechanical Engineering Science 2011; 225(4): 1015-20.

[7] Brodie M, Walmsley A, Page W. Fusion motion capture: a prototype system using inertial measurement units and GPS for the biomechanical analysis of ski racing. Sports Technology 2008; 1(1): 17-28.

[8] Moss AD, Fowler NE, Tolfrey VL. A telemetry-based velocometer to measure wheelchair velocity. Journal of Biomechanics 2003; 36: 253-7.

[9] Fuss FK. Influence of mass on the speed of wheelchair racing. Sports Engineering 2009;12(1): 41-53.

[10] Ayars E, Lai E. Using Xbee transducers for wireless data collection. American Journal of Physics 2010; 78(7):778-81.

[11] Xbee datasheet. Available: http://www.digi.com/pdf/ds_xbeemultipointmodules.pdf