PLUX REAL-TIME SPORTS EVALUATION

A New Real-time Tool for Sports Evaluation

Jo˜ao Santinha

, Rui Santos

, Joana Sousa

and Hugo Gamboa

1,2

Physics Department, FCT - New University of Lisbon, 2829-516, Caparica, Portugal

PLUX, Wireless Biosignals, S.A., Avenida 5 de Outubro n

70 - 6

, 1050-059 Lisboa, Portugal

Keywords:

Sports, Real-time, Intensity, Evaluation, METs, Crouter, HR, %HR

max

, %HRR, TRIMP.

Abstract:

In this paper we present a new tool for athletes performance evaluation in real-time using two different biosig-

nals (ECG and accelerometry). From the accelerometer signal, the level of activity was extracted based on

a validation protocol, in which metabolic equivalent tasks (METs) values were calculated in real-time from

a body-acceleration signal. METs values were obtained from various lifestyle and sporting activities, and

compared with the results from the reference work (Crouter et al., 2006b) for the same activities. The results

obtained showed correlation with the Crouter model. With the results we can conclude that present tool allows

the assessment of the athlete performance based on ECG and accelerometer signals, being a versatile tool,

which can be used by sports professionals and non-professionals.

1 INTRODUCTION

The need to obtain results in sports and a sustained

evolution of athlete’s physical condition lead to con-

tinuous training assessment.

Athletic performance is evaluated by measuring

speciﬁc variables which provide information about

the athlete physical condition (Morrow Jr et al.,

2010). Providing feedback to the athlete and coach is

a major factor in the improvement of sport skill per-

formance. Nowadays this feedback can provide the

athlete and the couch with a real-time assessment tool

of the athlete’s performance.

1.1 Athlete’s Performance Evaluation

The athlete’s performancecan be assessed through in-

tensity. Intensity is deﬁned as the amount of effort

that the body is exposed during the performance of a

certain task. However, while duration and frequency

are easily measured, the intensity is not. The amount

of effort can be assessed using internal, external loads

or both. Since the same external load applied to dif-

ferent athletes can frequently produce different physi-

ological responses, the importance of assessing inter-

nal load has increased. The intensity assessed using

internal load parameters is usually estimated through

physiological parameters, such as:

• Heart Rate (HR);

• Percentage of maximum Heart Rate (%HR

max

)

(Robergs and Landwehr, 2002);

• Percentage of Heart Rate reserve (%HRR) (Bor-

resen and Lambert, 2009);

• Training Impulse (TRIMP) (Borresen and Lam-

bert, 2009);

• Metabolic Equivalent Task (MET).

MET is utilized to quantify the intensity in terms

of the energy expenditure and deﬁned as the resting

metabolic rate, that is, the amount of oxygen con-

sumed at rest.

METs can be obtained from the magnitude of the

accelerometer signal. Using this magnitude, Counts

are determined and then converted into METs through

a non-linear signal processing algorithm using two re-

gression equations based on the method described by

Crouter et al. (Crouter et al., 2006b). For this oper-

ation, the range of the accelerometer is divided into

levels of 0.001664 g, being each level considered 1

Count. The number of Counts are determined by how

many levels the difference of the magnitude of the ac-

celeration between samples correspond to (Actigraph,

2011).

In the present work, a new tool to assess in real-

time the athlete’s performance is presented. This tool

allows the acquisition, visualization and processing of

biosignals in real-time, providing feedback of physi-

ological parameters both for athletes and coaches, in

389

Santinha J., Santos R., Sousa J. and Gamboa H..

PLUX REAL-TIME SPORTS EVALUATION - A New Real-time Tool for Sports Evaluation.

DOI: 10.5220/0003773103890392

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2012), pages 389-392

ISBN: 978-989-8425-89-8

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

order to evaluate the athlete’s performance.

The rest of the paper is organized as follows: Sec-

tion 2 describes the PLUX Real-Time Sports Evalu-

ation tool; Section 3 describes the validation proce-

dure of METs real-time algorithm; Section 4 details

the main results and Section 5 highlights the main re-

sults and future work.

2 PLUX REAL-TIME SPORTS

EVALUATION

PLUX Real-Time Sports Evaluation (PRTSE) is a tool

to assess athlete’s performance in real-time. This tool

was implemented and built in independent blocks,

which are responsible for different options (visualiza-

tion, processing, saving data). The processing option

features two types of algorithms to compute intensity

in real-time based on Heart Rate (HR-based Inten-

sity) and accelerometer signal (Accelerometer-based

Intensity).

2.1 HR-based Intensity

The HR-based intensity block processes, in real-time,

an ECG signal, in order to calculate the %HR

max

%HRR and TRIMP. For that, from the ECG signal,

R-peaks are detected using a buffer which contains

the acquired data from an ECG sensor. The R-peaks

will be used for calculating the R-R intervals in order

to obtain HR, in beats-per-minute (BPM). With HR,

%HR

max

, %HRR and TRIMP are calculated.

2.1.1 HR-based Intensity Real-time Algorithms

The algorithm used for detecting R-peak is based on

a method developed by Pan and Tompkins (Pan and

Tompkins, 1985).

The heart rate monitors show a relatively stable

HR between measures, by computing a mean R-R

interval to reduce the heart rate variability between

R-peaks (Achten and Jeukendrup, 2003). Therefore,

PRTSE obtains the R-R intervals from a 4 seconds

buffer and the HR and other HR-related parameters

are computed, and shown every 2 seconds.

The algorithm computes the HR from consecutive

R-peaks from equation (1):

) − (R

i+1

)

(1)

with R

being the time instant where beat i was de-

tected and HR

being HR calculated from the i-th R-R

interval, with i ranging from 1 to number of peaks-R

– 1 in the 4 seconds ECG-signal buffer. A mean HR,

mean

, is obtained in each 2 seconds using the HR

obtained from the 4 seconds buffer.

Using HR

mean

, the algorithm will compute the fol-

lowing parameters through equations (2), (3) and (4)

in each 4 seconds from the ECG-signal buffer:

%HR

max

mean

× 100

max

(2)

%HRR =

mean

− HR

rest

max

− HR

rest

× 100 (3)

TRIMP =

mean

− HR

rest

max

− HR

rest

× ∆t ×Y (4)

where the ∆t

is 2 seconds since this value is cal-

culated and accumulated each 2 seconds and Y

is the

lactate proﬁle determined using the athletes gender

and the ∆HR

ratio

2.2 Accelerometer-based Intensity

Accelerometer-based intensity block runs a real-

time algorithm for accelerometer signal processing.

There are several equations available to predict the

METs based on accelerometer-data (Crouter et al.,

2006b; Crouter et al., 2006a). The algorithm of

Accelerometer-based intensity block obtains the Act-

graph’s counts × min

−1

(Actigraph, 2011) and cal-

culates the corresponding METs using the model of

Crouter (Crouter et al., 2006b).

2.2.1 Accelerometer-based Intensity Real-time

Algorithms

For the algorithm developed in the present work, a

real-time band-passﬁlter was implemented based on a

Inﬁnite Impulse response model expressed as follows:

y[n] =



∑

i=0

x[n− i] −

∑

j=1

y[n− j]



(5)

where R is the feedforward ﬁlter order, b

are the

feedforward ﬁlter coefﬁcients, S is the feedback ﬁlter

order, a

are the feedback ﬁlter coefﬁcients, x[n] is the

input signal and y[n] is the output.

The coefﬁcients b

and a

are determined for a

band-pass Butterworth of 4

order with corner fre-

quencies of 0.25 and 2.5 Hz.

This ﬁlter records and uses the last input and out-

put samples, x[n − 1]...x[n − R] e y[n − 1]...y[n − S],

respectively, and update them each new sample ac-

quired.

The input signal, x, corresponds to the unﬁltered

z-axis acceleration signal and the output signal, y, to

the ﬁltered z-axis acceleration signal.

BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing

390

Then, the ﬁltered z-axis acceleration signal is sub-

sampled at 10 Hz and the counts × min

−1

and the

coefﬁcient of variation of counts, c

, are deter-

mined each 10 seconds over a period of one minute.

Using the obtained values of counts× min

−1

and c

the METs value is determined using Crouter’s model

(Crouter et al., 2006b).

The output values presented to the user, in visual-

ization, are the METs. If the user chooses to write

the data from Accelerometer-based intensity block,

the parameters recorded are the counts× min

−1

and

METs.

Although the presented tool assess the performance

of athletes based on the ECG and accelerometer sig-

nals, the lack of tools to evaluate in real-time the ath-

lete’s performance based on ECG parameters doesn’t

allow to compare our cardiovascular results with stan-

dard. However, from preliminar study no missing cy-

cles were veriﬁed. This shows that used algorithm is

a robust one with good accuracy; hence the cardio-

vascular parameters were extracted directly from the

computed R-peaks. Thus, the validation of PRTSE

outcomes is focused on METs procedures, which is

reported by Crouter model, allowing the comparisons

between our results and the literature.

3 METHODS

3.1 Acquisition System and Sensors

To acquire the acceleration signal necessary for this

validation a triaxial accelerometer sensor (ACC), xyz-

PLUX, was used. The acquisition system used was

the bioPLUX clinical (PLUX - Wireless Biosignals,

2007). This system is wireless and is responsible for

the signal’s analog to digital conversion, using a 12

bit ADC, and bluetooth transmition of data to a com-

puter. This system can acquire data at a maximum

sampling rate of 1000 Hz.

Since the accelerometer-based intensity uses the

inferior-superior axis, only the inferior-superior axis

of the accelerometer was connected to the bioPLUX.

3.2 Procedures

To validate the real-time algorithm for METs calcula-

tion based on the model deﬁned by Crouter (Crouter

et al., 2006b) two sets (Set 1 and Set 2) of vari-

ous lifestyle and sporting activities were performed.

These activities were selected based on those used by

Crouter to validate his model.

For the Set 1, each of these activities were re-

peated ﬁve times, except walking up and down the

stairs, walking at an average speed of 4.9 km/h and

6.2 km/h and running at an average speed of 9.5 km/h

that were repeat two times.

Each repetition had a duration of 10 minutes, pro-

ducing a total of 60 METs values per repetition, since

the algorithm determines the METs each 10 seconds.

The mean value and standard deviation of METs

per min were calculated to compare with Crouter

model results for each activity.

After that, modiﬁcations to the algorithm were

made and a new set of activities, Set 2, were per-

formed, this time with only one repetition. The values

obtained from the algorithm were, again, compared

with Crouter model.

4 RESULTS/DISCUSSION

In this section, the obtained outcomes will be present

and discussed based on Crouter study (Crouter et al.,

2006b).

For the activities Lying, Standing and Computer

work, Crouter reported results and the PRTSE results,

from both sets, were all equal to 1.00±0.00 METs.

For the Filling activity, Crouter’s results, PRTSE

ﬁrst set and PRTSE second set obtained METs val-

ues of 1.30±0.67, 2.44± 0.05 and 2.44±0.06, respec-

tively. For the activity of Slow walk, the results were

3.73±0.42, 9.38±0.54 and 5.36±1.31 and for the

Brisk walk, 4.71±0.58, 11.83±1.42 and 7.89±1.27.

For the Ascending/ Descending stairs the results were

6.08±1.29, 6.77±1.69 and 5.72±1.54. Finally, for

Slow run, the results were 7.76±0.96, 91.83±5.81

and 43.58±9.0.

For the PRTSE’s results of the ﬁrst set of activi-

ties, it’s possible to note that, for the ﬁrst three activi-

ties, the results from our tool show equal values when

compared with the results by Crouter (Crouter et al.,

2006b) for the same activities.

In the Filling activity, the PRTSE overestimates

the value of METs, when compared with the results

from Crouter. This might be explained by a different

intensity when performing this activity in the valida-

tion of PRTSE’s METs algorithm. Since a work rate

or intensity it’s not deﬁned for this activity, a different

work rate or intensity might explained the difference

between our results and Crouter’s.

For the Ascending/Descending stairs activity,

PRTSE showed a close value from Crouter’s result

for this activity. The difference between the results

can also be explained by the differences in intensity

of this activity performance.

PLUX REAL-TIME SPORTS EVALUATION - A New Real-time Tool for Sports Evaluation

391

For the Slow, Brisk walk and Slow run, an in-

crease of METs values is veriﬁed. This trend is re-

ported in the literature, since Crouter results show

that there is an increase of the METs values with an

increase of the activity intensity. Despite the same

trend, the obtained results are higher than Crouter’s,

but are correlated in most of the situation.

The overall difference in the results for walking

and running may be due to the ﬁlter applied, pos-

sibly, because it allows higher frequencies than the

Actigraph thus verifying the high levels of METs for

these activities.

Therefore, another set of activities were per-

formed using the METs algorithm with a lower upper

cutoff frequency.

The values of Lying, Standing and Computer work

activities, from the second set of activities, estimated

using the METs algorithm were not signiﬁcantly dif-

ferent from the METs, in (Crouter et al., 2006b).

All other estimated values, using METs algorithm

from PRTSE, overestimate the measured values pre-

sented by Crouter for these activities (Crouter et al.,

2006b) with exception of Ascending/descending

stairs that was underestimated. Although these re-

sults, it’s possible to note that the overall results im-

prove when compared with the results from the ﬁrst

set.

According to these results, we can state that our

hypothesis concerning the ﬁlter was correct and the

METs algorithm will need more adjustments to al-

low the use of the EE-based Intensity monitor from

PRTSE.

5 CONCLUSIONS AND FUTURE

WORK

The PRTSE represents a new tool capable of assessing

athletes performance using different physiologicalpa-

rameters, allowing to improve the athlete evaluation.

Concerning to METs results, in activities with high

level of intensity, higher values than Crouter model

were veriﬁed, showing the need of further adjust-

ments in the ﬁltering, in order to obtain results closer

to Crouter model. This tool can be used both by pro-

fessionals and non-professionals of sports.

In the future work we plan to improve the

EE-based intensity monitor algorithm based on the

Crouter model and create our own METs model using

the 3-axis accelerometer signal for various activities,

trying to ﬁll the gap of the Crouter model in activities

with METs values between 1 and 2.4 METs. Since

this new tool is capable of expansion, we intend to

allow the assessment of the athletes performance us-

ing other parameters, such as the critical velocity and

power, instantaneous acceleration, velocity and dis-

tance.

ACKNOWLEDGMENTS

This work was partially supported by National Strate-

gic Reference Framework (NSRF-QREN) under

project ”LUL”, and Seventh Framework Programme

(FP7) program under project ICT4Depression, whose

support the authors gratefully acknowledge.

REFERENCES

Achten, J. and Jeukendrup, A. (2003). Heart rate moni-

toring: applications and limitations. Sports Medicine,

33(7):517–538.

Actigraph (2011). Actigraph Support - What are counts?

[online] Available at: http://support.theactigraph.com/

faq/counts [Accessed 21 July 2011].

Borresen, J. and Lambert, M. (2009). The Quantiﬁcation of

Training Load, the Training Response and the Effect

on Performance. Sports Medicine, 39(9):779–795.

Crouter, S., Churilla, J., and Bassett, D. (2006a). Estimating

energy expenditure using accelerometers. European

journal of applied physiology, 98(6):601–612.

Crouter, S., Clowers, K., and Bassett, D. (2006b). A

novel method for using accelerometer data to predict

energy expenditure. Journal of applied physiology,

100(4):1324.

Morrow Jr, J., Jackson, A., Disch, J., and Mood, D. (2010).

Measurement and evaluation in human performance.

Human Kinetics Publishers.

Pan, J. and Tompkins, W. (1985). A real-time qrs detection

algorithm. Biomedical Engineering, IEEE Transac-

tions on, (3):230–236.

PLUX - Wireless Biosignals (2007). PLUX - Systems [on-

line] Available at: http://www.plux.info/systems [Ac-

cessed 14 July 2011].

Robergs, R. and Landwehr, R. (2002). The surprising his-

tory of the HRmax= 220-age equation. Journal of Ex-

ercise Physiology Online, 5(2):1–10.

BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing

392