GB2565567A

GB2565567A - System for analysing efficiency of motion

Info

Publication number: GB2565567A
Application number: GB1713142.6A
Authority: GB
Inventors: Howe Robin
Original assignee: Riff Kiddy Ltd
Current assignee: Riff Kiddy Ltd
Priority date: 2017-08-16
Filing date: 2017-08-16
Publication date: 2019-02-20
Also published as: GB201713142D0

Abstract

Method for analysing efficiency of motion, comprising: capturing a subject’s motion; measuring displacement characteristics of the captured motion; processing the displacement characteristics to identify and quantify inefficiency in the motion; providing feedback representing the identified inefficiency to a user. Movement components representing optimal motion may be filtered out. Displacement characteristics may comprise velocity, acceleration, third order and higher order derivatives of displacement of a node such as subject’s joint. Energy ratio, frequency domain function, coherence metric, incoherence metric, cross bispectrum, bicoherence, excess phase response, approximate vector motion quality metric and squared normalised metric based on cross spectrum may be calculated, comparing observed motion with ideally stabilised node. Piezoelectric vibrator having output amplitude that varies with level of inefficiency may be provided in wearable belt or vest to provide feedback. Feedback may be through a gauge, meter, graphical representation or may be visual, audible tone getting louder, haptic, biofeedback, electro-stimulation of muscles. May help optimise human body motion and establish ideal musculoskeletal control for elite athletic sports including jumping, throwing, shot put, lifting, running stride, golf swing, swimming stroke. May help with trauma or disease, or with real-time movement planning in robotics. Motion capture may employ video camera, smartphone, laptop, tablet.

Description

SYSTEM FOR ANALYSING EFFICIENCY OF MOTION

Technical Field

The present application relates to a system for analysing efficiency of motion and for providing feedback on identified inefficiency to a user.

Background to the Invention

Efficiency of motion is a key differentiator of elite performers from the general human population. In skilled motions (and often 'basic' motions such as running), high efficiency motion is not guaranteed. Less-than-optimal motion is the norm and contributes to a lack of health and fitness in the population as well as limiting the extent of participation in and enjoyment of physical activities. Increasing motion efficiency is therefore a worthwhile objective that extends far beyond the realm of elite sports and like activities.

For optimising human motion certain parts of the body must be stabilised whilst others are mobilised. Motion in stabilised joints is also used as a feedback system driver in the nervous system to increase joint movements where and in the direction in which they are required to be mobilised. For example, most actions require minimising lumbar spine movement and extensions in the sacroiliac joints whilst simultaneously maximising motion in the hip joint in the required direction (ideally perpendicularly to the joint axis).

For many skilled motions learning is achieved using mimicry - something that proves very difficult where something is not moving or where that movement is very small such as in an ideally stabilised joint. This observation is not widely known, and coaching practise typically ignores such effects. Moreover, the ability of a subject to learn via mimicry is often reduced where the subject's ability to discern motion efficiency is impaired or absent, for example where physiological degeneration is apparent or where cues for learning efficient motion were absent in the original learning environment.

Summary of Invention

According to a first aspect of the invention there is provided a system for analysing efficiency of motion, the system comprising: means for capturing motion of a test subject; means for measuring one or more displacement characteristics of the captured motion; means for processing the captured displacement characteristic(s) to identify and quantify inefficiency in the motion; and means for providing feedback representing the identified inefficiency to a user.

The analysis of human motion is commonplace. The system of the present invention adds the further ability firstly to ascertain the source of inefficiency and secondly to provide some form of feedback to the test subject that enables them to develop more optimal motion to counter the inefficiency.

The processing means may be operative to filter out components of the measured displacement characteristics that represent optimal motion.

The displacement characteristic(s) may comprise one or more of: displacement of a node of the test subject; velocity of the node of the test subject; acceleration of the node of the test subject; and third- and higher-order derivatives of displacement of the node of the test subject with respect to time.

The node of the test subject may be a joint in the body of the test subject.

The means for processing the captured displacement characteristic(s) may be operative to calculate the ratio of energy in the captured motion of the node to energy in an ideally stabilised node during the captured motion and to output the calculated ratio to the means for providing feedback.

The means for processing the captured displacement characteristic(s) may be operative to

H_c0(/) = calculate the frequency domain function u₀(f) u_c(f) , where U_c(f) is the energy in the captured motion of the node and Uo(f) is the energy in the ideally stabilised node.

The means for processing the captured displacement characteristics may be operative to “ -^f) |₍; _{(f )}P calculate the frequency domain function ¹ ' , where U_c(f) is the energy in the captured motion of the node and Uo(f) is the energy in the ideally stabilised node.

The means for processing the captured displacement characteristics may be operative to calculate a metric based on a cross spectrum between a measured velocity of the node and a velocity of the ideally stabilised node.

The means for processing the captured displacement characteristics may be operative to calculate a squared normalised metric based on a cross spectrum between a measured velocity of the node and a velocity of the ideally stabilised node.

The means for processing the captured displacement characteristics may be further operative to calculate a coherence metric based on the squared normalised metric.

The means for processing the captured displacement characteristics may be further operative to calculate an incoherence metric, the incoherence metric being calculated as the complement of the coherence metric.

The means for processing the captured displacement characteristics may be operative to calculate a cross-bispectrum based on a measured velocity of the node and a velocity of the ideally stabilised node.

The means for processing the captured displacement characteristics may be operative to calculate a bicoherence based on a measured velocity of the node and a velocity of the ideally stabilised node.

The means for processing the captured displacement characteristics may be operative to calculate an excess phase response based on a measured velocity of the node and a velocity of the ideally stabilised node.

The means for providing feedback may comprise one or more piezoelectric vibrators.

The one or more piezoelectric vibrators may be provided on one or more wearable devices.

The one or more wearable devices may comprise a belt or vest.

An amplitude of an output of the piezoelectric vibrators may vary in accordance with a level of identified inefficiency.

According to a second aspect of the invention there is provided a method for analysing efficiency of motion, the method comprising: capturing motion of a test subject; measuring one or more displacement characteristics of the captured motion; processing the captured displacement characteristic(s) to identify and quantify inefficiency in the motion; and providing feedback representing the identified inefficiency to the test subject.

Processing the captured displacement characteristic(s) may comprise filtering out components of the measured displacement characteristics that represent optimal motion.

The node of the test subject may be a joint in the body of the test subject.

Processing the captured displacement characteristic(s) may comprise calculating the ratio of energy in the captured motion of the node to energy in an ideally stabilised node during the captured motion and outputting the calculated ratio to provide feedback.

Processing the captured displacement characteristic(s) may comprise calculating the frequency domain function

H_c0(f) =

U₀(f)

U_c(f) , where U_c(f) is the energy in the captured motion of the node and Uo(f) is the energy in the ideally stabilised node.

Processing the captured displacement characteristics may comprise calculating the frequency „ ,,_v Kuf ”^(/ It/ (ff domain function I ^{c 1} , where U_c(f) is the energy in the captured motion of the node and Uo(f) is the energy in the ideally stabilised node.

Processing the captured displacement characteristics may comprise calculating a metric based on a cross spectrum between a measured velocity of the node and a velocity of the ideally stabilised node.

Processing the captured displacement characteristics may comprise calculating a squared normalised metric based on a cross spectrum between a measured velocity of the node and a velocity of the ideally stabilised node.

Processing the captured displacement characteristics may comprise calculating a coherence metric based on the squared normalised metric.

Processing the captured displacement characteristics may comprise calculating an incoherence metric, the incoherence metric being calculated as the complement of the coherence metric.

Processing the captured displacement characteristics may comprise calculating a crossbispectrum based on a measured velocity of the node and a velocity of the ideally stabilised node.

Processing the captured displacement characteristics may comprise calculating a bicoherence based on a measured velocity of the node and a velocity of the ideally stabilised node.

Processing the captured displacement characteristics may comprise calculating an excess phase response based on a measured velocity of the node and a velocity of the ideally stabilised node.

The feedback may be provided using one or more piezoelectric vibrators.

The one or more wearable devices may comprise a belt or vest.

An amplitude of an output of the piezoelectric vibrators may in accordance with a level of identified inefficiency.

Brief Description of the Drawings

Embodiments of the invention will now be described, strictly by way of example only, with reference to the drawings, of which:

Figure 1 is a schematic block diagram illustrating a system for analysing efficiency of motion;

Figure 2 is a schematic representation of a test subject showing analysis nodes used by the system of Figure 1 for analysing efficiency of motion;

Figure 3 is a schematic representation of a feedback device used in the system of Figure 1;

Figure 4 is a schematic representation of a test subject performing a deadlift; and

Figure 5 is a graph showing a bispectral plane.

Description of the Embodiments

Referring first to Figure 1, a system for analysing efficiency of motion is shown generally at 100. The system is for analysing the motion of a test subject 120, for example a person performing an exercise such as running, lifting a weight or the like.

Motion capturing means 140 such as a video camera, which may be provided as standalone device or integrated in another device such as a smartphone, laptop or tablet computer or the like, is provided to capture the motion of the test subject 120. The motion capturing means 6

140 outputs a signal (e.g. a video signal) representing the captured motion to processing means 160, which may be a standalone computer system or may be a processor integrated within a smartphone, laptop or table computer in which the motion capturing means is incorporated.

The processing means 160 is operative to measure one or more displacement characteristics of the captured movement of the test subject. The displacement characteristics may be, for example as velocity components (i.e. displacement with respect to time) in three mutually orthogonal directions of nodes, which are typically joints 122 in the body of the test subject 120 (see Figure 2). Alternatively, displacement and/or acceleration and/or third-order and/or higher-order derivatives of displacement with respect to time can be measured by the processing means 160. Velocity is preferred, however, simply because under constant load conditions, power output is proportional to the square of velocity.

The processing means 160 further processes the measured displacement characteristic(s) to identify and quantify inefficiency in the captured motion of the test subject 120, and to generate a signal representative of the identified inefficiency. Thus, the processing means 160 is operative to filter out components filter out components of the measured displacement characteristics that represent optimal motion, leaving only those components representing sub-optimal movement. This processing can be performed in either the time domain or the frequency domain. For example, if movements are repeated, such as is the nature of a running stride, for example, or if multiple attempts at non-repetitive actions such as a golf swing are sampled in time and synchronously overlaid, then there exists the possibility of analysing and metering the data in the frequency domain rather than the time domain. In the frequency domain, the relevant information will be largely apparent as harmonics of a fundamental repetitive movement frequency.

The resulting signal is output to a feedback means 180, which may be, for example, a gauge or meter 182 (either a physical device or a graphical representation of a gauge presented on a screen of a device), as shown in Figure 3, or some other form of visual, audible or haptic feedback device, which provides feedback to the test subject 120 as to the level of inefficiency in the captured motion. This feedback can be used by the test subject 120 to enable the test subject 120 to adapt their motion in order to reduce the level of inefficiency. As the system 100 continues to capture the test subject's motion, the feedback means 180 provides dynamic feedback to the test subject 120 as to whether adaptations to their motion are reducing the level of inefficiency.

The operation of the system 100 will now be explained with reference to Figure 4, which is a schematic illustration of a test subject 120 performing a deadlift, that is, lifting a load such as a weight from the floor in front of them in a bilaterally symmetrical fashion and finishing in a standing position.

Whilst ideally such a lift is dominated by hip flexion then extension, spinal movement is often observed that detracts from efficiency - and also (not by chance) increases the chance of injury.

The component velocity u_c(t) = u_s(t) - Uh(t) in Figure 4 represents the combined movement of the spine, pelvic and shoulder girdles. For the ideally performed deadlift u_c(t) is zero, such that the spine maintains a straight alignment and core symmetry is maintained. Commonly for a non-ideally executed deadlift, the curvature of the spine causes a non-zero u_c(t). For the test subject and their coach alike, it is difficult to know the extent of u_c(t).

One convenient method for quantifying u_c(t) would be to calculate the ratio of energy in the output movement to that in the ideally stabilised spine. Calculating and minimising the frequency domain function H_co(f), where #_c0(/) =

Up(/)

U_c(/) (with Uc(f) being the frequency-domain energy in the output movement and Uo(f) being the frequency domain energy in the ideally stabilised spine), is one such method, though it is often easier to derive the square magnitude H_co(f)², where

(/)²

Κ(/)Γ R₀(t) |U_C(/)|^{2 R}e7) where the function Ri(t) denotes the autocorrelation of the time sequence Ui(t).

A meter 182 or other form of feedback device is set to display such a measure of the motion as Hco(f)² as shown in Figure 3. A “good” reading denotes a high value of Hco(f)² and low unwanted kinetic energy in stabilised joints, whilst a “bad” reading shows a low value Hco(f)²and relatively high motion in these joints. The meter 182 then provides visual feedback to the test subject 120 that promotes the learning of optimal motion in a manner that is not easily accomplished by other means (excepting chance occurrences, an already elite performer or previous good learning experiences).

Using Hco(f)² to drive the meter has, however, a number of problems. The value of Hco(f)²(though positive only) is not bounded, hence some form of variable meter scaling will be required. Ideally uc(t) will also tend to zero with accompanying issues in the calculation of Hco(f)²· There are also issues from measurement noise as u_c(t) is small and from errors caused by u_c(t) simply being small compared to the two quantities from which it is defined as the difference, in this case u_s(t) and Uh(t).

Furthermore, difficulties in motion analysis prevent the accurate capture of small movements directly, however. Adding a sensor to the hip joint, for example, is prone to error since there can be varying amounts of skin, body fat and clothing between the sensor and the joint. Small movements can then be easily misinterpreted.

Instead of H_co(f)², it is convenient to use a measure based on the cross-spectrum a_co(f) defined between the velocities uo(t) and u_c(t), as the Fourier Transform of their cross-correlation, namely:

a_c0(f) = U_c(f).Uo(-f) where Uo(-f) represents the complex conjugate of Uo(f) that happens in these cases to equal Uo(-f).

It is further convenient to use instead a normalised cross-spectrum measure, C_co(f), where usn-us-n ^coU)’K(/)|.|u_o(f)| or indeed its square where

(/)² k(/)²|-k(/)²|

An advantage of the squared normalised measure, C_co(f)², is that it is a positive, real quantity bounded by zero and unity. In order to drive the meter 182 an ensemble average over f known as the coherence, c_co, is then calculated, where ^Ce0 = (^Ce0(.f)²) f

The average coherence, c_co, is similarly bounded, and a value of unity represents a bad meter reading and poor movement efficiency whilst a value of zero represents a good meter reading and high movement efficiency.

It is further convenient to define an incoherence measure Q_co, as the complement of c_co and therefore representative of the normal meter driving signal, where

QcO - 1-CcO

Given the circumstances of the test subject 120 it might also be pertinent to define other values of Q. For the deadlift example, using a measure of hip velocity, Uh(t), relative to knee velocity, Uk(t), is another pertinent metric; a poorly executed deadlift often utilises knee flexion over and above using hip flexion. A further incoherence measure, Qhk, can therefore be usefully defined as hk (\u _k(f).u (Κ(η²|·Κ(η²|\

The meter 182 can be used to show both Q_co and Qhk via a switch, or alternatively two separate meters 182 can be used. Alternatively a compound quality factor Qt can be defined as the product of Q_co and Qhk such that a single meter 182 showing Qt can be adopted. The effective use of compounded measures is dependent on learned coaching practise and any number, N, of individual measures, Qa.p, can be defined such that

N

Qt ⁼ FI Qafi

It is also often advantageous to consider for one or more measures of used in isolation or in a compound measure to use the bispectrum and the related measure of bicoherence. The crossbispectrum, pco(f), represents energy as a series of bifrequency pairs fi and f₂, and for the measures uo(t) and u_s(t) is defined as

3co(f)(fl, f₂) = Uc(fl).Uc(f₂).Uo(-fo) where fo = fi + f₂.

The normalised measure bicoherence is then defined as _n , \U_c(fQ-U_c(f₂).U₀(-f₀)\ ^{coU ,/2)}Ί£ΜΛ)|Κ(Λ>ο(-/ο)| and its squared counterpart as „ _rf1 . \uffZu_c(f₂).u₀(-f₀)\² &cO \ J J 2 ) I 12 I |2 I 12

Κ(Λ)| ·Κ(Λ)| -|e_o(-/o)|

The bispectrum makes use of information discarded in the conjugation operation involved in the calculation of a_co(f). In the calculation of p_co(fi Jl·) phase information is preserved such that different mechanisms with different phases can be ascertained by the degree to which they are 'coupled'. A further advantage of the bispectrum is that (Guassian) noise with its random phase averages out to zero in the calculation.

The bispectrum and its normalised coherence measure, βεοζίιΤ), are represented in the bispectral plane (Figure 5) where diagonal lines satisfying fo = fi + f2 show the interactions of frequencies fi and f2 that contribute to fo. A useful way of analysing the bispectrum is therefore to take an ensemble average over such values of fo (that is, summing along the lines fo = fi + f2) yielding the quantity, Φ_εο(ίο), defined as

Taking a further ensemble average of <D_co(fo) over fo then yields the single [0,1] bounded value of bicoherence, b_co, where

Uo =Ko(/0)>_/o

Other useful manipulations of data on the bispectral plane can also yield important information. If b_co is used as the target function for biofeedback then, just as with c_co, Qco is defined to drive the meter 182 as

Qco =l-bco bco can also be used as a compound measure with other values of Q_ap as required. There is no restriction on compounding measures of coherence and bicoherence in deriving the signal Qt that best befits the current application of the meter 4. One useful measure that utilises both domains for a single velocity measure is φ_εο where = (φ „ (A)-C_c0(f₀)²>_/o and then as previously for the meter 182

QcO ⁼ 1 ^_ φεΟ

A further useful processing technique is so-called “cepstral” filtering. A minimum phase response is that with the minimum possible delay for each frequency component. The causal delays necessary for corrections to unstable movements tend to cause deviations from minimum phase behaviour and manifests themselves in so-called excess phase components. Hence excess phase components can be used to drive the meter 182.

For a measured response, u_c(t), for example, its minimum phase response, M_c(f), can be found from U_c(f) via the Hilbert Transform denoted by H{}, where

ΛΜ/) = Κ(/)Κ^Μ1ηΚ(/)|}

The (unity magnitude) excess phase response, A_c(f), is then simply

4(/) =

4(/)

M_c(f>

Spectral and bispectral analysis can then be carried out with A_c(f) in place of U_c(f) yielding further discernment of unstable movements as required for driving the meter 182.

In more complex movements there might be further cause to use higher order (poly-)spectral analysis although the same process of isolating the undesirable sections of the polyspectra and using them potentially as feedback for minimisation remains the same. In essence each order of the analysis can be considered as an independent moving part. The basis of this application does not restrict the analysis to that of lower order spectra or bispectra.

More conventional preprocessing of the measured data can also be useful. Windowing in the time domain over the fundamental movement period reduces spectral leakage and so yields better resolution of higher frequency corrective components. Care must be exercised in window selection, however, since better frequency resolution occurs at the expense of blurring information in time (as described by Heisenberg's Uncertainty Principle) and this might in turn disguise corrective movement components rather than highlight them.

In a more complex movement example, such as a golf swing, where there is no lateral symmetry (or in a bilaterally symmetric but dysfunctionally performed movement), the formulation of Q_co is also more complex - indeed the simple deadlift example is merely an example restricted to two dimensions. In the golf swing example, the initial virtually unloaded acceleration and final deceleration of the club is performed with a core rotation. In this or like cases, the whole or just the rotational component of the measured velocity can be windowed to make higher frequency corrective components more obvious in the bispectral plane.

There also exists the possibility to filter the measured velocities in an attempt to highlight higher frequency corrective components. Simple, arbitrary filtering, however, must be applied with caution. The impulsive nature of the ground strike of a runner's foot, for example, generates significant higher frequency energy that is compensated in an ideal movement pattern. The existence of high frequencies is not therefore evidence of correctable instability.

One more useful technique arises from using the autobispectrum. Since measuring u_c(t) is prone to unavoidable errors, a useful enhancement would be to measure and analyse just uo(t) or some such easy to measure velocity. The bispectrum allows discernment of coupled and uncoupled components, so that instability and its attempted correction can be evident in the self-referred autobispectral plane. For such a measure, the autobispectrum, βοο(ί), is defined as

Boo(fi ,f₂)=Uo(fi).Uo (f₂ ).Uo(-fo) so that the autobicoherence, boo, can be calculated as /’oo ⁼ 00 fo ) γθ where Φοο(ίο) is defined as ' ' Jq-J]⁺J2 which in turn leaves the meter driving signal. Qoo, for the meter 182 to be calculated from

Qoo =1-φοο

If a corrective component of movement is considered that compensates for some undesirable instability, it is likely that movement will be (to some degree) reflected in at least one other component of the movement, if not indeed the intended load. A simple analysis of just the output velocity, uo(t), might then show no or very little bispectral content by virtue of a good corrective response by the test subject 120. Often the test subject 120 will have a wellpractised corrective response that appears to the untrained observer to be optimal already. It may well prove beneficial therefore to employ the autobicoherence of a joint elsewhere in the body, such as using the knee velocity, Uk(t), in the deadlift example, where any instability will likely prove more evident. A compound measure Qt ideally then include such measures.

Unusually in a coaching method, little reference has been made herein to an “ideal” motion that serves as a target to the test subject 120 and possibly also the coach. Generally speaking the target of a movement will be easily defined - such as jumping high, running fast or lifting weight. It has been found most useful to leave normal biofeedback and learning mechanisms to gauge success in this aspect. The subject of the current invention is largely to aid the mechanisms that are not easily discerned. It is worthy of note that in many applications, it is often only necessary (often even best) to just provide feedback of the problematic area or isolate the problem in a less complex training movement. For the most part, the resultant effect of a movement is self-evident - throwing an object, for example - and normal biofeedback can be augmented by the feedback provided by the invention. Thus it is not necessarily the case that an ideal movement pattern be known in advance - or indeed known at all.

Using the invention to its greatest benefit will yield movements where the core of the body is completely stabilised - and where the test subject 120 has therefore learned to make maximum use of joints that are best mobilised. At such point in time, it is then feasible to allow the test subject 120 to use the core of the body to provide additional movement to increase the power output. Any attempt to use the core of the body before this point in time is known, however, to curtail the movement efficiency in the long-term and remove the ability of the subject 120 to discern ideal musculoskeletal control.

The exception to the stated coaching methodology therefore arisess when assisting a test subject 120 already possessing high skill levels. In such cases, an ideal motion model (or one adapted according to some form of averaged measures) can be used as a predefined optimal measure of Q, Q_opt, where the meter 182 can then be usefully employed to show the quotient Qt, that is the deviation from the optimal technique:

However, we reiterate that learning such core movements prior to optimising more appropriate measures of Q is known to lead to inefficient and sub-optimal patterns of movement. For the majority of minors and adults outside of elite performance environments, employing Q_opt (or equivalent) will most likely be detrimental to performance in the longterm. A good training exercise is designed to emphasise correcting suboptimal movements and instability. It is not always good therefore for, for example, a prospective runner simply to begin by running.

As will be apparent from the foregoing description , the present invention aids the learning of more optimal movement by developing biofeedback that enables the test subject 120 to discern the differences of relatively small movements, in turn augmenting the natural learning process thereby developing more efficient movements.

The present invention first allows the measurement and analysis of motion where suboptimal components of motion can be ascertained. This type of analysis may also be used to isolate a problematic area by requiring the test subject to perform ever more simple movements until the lack of proper innervation is isolated. The invention subsequently allows feeding back the energy level of the suboptimal components to the test subject in a manner that enhances the chances of learning more optimal motion.

For example, a complex motion such as running stride can be analysed using the system in order to isolate the source of inefficiency. Having identified the source of inefficiency a corrective exercise strategy can be developed to correct the issue. The system described herein can be employed firstly to ensure correct diagnosis of the problem, by analysing progressively more simple movements in order to identify the motion that is at the root of the inefficiency problem, and secondly implement an appropriate corrective exercise strategy, as the user receives direct feedback during the corrective exercise to assist the user in completing the exercises correctly.

Whilst the invention as described produces a scalar index describing the quality of motion, in activities with a defined objective, such as in athletic sports, some form of vectorisation might be desirable. Whilst optimising motion without reference has distinct advantages, it also leaves the possibility of learning an optimal movement pattern that achieves the wrong objective. For an extreme example, a user might develop an optimal techinque at running sideways that will be anything but optimal in a race where they have to run forwards.

In such performance-dependent cases, where there exists a well-defined and appropriate performance measure, such as running speed, jumping height or throwing distance, for example, it is further possible to multiply that end-performance measure, or some scaled version thereof, with the scalar motion quality index to calculate an approximate ‘vector motion quality index’ that may further serve in providing the relevant feedback to maximise the efficiency of the user in that movement and therefore maximising the performance that can be achieved.

Other means of “biofeedback” are feasible besides a simple visual meter or gauge. For example, a more complex display showing the body and indicating where correction needs to be applied could be used. An audible tone that gets louder as suboptimal motion increases will likely suffice, for example, and may be used in conjunction with visual feedback to further enhance the response of the test subject. Direct feedback via electro-stimulation of the appropriate muscles is a further possibility. Instructions from a coach more able to interpret the data may also be considered suitable feedback.

Another direct means of feedback might be provided by, for instance, piezo-electric vibrators attached directly to the user’s body. For example, the piezoelectric vibrators may be provided on a so-called ‘wearable’ (e.g. a belt or vest), such that it becomes possible to obtain the benefits of the invention discussed above without requiring further explanation or involvement of a knowledgeable person, such as a coach, for example.

The output of the piezo-electric device can have its amplitude modulated by the magnitude of the error function (or quality factor) derived from one or more of the above analyses, such that zero error is accompanied by zero vibration and an increasing error is signalled by increasing vibration, for example.

Further to simple amplitude modulation, an automatic gain adjustment servo enables the feedback sensitivity to be increased such that as movement efficiency increases with learning so the resolution of residual errors also increases, thereby maintaining the devices learning stimulus as movement skills increase.

A number of possible formats for such wearables exists. For example, a one such possible wearable incorporates a number of piezoelectric vibrators or other actuators that map the effective lines of force through the body - and in particular its core - in order to alert the user to which part of their musculature needs better innervation. Such a device requires further decoding of the error signal in order that each particular line of force is excited in an optimally proportionate manner.

An alternative wearable device, a simple waist belt of one or more vibrating devices such as piezoelectric vibrators, or a vest including one or more vibrating devices such as piezoelectric vibrators, may be used to convey the requisite feedback direct to the user. In the case where the waist belt or vest comprises more than one vibrating device, such vibrating devices may be arranged circumferentially. It is possible to decode some directional information regarding how the core musculature is better innervated, although success may still be achieved where no such information is encoded and the one or more vibrating devices provide feedback only on the magnitude (i.e. a scalar quantity) of the movement error.

A further alternative wearable device is a wrist band including one or more vibrating devices such as piezoelectric vibrators. Again, such a wearable device may be used to convey feedback direct to the user. It is important to note, however, that these exemplary wearable devices intended to provide haptic feedback to the user in a general sense, rather than specifically targeting particular areas of the body or muscles. In other words, these exemplary wearable devices merely provide feedback to a user that inefficiency has been detected, to encourage the user to modify their movements, rather than indicating to the user that a particular area of the body at which the vibrating devices are located is the source of the detected inefficiency.

As the wearable (and therefore feedback information) becomes simpler, the requirement for training exercises to be performed at a complexity level where the fundamental source of error is more apparent becomes more important. However, once improvements in the subject 18 occur, simpler devices are capable of performing just as successfully as more complex arrangements, even as movement complexity increases.

One advantage of employing a simpler device is that it does not need to be individually fitted to the user, who may range from a small toddler to a large adult - thereby obviating the need for a practitioner or other knowledgeable person such as a coach to fit the device. It will be appreciated that application of the invention in high skill dependent movement, such as in sports performance coaching, may benefit from more complex wearables.

Other forms of direct feedback operating under the guidelines discussed above may also be utilised.

The invention allows the learning of motion ranging from basic movements such as standing up to skilled movements such as throwing a shot put and repeated movements such as the stride of a runner. It can be used in coaching children when teaching from first principles or in adults where a degree of unlearning dysfunctional motion patterns will likely also be encompassed in the process. The invention may also help those afflicted by trauma or disease in best optimising the resulting limited muscular forces to achieve a desired motion.

Additionally, the invention can be used for assessing “motion quality” in one or more movements to use for comparative analyses such as in talent identification or in separating groups in class environments in order to make the best use of available coaching resources

The invention itself maybe embodied as a stand-alone test apparatus or constructed in a mobile rig such that it can track swimmers in a pool, for example. It can also be attached to one or more pieces of gym apparatus or used with face identification, for example, to track the motion of a number of people in a class environment simultaneously. The invention may also be used without the biofeedback where it is simply an effective means of motion analysis.

The analysis methods described also provide a means for real-time movement planning such as required in robotics. Rather than knowing or having to calculate all the components of a movement in advance, preprocessing is reduced to simply defining the target motion. Once initiated, the polyspectral data (or its time domain equivalent) can be used to best adapt the 19 motion as required and provide the basis for learning movements where some memory is utilised. It is not then requisite to know the ideal motion of a complex set of levers in advance.

Claims

1. A system for analysing efficiency of motion, the system comprising:

means for capturing motion of a test subject;

means for measuring one or more displacement characteristics of the captured motion;

means for processing the captured displacement characteristic(s) to identify and quantify inefficiency in the motion; and means for providing feedback representing the identified inefficiency to a user.

2. A system according to claim 1 wherein the processing means is operative to filter out components of the measured displacement characteristics that represent optimal motion.

3. A system according to claim 1 or claim 2 wherein the displacement characteristic(s) comprise one or more of:

displacement of a node of the test subject;

velocity of the node of the test subject;

acceleration of the node of the test subject; and third- and higher-order derivatives of displacement of the node of the test subject with respect to time.

4. A system according to claim 3 wherein the node of the test subject is a joint in the body of the test subject.

5. A system according to any one of claims 2 to 4 wherein the means for processing the captured displacement characteristic(s) is operative to calculate the ratio of energy in the captured motion of the node to energy in an ideally stabilised node during the captured motion and to output the calculated ratio to the means for providing feedback.

6. A system according to claim 5 wherein the means for processing the captured displacement characteristic(s) is operative to calculate the frequency domain function

Hc0(f) = u0(f) uc(f) , where Uc(f) is the energy in the captured motion of the node and Uo(f) is the energy in the ideally stabilised node.

7. A system according to claim 5 wherein the means for processing the captured displacement characteristics is operative to calculate the frequency domain function „ ^2_|4(/)Γ n co\J > - |^ , where Uc(f) is the energy in the captured motion of the node and Uo(f) is the energy in the ideally stabilised node.

8. A system according to claim 5 wherein the means for processing the captured displacement characteristics is operative to calculate a metric based on a cross spectrum between a measured velocity of the node and a velocity of the ideally stabilised node.

9. A system according to claim 5 wherein the means for processing the captured displacement characteristics is operative to calculate a squared normalised metric based on a cross spectrum between a measured velocity of the node and a velocity of the ideally stabilised node.

10. A system according to claim 9 wherein the means for processing the captured displacement characteristics is further operative to calculate a coherence metric based on the squared normalised metric.

11. A system according to claim 10 wherein the means for processing the captured displacement characteristics is further operative to calculate an incoherence metric, the incoherence metric being calculated as the complement of the coherence metric.

12. A system according to claim 5 wherein the means for processing the captured displacement characteristics is operative to calculate a cross-bispectrum based on a measured velocity of the node and a velocity of the ideally stabilised node.

13. A system according to claim 5 wherein the means for processing the captured displacement characteristics is operative to calculate a bicoherence based on a measured velocity of the node and a velocity of the ideally stabilised node.

14. A system according to claim 5 wherein the means for processing the captured displacement characteristics is operative to calculate an excess phase response based on a measured velocity of the node and a velocity of the ideally stabilised node.

15. A system according to any one of claims 8-14 wherein the means for processing the captured displacement characteristics is operative to multiply the metric with an endperformance measure, or a scaled version of an end-performance measure, to calculate an approximate vector motion quality metric.

16. A method for analysing efficiency of motion, the method comprising:

capturing motion of a test subject;

measuring one or more displacement characteristics of the captured motion;

processing the captured displacement characteristic(s) to identify and quantify inefficiency in the motion; and providing feedback representing the identified inefficiency to the test subject.

17. A method according to claim 16 wherein processing the captured displacement characteristic(s) comprises filtering out components of the measured displacement characteristics that represent optimal motion.

18. A method according to claim 16 or claim 17 wherein the displacement characteristic(s) comprise one or more of:

displacement of a node of the test subject;

velocity of the node of the test subject;

19. A method according to claim 18 wherein the node of the test subject is a joint in the body of the test subject.

20. A method according to any one of claims 17 to 19 wherein processing the captured displacement characteristic(s) comprises calculating the ratio of energy in the captured motion of the node to energy in an ideally stabilised node during the captured motion and outputting the calculated ratio to provide feedback.

21 A method according to claim 20 wherein processing the captured displacement ^co(/) = characteristic(s) comprises calculating the frequency domain function

U0(f)

Uc(f) where Uc(f) is the energy in the captured motion of the node and Uo(f) is the energy in the ideally stabilised node.

22. A method according to claim 20 wherein processing the captured displacement „ mi |CW)|2 1 Itz < f >1’ characteristics comprises calculating the frequency domain function 1 c , where Uc(f) is the energy in the captured motion of the node and Uo(f) is the energy in the ideally stabilised node.

23. A method according to claim 20 wherein processing the captured displacement characteristics comprises calculating a metric based on a cross spectrum between a measured velocity of the node and a velocity of the ideally stabilised node.

24. A method according to claim 20 wherein processing the captured displacement characteristics comprises calculating a squared normalised metric based on a cross spectrum between a measured velocity of the node and a velocity of the ideally stabilised node.

25. A method according to claim 24 wherein processing the captured displacement characteristics comprises calculating a coherence metric based on the squared normalised metric.

26. A method according to claim 25 wherein processing the captured displacement characteristics comprises calculating an incoherence metric, the incoherence metric being calculated as the complement of the coherence metric.

27. A method according to claim 20 wherein processing the captured displacement characteristics comprises calculating a cross-bispectrum based on a measured velocity of the node and a velocity of the ideally stabilised node.

28. A method according to claim 20 wherein processing the captured displacement characteristics comprises calculating a bicoherence based on a measured velocity of the node and a velocity of the ideally stabilised node.

29. A method according to claim 20 wherein processing the captured displacement characteristics comprises calculating an excess phase response based on a measured velocity of the node and a velocity of the ideally stabilised node.

30. A method according to any one of claims 23- further comprising multiplying the metric with an end-performance measure, or a scaled version of an end-performance measure, to calculate an approximate vector motion quality metric.

31. A system according to any one of claims 1 to 15, wherein the means for providing feedback comprises one or more piezoelectric vibrators.

32. A system according to claim 31, wherein the one or more piezoelectric vibrators are provided on one or more wearable devices.

33. A system according to claim 32, wherein the one or more wearable devices comprises a belt or vest.

34. A system according to claim 31, wherein an amplitude of an output of the piezoelectric vibrators varies in accordance with a level of identified inefficiency.

35. A method according to any one of claims 16 to 30, wherein the feedback is provided using one or more piezoelectric vibrators.

36. A method according to claim 35, wherein the one or more piezoelectric vibrators are provided on one or more wearable devices.

37. A method according to claim 6, wherein the one or more wearable devices comprises a belt or vest.

38. A method according to claim 35, wherein an amplitude of an output of the piezoelectric vibrators varies in accordance with a level of identified inefficiency.