Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
ATHLETE MONITORING SYSTEM
Document Type and Number:
WIPO Patent Application WO/2015/035473
Kind Code:
A1
Abstract:
The present invention relates to a method for determining a parameter of an athlete. The method involves monitoring acceleration of the athlete using an accelerometer. The monitored acceleration and a model associated with the athlete are used to determine the parameter. Advantageously, the method is improved over known methods as it need not require the continuous use of a GPS when monitoring the athlete's performance. In one embodiment, a GPS is not provided at all.

Inventors:
NEVILLE, Jonathan (C/- Cullens Patent and Trade Mark Attorneys, Level 32239 George Stree, Brisbane Queensland 4000, AU)
JAMES, Daniel (C/- Cullens Patent and Trade Mark Attorneys, Level 32239 George Stree, Brisbane Queensland 4000, AU)
ROWLANDS, David (C/- Cullens Patent and Trade Mark Attorneys, Level 32239 George Stree, Brisbane Queensland 4000, AU)
Application Number:
AU2014/050230
Publication Date:
March 19, 2015
Filing Date:
September 15, 2014
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
GRIFFITH UNIVERSITY (C/- Cullens Patent and Trade Mark Attorneys, Level 32239 George Stree, Brisbane Queensland 4000, AU)
International Classes:
G01C22/00; G01C23/00; G01P15/02; G01P21/00
Foreign References:
US6145389A2000-11-14
US20130085711A12013-04-04
US7057551B12006-06-06
Attorney, Agent or Firm:
CULLENS PATENT AND TRADE MARK ATTORNEYS (Level 32, 239 George StreetBrisbane, Queensland 4000, AU)
Download PDF:
Claims:
The claims defining the invention are as follows:

1. A method for determining a parameter of an athlete, the method involving:

monitoring acceleration of the athlete using an acce!erometer; and

using the monitored acceleration and a model associated with the athlete to determine the parameter.

2. A method as claimed in claim 1 , wherein the model is personalized to the athlete.

3. A method as claimed in claim 1 , further including the step of creating the model.

4. A method as claimed in claim 3, wherein the step of creating involves:

(1 ) recording acceleration from the acceierometer as the athlete performs running trials at constant speeds;

(2) determining the athlete step rates from the recorded acceleration for the corresponding speeds; and

(3) creating the model using the determined step rates and corresponding speeds.

5. A method as claimed in claim 1 , wherein the model includes th variance of the athlete's step frequency with speed.

6. A method as claimed in claim 1 , wherein the model includes linear model, a non-linear model or a quadratic model.

7. A method as claimed in claim 1 , wherein the monitored acceleration is used to determine athlete step frequency which, in turn, is used with the model to determine the parameter.

8. method as claimed in claim 7, wherein the step frequency is determined using a zero crossing step detection process, a fast fourier transform approximation process or a peak detection process.

9. method as claimed in claim 7 wherein, for contact sports, the determination of athlete step frequency involves any one or more of: using a FFT process to determine a magnitude of dominant frequency and/or frequency of dominant frequency from filtered accelerometer data; identifying and excluding frequency outliers; monitoring stride consistency and excluding determined variatio between strides; detecting peaks in filtered accelerometer data; and correlating frequency and time determinations of step frequency.

10. A method as claimed in claim 1 , wherein the monitored acceleration is processed using a sliding window.

11. A method as claimed in claim 1 , wherein the monitored acceleration has its baseline removed.

12. A method as claimed in claim 1 , wherein the monitored acceleration undergoes any one or both of a high pass filter stage or a low pass filter stage.

13. A method as claimed in claim 1 , wherein the parameter is speed or distance travelled.

14. A system for determining a parameter of an athlete, the system including:

an accelerometer for monitoring acceleration of the athlete; and

a processor for processing data associated with the monitored acceleration and a model associated with the athlete to determine the parameter.

15. A system as claimed in claim 14, further including a vest to whic the

accelerometer can be fastened,

16. A system as claimed in claim 15, further including a transmitter for transmitting data from the vest.

17. A system as claimed in claim 16, further including a receiver for receiving the transmitted data.

18. A system as claimed in claim 14, wherein the accelerometer forms part of a mobile phone.

19. A system as claimed in claim 14, further including a global positioning system (GPS) to be used in creating the model.

20. A system as claimed in claim 14, further including any one or more of a low pass filter, a high pass filter, a fastfourier transform (FFT) block, a zero crossing detection block and a peak detection block.

Description:
ATHLETE MONITORING SYSTEM

TECHNICAL FIELD

[0001 J The present invention generally relates to athlete monitoring systems.

BACKGROUND

[0002} The reference to any prior art in this specification is not, and should not be taken as an acknowledgement or any form of suggestion that the prior art forms part of the common general knowledge,

[0003] Professional athletes seek every competitive advantage to win. Athlete monitoring systems in the form of pedometers were traditionall used to log distance travelled, although often prove to be inaccurate.

[0004] In recent times, athletes have been fitted with data loggers including a Global Positioning System (GPS) such as those of US 7715982 and US 8036826 to provide more accurate athlete parameters to athletes and coaches.

[0005] A GPS can be unreliable in some locations such as roofed stadiums, and is somewhat expensive thereby adding to equipment cost and placing equipment beyond the budget of many non-professional athletes. Furthermore, a GPS is comparativel power hungry and thereby drains a battery quickly,

[0006] Embodiments of the present invention provide improved techniques for determining one or more athlete parameters,

SUMMARY OF THE INVENTION

[0007] According to one aspect of the present invention, there is provided a method for determining a parameter of an athlete, the method involving:

monitoring acceleration of the athlete using an accelerometer; and

using the monitored acceleration and model associated with th athlete to determine the parameter. [0008] Advantageously, the method is improved over known methods as it need not require the continuous use of a GPS when monitoring the athlete's performance. In one embodiment, a GPS is not provided at all.

[0009] Preferably, the model is personalized to the athlete. By using the

personalized model, the accuracy of the determined parameter is far better than traditional pedometers and at least as good as data loggers with a GPS.

[00010] The method may further include the step of creating the model. The step of creating may involve: (1 ) recording acceleration from the accelerometer as the athlete performs running trials at constant speeds; (2) determining the athlete step rates from the recorded acceleration for the corresponding speeds; and (3) creating the model using the determined step rates and corresponding speeds.

[00011 ] The model may include the variance of the athlete's step frequency with speed. The model may include a linear model, a non-linear model or a quadratic model.

[00012] The monitored acceleration may be used to determine at lete step frequency which, in turn, is used with the model to determine the parameter. The monitored acceleration may be processed using a sliding window. The monitored acceleration may have its baseline removed. The mortitored acceleration may undergo any one or both of a high pass filter stage or a low pass filter stage. The step frequency may be

determined using a zero crossing step detection process, a fast fourier transform approximation process or a peak detection process.

[00013] For contact sports, the determination of athlete step frequency may involve any one or more of: using a FFT process to determine a magnitude of dominant frequency and/or frequency of dominant frequency from filtered accelerometer data; identifying and excluding frequency outliers; monitoring stride consistency and excluding determined variation between strides; detecting peaks in filtered

accelerometer data; and correlating frequency and time determinations of step frequency.

[00014] The parameter may be speed or distance travelled. [00015] According to another aspect of the present invention, there is provided a system for determining a parameter of an athlete, the system including:

an accelerometer for monitoring acceleration of the athlete; and

a processor for processing data associated with the monitored acceleration and a model associated with the athlete to determine the parameter.

[00016] The system may further include a vest to which the accelerometer can be fastened. The system may include a transmitter for transmitting data from the vest, and a receiver for receiving the transmitted data.

[00017] The accelerometer may form part of a mobile phone.

[00018] Any of the features described herein can be combined in any combination wit any one or more of the other features described herein within the scope of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

[00019] Preferred features, embodiments and variations of the invention may be discerned from the following Detailed Description which provides sufficient information for those skilled in the art to perform the invention, The Detailed Description is not to be regarded as limiting the scope of the preceding Summary of the Invention in any way. The Detailed Description will make reference to a number of drawings as follows:

[00020] Figure 1 is a block diagram of an athlete monitoring system in accordance with an embodiment of the present invention;

[00021] Figure 2 is a flowchart of a model creation process performed with the athlete monitoring system of Figure 1 ;

[00022] Figure 3 is a subsystem of the athlete monitoring system of Figure 1 for creating a model in accordance with the model creation process of Figure 2;

[00023] Figure 4 shows recorded acceleration in the form of raw data input into the subsystem of Figure 3; [00024] Figure 5 shows the acceleration data of Figure 4 after a High Pass Filter stage of the subsystem of Figure 3;

[00025] Figure 6 shows the acceleration data of Figure 5 after a Low Pass Filter stage of the subsystem of Figure 3;

[00026] Figure 7 shows the acceleration data of Figure 6 after a peak detection stage of the subsystem of Figure 3;

[00027] Figure 8a is a linear model (step frequency versus speed) of the athlete monitoring system of Figure 1 ;

[00028] Figure 8b is a quadratic model (step frequency versus speed) of the athlete monitoring system of Figure 1 ;

[00029] Figure 9 shows comparative athlete speeds over time determined using both the accelerometer and GPS of the system of Figure 1 ;

[00030] Figure 10 shows a plot of athletes step rate versus constant speed running sample with false negatives (i.e. step rate - 0);

[00031] Figure 1 1 is a subsystem of the athlete monitoring system of Figure 1 , including the subsystem of Figure 3 with enhanced output processing to determine step rate whilst accommodating for impacts of contact sport;

[00032] Figure 12 is a plot of magnitude versus frequency for each accelerometer axis showing band pass filtering applied to the y and z axes; and

[00033] Figure 13 is a plot showing FFT step rate approximation against the corresponding peak detection step rate.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

[00034] According to an embodiment of the present invention, there is provided an athlete monitoring system 100 shown in Figure 1 for determining a speed (i.e.

parameter) of an athlete. The system 100 includes an athlete unit 102 in radio frequency (RF) communication with a coach's computer 104 with which the athlete's performance can be assessed. [00035] The athlete unit 102 is fastened to a vest worn by the athlete and includes an accelerometer 106 mounted between the T3 and T4 vertebrae for monitoring

acceleration of the athlete. Although not typicaily used when the athlete is competing, the athlete unit 102 further includes a global positioning system (GPS) 108 for use in creating a stored characteristic mode! 110 associated with the athlete.

[00036] The athlete unit 102 further includes a processor 1 12 loaded with software 1 14, in the form of computer readable instructions. The processor 112 processes data associated with the monitored acceleration from the accelerometer 106 and the athlete's model 110 to determine the athlete's instantaneous speed.

[00037] Advantageously, the athlete monitoring system 100 is improved over known athlete monitoring systems as it need not continuously use the GPS 108 when monitoring the athlete's performance. The GPS 108 drains a unit battery 1 16 more quickly when compared with th accelerometer 106 and therefore GPS use should be avoided or at least minimized wherever possible.

[00038] Further, the model 110 is personalized to the athlete. By using the

personalized model 110, the accuracy of the determined speed is far better than traditional pedometers and at least as good as using the GPS 108.

[00039] The system 100 further includes a RF transmitter 118 for transmitting the determined speed from the athlete unit 102, and a RF receiver 120 for receiving the transmitted speed. The coach's computer 104 stores a history 122 of the athlete's speed and other determined parameters such as distance travelled.

[00040] The athlete monitoring system 100 is used in creating the model 110 as well as monitoring the performance of the athlete as indicated below.

[00041] MODEL CREATION

[00042] The model creation process 200 is explained in detail below with reference to Figures 2 and 3.

[00043] At step 202, the processor 112 records in database 10 the acceleration from the accelerometer 106 as the athlete performs a number of running trials. The athlete runs at a constant speed over a predetermined distance (say 20 meters) for a series of trials, with each trial increasing in speed (by say i -2km/haur increments). The running speed is kept constant (±5-10%) for each trial over the distance, and the acceleration is recorded over a period of time (ideall between 3 and 5 seconds).

[00044] The GPS 108 can be used to verify the constant running speed for each speed trial. Alternatively, the any external speed measurement device (e.g. treadmill, radar gun, etc.) can be used to ensure that the athlete is maintaining a constant speed.

[00045] At step 204, the processor 112 implements a subsystem 300 shown in Figure 3 to determine the athlete step rate 302 from the recorded acceleration 304 for each constant speed running trial.

[00046] An example of the recorded acceleration 304 in the form of respective x, y and z Cartesian Coordinate raw data for a given constant speed trial is shown in Figure 4. The raw accelerometer data contains external forces and noise not associated with running.

[00047] The two filter stages 306, 308, and a peak detection of fast fourier transform (FFT) block 310 are used to determine step rate 302.

[00048] The 0.9 Hz High Pass Filter (HPF) stage 308 removes the majority of sensor orientation and gravitational components as shown in the signal of Figure 5. An orientation data signal is produced. Uncompensated gravity can often offset the axes making step detection occasionally challenging. This process isolates the slow changes of sensor often associated with the athlete's orientation, found during accelerating and breaking. As 0.9Hz is significantly slower than the minimum expected step frequency of 2 Hz, forces associated with ground impacts are not filtered out. For regulated running data (on track) this filtering stage 306 can be skipped.

[00049] Although useful for detecting events including accelerating, braking, tackling and other sporting activities, the determined orientation data is subtracted from the accelerometer data for the purposes of detecting steps below. This subtraction removes any effects due to the athlete's running angle that may impact on the zero- crossing algorithm below. [00050] The βΗζ Low Pass Filter (LPF) stage 308 effectively turns the noisy accelerometer data, including signals induced by the vibrations in the skin and vest, into clearly identifiable repetitive signals as shown in Figure 6. Each repetition in the y and z axis indicates an impact with the ground. The x-axis corresponds to left-right body rotations , in turn, coinciding with strides {i.e. one stride = 2 steps = 2 ground impacts). Each peak and trough pair in the x-axis corresponds to a ground impact in the and z axes. As the maximum typical step frequency for running is around 4-5Hz, a low pass filter of 6Hz does not remove the frequency components containing th athlete's impacts with the ground.

[00051] Peak detection can be used to identify peaks in the acceleration data as shown in Figur 7, and the step rate 302 (or frequency) is determined by counting the peaks in the designated time period. Alternatively, A Fast Fourier Transform (FFT) block 310 can be used to determine the step rate 302. A zero crossing step detection block 31 can also be similarly used to determined step rate 302,

[00052] At step 206, the processor 11 creates and stores the mode! 110. The model 1 10 is created by plotting the step rates 302 determined at step 204 against the associated constant speeds and shown in in Figures 8a and 8b.

[00053] Figure 8a shows a linear approach (y∞mx+c) to a mode! 110 for track running showing high correlation (r2=0.9712).

[00054] Figure 8b shows quadratic fit model 1 10 (y=ax 2 +bx+c) for Australian Football League competition and training data at the elite level. A high correlation is observed (r2«0.92614). High model correlation indicates repeatability in measuring athlete step rate across a range of speeds and indicates improved accuracy when implementing the model 10. The equation representing the linear or quadratic models 1 10 can be used during the performance monitoring indicated below. The choice of model 110 is dependent upon the sporting activity being undertaken and can include a non-linear model.

[00055] PERFORMANCE MONITORING

[00056] A method for determining the speed (i.e. parameter) of the athlete is now described. The method involves monitoring acceleration of the athlete using the accelerometer 106. The processor 112 undertakes the same filtering process described above in relation to steps 204 and 206 of Figure 2, and Figure 3, to derive athlete step frequency from input real-time raw accelerometer data 304 as the athlete performs at often varying speed. Preferably, the raw accelerometer data 304 is processed to hav its baseline (mean) removed to account for errors associated with the mounting of the accelerometer integrated circuit on the printed circuit board (PCB), the mountin of the PCB within the device casing, and the attachment of the unit 102 to the athlete.

[00057] A similar time window (3-5 seconds) for calculating real-time step rate 302 is used as above. However, the window is applied to ai! data and is shifted at a

predetermined number of samples. For a 100Hz accelerometer 106, the sliding time window is typically shifted every 20 samples for each step rate determination. However, size can be implemented at the user's discretion.

[00058] The determined step rate 302 is the y component of the linear equation (y-mx+c), or quadratic equation (y~ax 2 +bx+c) keeping only the positive value fo x. In turn, the ste rate 302 can be used to determine the associated athlete speed from the model 110.

[00059] The determined athlete speed is transmitted from the processor 1 12 to the coach's computer 104 where is it stored in the history database 122. The computer 104 can also determine the distance travelled and other performance parameters.

[00060] Figure 9 shows exemplary determined athlete speeds over time with both the accelerometer 106 and GPS 108 showing a high degree of correlation.

[00061 ] ENHANCED PERFORMANCE MONITORING FOR CONTACT SPORT

[00062] The foregoing embodiment was suited to non-contact running sports, with the monitored acceleration having minimal disturbances caused by tackling or a sudden change in the direction of travel associated with contact sports. The monitored acceleration during contact sports includes more disturbances as indicated below.

[00063] External factors relate to conditions of the sport that impact the athlete's natural gait. These factors can include player contact, ground surface, obstructions and ball control skills. Individual factors include events such as a change in a player's objective {change in ball location or creating a play) and intentional change to dodge or mislead opponents.

[00084] False negative values (shown in Figure 10} can be identified that equate to periods of constant speed where no clea ground impacts can be observed post filtering. This effect is most commonl observed when analysing walking data as shock attenuation through the lower limbs and core, reduce ground impact forces making them difficult to detect. It is uncommon however, the same effect can occur during running when significant events impact on athlete gait or add noise to the accelerometer signal.

[00065] Although tare, false positive step detection can occur in one of two ways. One way to produce false positive step detection in the accelerometer data is to create a strong repetitive tmpact force that mimics ground impact forces produced by running. False ste rates can also occur when accelerometer signals are under-filtered and double peaks are observed for some or all of the actual ground impacts. This effect is often seen at high running speeds and will often produce step-rates twice the value expected.

[00066] Figure 1 1 is an optional subsystem 800 of the athlete monitoring system 100 of Figure 1 , including the subsystem of Figure 3 with enhanced output processing to determine step rate whilst accommodating for the foregoing external factors, individual factors, faise negative and false positive detection. Like referenc numerals in Figure 10 refer to like features previously described.

[00067] Step rate selection processes 602, 604, 606 were therefore effectively added to the subsystem 300 of Figure 3. In particular, the FFT selection process 602, a step rate consistency selection process 604 and a time and frequency step rate correlation check 606 were added. The zero crossing process 312 was also substituted for a peak detection process 608 to reduce the likelihood of noise producing false positives.

[00068] FFT processing

[00069] Individual variability in running styles of athletes required post processing in the frequency domain to improve the accuracy of the automated method for extracting step frequency. The extracted FFT data from block 310 is used in two ways to provide accurate step detection. These two methods are the FFT selection process 602 and the time and FFT step rate match 606 and are discussed in further detail below. The FFT process 310 was therefore required to extract two different metrics from the filtered accelerometer data: the magnitude of dominant frequency (signal power at step rate) 610 and the frequency of the dominant frequency (step rate) 612.

[00070] Magnitude of the dominant frequency peak 610 provides information on the distribution of power within the signal being analysed. In this case, provided the dominant frequency observed is in fact the athletes step frequency, the magnitude represents the signal to noise ration of the athlete's impacts with the ground.

[00071 ] The frequency of the dominant peak 612 is directly related to the step rate during over-ground running limited only by the resolution the FFT provides. If the dominant frequency is not the athlete's actual step frequency, the following filters are designed to remove the data from creating athlete speed to step frequency models.

[00072] Both the magnitude of the dominant frequency peak 610 and the frequency of the dominant peak 612 can be obtained by detecting the primary peak in the FFT for each acceleration channel ( y, z) and the stride rate channel (x). This method functions by identifying the maximum in each channel (the dominant frequency magnitude) and the location of the maximum value (FFT step rate approximation).

[00073] Local Peak Step Detection 60S

[00074] Local peak step detection 608 is used in place of the previously discussed zero crossing method. Elite level competition running data is often intertwined with collisions and thus, random spikes in acceleration during running are frequently seen. These noisy influences can produce false positive step detections or incorrectl increase the extracted step rate. Detecting peaks as opposed to zero crossings provide a reduction in the previously mentioned errors. The function of the peak detection algorithm was previously shown in Figure 7. By totaling the number of peaks and dividing by the time between the first and last peak the step rate 614 can be extracted from the time domain accelerometer data.

[00075] FFT selection process 602 [00076] An FFT selection filter 602 was designed to identify outliers both in step frequency 614 and the magnitude of the signal 610. By plotting a large collection of constant speed accelerometer data, processed according to the controlled environment study, in the frequency domain, outliers can be identified and excluded using band pass filtering or simply a ptecewise equation. This plot is shown in Figure 12, where each circle denotes the dominant frequency for a section of running data. The two clusters of dominant peaks have been highlighted (in the y-axis and z-axis) and can be used as limits to automatically filter out noisy data and outliers according to the following criteria:

[00077] (2≤ f ? 1≤4) and (1 j

where

f i is the frequency of the primary peak

|fpi| Is the magnitude of the primary frequency peak

|f p2 | is the magnitude of the secondary frequency peak

[00078] Consistent Step Rate Selection Process 604

[00079] The techniques used in the foregoing frequency domain selection process 602 were designed to automatically remove noisy accelerometer data or accelerometer data where a clear step frequency was not observable. The step frequency consistency check 604 is used to filter out data when the accuracy of the step rate with respect to the GPS calculated speed is potentially compromised. A good example of the desirability for this selection process is when considering running with changes of direction. Although speed may not be lost during change of direction running, the athletes stride pattern may be affected and therefore should be excluded from creating the model. Measuring the time between each step 616 during the period of constant speed provides an automated method for monitoring stride consistency and the data is excluded if the variation between strides is greater than 20% of the mean inter-step time.

[00080] Time and frequency domain step rate correlation 606

[00081] The FFT step rate approximation method 310 is largely resistant to noise and collisions. However, without zero padding, its frequency resolution can be limited. By comparing the FFT approximated step rate (f p i) 612 and the local peak detection step rate 614, errors in step count in the time domain or outliers in the dominant frequency detection can be identified and excluded. Figure 13 shows an example of an athlete's step rate calculated in both the time and frequency domain for n = 867 periods of constant speed across several elite level games. A 1 :1 ratio line has been added along with a ±10% error margins as cut off points. Also, < 2.5Hz and > 4.5 Hz logical filters have been added to remove non-running step rates. The filters are summarized in equation (2) below.

. (ί ρ1 >09 ί ρ1 ).&(ί ρ1 < 1.1χ ί ρ1 )

. (2,5≤t pl <4,5)

[00082] A person skilled in the art will appreciate that many embodiments and variations can be made without departing from the ambit of the present invention.

[00083] The system 100 of Figure 1 included a GPS 108. However, other

embodiments may not include a GPS at all thereby reducing equipment cost and placing equipment within the budget of more non-professional athletes. Furthermore, removal of the GPS will desirabl reduce the size and weight of the athlete unit 102.

[00084] In one embodiment, the raw accelerometer data may be wirelessly transmitted to the computer 104 which instead processes the data to determine athlete speed and distance travelled as described above.

[00085] In another embodiment, the system 100 may be wholly incorporated on a smart phone equipped with an accelerometer 106 and optionally a GPS 108.

Accordingly, the foregoing processing performed by the processor and 1 12 and computer 104 can instead be performed by the phone.

[00086] The present invention may be used by swimmers with the step rate being replaced by a stroke rate.

[00087] In compliance with the statute, the invention has been described in language more or less specific to structural or methodical features, it is to be understood that the invention is not limited to specific features shown or described since the means herein described comprises preferred forms of putting the invention into effect. [00088] Reference throughout this specification to 'one embodiment' or 'an embodiment' means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment ot the present invention. Thus, the appearance of the phrases 'in one embodiment' or 'in an embodiment' in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more combinations.