LOKOGRAM SOFTWARE TOOL FOR ANALYSIS OF WALKING EXERCISES

Description

[0001] The present invention relates to a method and a device for analyzing the regularity and symmetry of a sequence of walking or running cycles of a person. STATE OF THE ART

[0002] Recently, advances in electronics and computing have led to the development of new sensors that make gait measurement possible in routine clinical consultations (e.g., using inertial sensors). These sensors provide access to signals that must be synthesized to draw conclusions about a subject's gait.

[0003] Gait parameters such as speed or cadence are the most synthesized form of gait signals because they reduce them to a number. Thus, manufacturers of gait measurement tools provide comprehensive reports in the form of a list of parameters reflecting the different aspects of a subject's gait such as symmetry, regularity or speed of onset.

[0004] There can be more than a hundred parameters. They are conveniently represented in the form of a table, bar chart or curve.

[0005] However, the topic of visualizing intermediate data between uninterpretable raw time signals and parameters is little explored.

[0006] Some parameters require thresholds to be defined in order to be calculated. For example, a subject is defined as having reached a steady state when the amplitude of their cycles has exceeded 67% of the amplitude of cycles 5 to 10 (it is considered that there is no longer any need to speak of initiation at more than 5 cycles). Some cycles may have limit values ​​(60% or 70% of the amplitude of the remote cycles).

[0007] This leads to errors related to the chosen threshold which could be avoided by looking directly at the raw data.

[0008] The parameters average the gait characteristics over the entire exercise, and thus obscure the progress of the exercise. For example, to assess the regularity of a subject's gait, we use the standard deviation of the duration of the gait cycles. If the cycles are generally regular and one cycle is longer than the others, the standard deviation will be increased. We will not know that this increase is due to a single isolated cycle, nor at what point in the gait it occurred. However, the number of erratic cycles, and their time of occurrence, are important analytical information.

[0009] Parameters are good indicators of a subject's walking performance. However, they do not provide analytical information about the origin of the change in gait. For example, walking speed is a good overall assessment; however, speed can be reduced by shorter steps or a decreased cadence. In turn, step size and cadence can be affected by various factors.

[0010] Gait measurement presents significant inter-individual fluctuations due mainly to the anatomical and functional variabilities of each individual which influence the positioning of the sensors, and to the locomotor style variabilities of each individual. The parameters are affected by this variability and there is no reference method to overcome it.

[0011] MOE-NILSSEN R ET AL, "Estimation of gait cycle characteristics by trunk accelerometry", JOURNAL OF BIOMECHANICS, PERGAMON PRESS, NEW YORK, NY, US, (2004), vol. 37, no. 1, doi:10.1016 / S0021-9290(03)00233-1, ISSN 0021-9290, pages 121-126, XP004895603, is cited to define the general state of the art.

[0012] CHE-CHANG YANG ET AL, "Real-Time Gait Cycle Parameter Recognition Using a Wearable Accelerometry System", SENSORS, (20110725), vol. 11, no. 12, doi:10.3390 / s110807314, pages 7314 - 7326, XP055364449, is cited to define the general state of the art.

[0013] The document YANG MINGJING ET AL, "iGAIT: An interactive accelerometer based gait analysis system", COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE, (2012), vol. 108, no. 2, doi:10.1016 / J.CMPB.2012.04.004, ISSN 0169-2607, pages 715 - 723, XP028974078, discloses a device and a method allowing, by a visual representation, the analysis of the regularity and symmetry of a sequence of N walking or running cycles of the same person, the device comprising sensors for measuring raw inertial time signals or signals relating to a physical quantity of an anatomical segment of the person, and a processing and calculation unit, connected to the measurement sensors. STATEMENT OF THE INVENTION

[0014] There is a need in the state of the art for a method that allows the temporal study of the regularity and symmetry of a sequence of walking or running cycles of a person.

[0015] One aim of the invention is to enable the shape of the signal to be taken into account in order to account for the regularity of walking or running, while avoiding the variability of the duration or amplitude of the walking or running cycles.

[0016] In the invention, a visualization matrix or table of a walking or running exercise is constructed. This visualization matrix M(i,j) is called a "locogram", and is a Gram matrix in the sense that xij=xji with xij being the value located in the ith row and the jth column in the matrix M(i,j).

[0017] The locogram is robust to the type of sensors used (accelerometers, gyroscope, infrared tracer, as well as to their anatomical locations for recording walking (foot, leg, belt, wrist, head).

[0018] Thanks to the visualization matrix of the invention, the professional who analyzes the gait (doctor, physiotherapist, nurse, podiatrist, coach, sports trainer, shoe designer, etc.) can have an overall view of the walking exercise from which he can perform semiology in order to get an idea of ​​the quality of a patient's gait and easily identify atypical steps in a given exercise. Parameters alone are not enough to have a satisfactory overview of the patient. The visualization matrix is ​​halfway between parameters and raw time signals, and gives the doctor access to the patient's gait data.

[0019] The visualization matrix is ​​based on dividing walking or running into cycles. The aim is to compare the cycles using "distances" or mathematical metrics. Depending on the choice of distance, this or that aspect of the signal will be compared.

[0020] The present invention relates to a method as defined in the appended claims.

[0021] The present invention also relates to a device as defined in the appended claims. DESCRIPTION OF FIGURES

[0022] Other objectives, characteristics and advantages will emerge from the detailed description which follows with reference to the drawings given for illustrative and non-limiting purposes, among which: there Figure 1a represents the device according to the invention; the Figure 1b represents the method according to the invention; the Figure 2 represents the definition of the terms amplitude, duration and shape of walking or running cycles; Figure 3 represents the visualization tool according to the invention of the walking cycle of a person's foot on a given walking exercise comprising a back and forth over a defined distance; the Figures 4a and 4brepresent the visualization tool F(i,j) which relates to the shape similarity coefficient according to the invention for a healthy subject, respectively for a Parkinson's subject, comparing for each subject, the shape of their walking cycles on a given walking exercise comprising a back and forth over a defined distance; Figures 5a and 5b represent the gait visualization tool or square matrix A(i,j) which relates to the amplitude similarity coefficient according to the invention for a healthy subject, respectively for a Parkinsonian subject, comparing for each subject, the amplitude of their gait cycles on a given walking exercise comprising a back and forth over a defined distance; Figures 6a and 6brepresent the gait visualization tool or square matrix D(i,j) which relates to the duration similarity coefficient according to the invention for a healthy subject, respectively for a Parkinson's subject, comparing for each subject, the duration of their walking cycles on a given walking exercise comprising a round trip over a defined distance; the Figure 7 represents, as a function of time, the raw time signals of acceleration from a triaxial accelerometer along the three axes x, y and z, for the right foot of a person moving on a given walking exercise involving a back and forth over a defined distance; figure 8 represents, as a function of time, the raw time signals of the acceleration without gravity corresponding to the raw signals of the Figure 1 from which the raw time signals from the triaxial accelerometer along the three axes x, y and z were subtracted for the right foot of a stationary person; Figure 9represents, as a function of time, the standard of the acceleration without gravity of the person's right foot on a given walking exercise involving a round trip over a defined distance; Figures 10a, 10b represent, as a function of time, the standard of the acceleration without gravity of the person's right foot, the standard of the acceleration without gravity of the person's left foot and the location of the times τ_right and τ_left corresponding to the walking cycles C1, C2....Cnd+ng on a given walking exercise involving a round trip over a defined distance; Figure 10c represents the resampling of the walking cycles C1, C2....Cnd+ng.

[0023] The figures presented in this document are for illustrative purposes only. DETAILED DESCRIPTION OF THE INVENTION

[0024] The present invention relates to a device 1 allowing the temporal analysis of the regularity and symmetry of a sequence of N walking or running cycles of a person, without calculating parameters based on the global consideration of the N walking or running cycles, by comparing each cycle with each of the other cycles.

[0025] The term 'walking or running cycle' refers to the phases defined between two heel strikes of the same foot on the ground when a person is walking or running and which occur at certain times between the two heel strikes. It is therefore representative of the behavior of the person's foot during a step.

[0026] As is known from the state of the art, the different phases can be defined as a percentage, with 0% corresponding to the first heel strike and 100% to the second.

[0027] The cycle can be separated into two phases: the support phase (0 to 60%) where the foot is in contact with the ground, and the swing or pendulum phase (60% to 100%) where the heel is in the air.

[0028] This expression defines all walking styles and running styles; for example, stamping should be considered as part of the category defined by the expression 'gait cycle'.

[0029] Device 1 illustrated on the Figure 1a comprises sensors 2 for measuring raw signals i of a physical quantity of an anatomical segment, measured during a sequence of N walking or running cycles of at least one foot of the person.

[0030] The measurement sensors 2 can be any sensor, for example such as: an accelerometer, a gyroscope, electromyography, pressure insoles or infrared tracers / video or IR acquisition devices.

[0031] They can be placed for example on the foot, ankle, belt, wrist, head.

[0032] The measured physical quantity can, for example, be the acceleration norm, the acceleration norm without the Earth's gravity component, speed, angular speed, displacement, position, etc.

[0033] For example, the quantities from inertial sensors can be linear acceleration or angular velocity along one of the 3 axes of space. With other sensors, other quantities can be the position of the foot in space with stereophotogrammetry or ground pressure with pressure sensor force insoles.

[0034] As shown in the Figure 1a , the device 1 comprises a processing and calculation unit 3 arranged for: processing and separating the raw data into distinct time signals Ci, each time signal Ci being a series of points of the measured physical quantity as a function of time and having a given shape, amplitude and duration, the series Ci being associated with a single given cycle i of the person among the N walking cycles as represented on the Figure 2 ; calculate at least one similarity coefficient between a signal Ci associated with a cycle i, with another signal Cj associated with another cycle j, representative of the similarity between the two signals Ci and Cj; store in a square matrix M(i,j) indexed in row i and column j, and indicating for each cell (i, j), the value of the similarity coefficient between cycles i and j.

[0035] In these calculations and the arrangement of the determined values ​​in the matrix M(i,j), the natural integers with i and j vary from 1 to N, and the N walking cycles studied are ordered for the given foot, chronologically according to their order in the walking sequence.

[0036] For example, cycle 1 is compared to cycle 2 ( m[1,2] or m[2,1] ), cycle 3 ( m[1,3] or m[3,1] ), cycle 4 ( m[1,4] or m[4,1] , etc... cycle 2 is compared to cycle 1 ( m[2,1] or m[1,2] ), cycle 3 ( m[3,1] or m[1,3] ), cycle 4 ( m[2,4] or m[4,2] , etc...

[0037] And so on.

[0038] The matrix visualizes by construction also the case where cycle 1 ( m[1,1] ), cycle 2 etc. are compared to themselves.

[0039] The processing (“compend resampling”) and separation of raw data is illustrated in the Figures 10A to 10D , and the distinct time signals Ci, after processing above are named C̃ i on the Figure 10D (which corresponds to the C'i terminology in the text).

[0040] The similarity coefficient can be: a similarity coefficient of the form fij of the two signals Ci and Cj (as represented in the Figures 4a and 4b ), a similarity coefficient of the amplitude aij of the two signals Ci and Cj (as represented on the Figures 5a and 5b ), or a similarity coefficient of the duration dij of the two signals Ci and Cj (as represented on the Figures 6a and 6b ).

[0041] A synonym for 'coefficient' is 'algebraic distance' in the mathematical sense of the term, for example, and for 'similarity' is resemblance.

[0042] The 'similarity' expression implies the symmetry of the coefficients and therefore that fij=fji, aij=aji and dij=dji.

[0043] The similarity coefficient can be considered as a correlation coefficient in a number of cases.

[0044] Other similarity coefficients can be used depending on the aspect of the shape that we want to compare between two signals: Dynamic Time Wrapping (DTW), Spearman's coefficient, Euclidean distances (L1, L2 and L∞).

[0045] Thus, with a similarity coefficient that varies in an interval [a; b], if two signals (or the two walking or running cycles) are identical, the coefficient can be equal to b, and if two signals (or the two walking or running cycles) have no resemblance, the coefficient can be equal to a, the similarity between the signals increasing linearly between a and b.

[0046] Advantageously, the processing and calculation unit 3 is arranged to: calculate several different similarity coefficients fij, aij, dij, each similarity coefficient fij, aij, dij being associated with the same color scale, each color having the same similarity meaning regardless of the coefficient used fij, aij, dij, as illustrated in the Figures 4a to 6b ; present the values ​​of the similarity coefficients fij, aij, dijj in different square matrices F(i,j), A(i,j) D(i,j), one square matrix per type of similarity coefficient.

[0047] The device 1 also has display means 4 connected to the processing unit and displaying the matrix M(i,j) for i and j varying from 1 to N, each value of the similarity coefficient being represented in the matrix M(i,j) by a graduated visual representation to enable the similarity between the walking or running cycles i and j to be visualized with the naked eye.

[0048] Advantageously, the similarity coefficients (fij, dij, aij) are chosen: so that once calculated, all the values ​​of the similarity coefficients (fij, dij, aij) are in the same interval [a; b] and so that the higher the values ​​of the similarity coefficients (fij, dij, aij) are, the more similar the signals Ci, Cj are.

[0049] Advantageously, this visual representation is graduated to represent the value of the similarity coefficient. In other words, there is a correspondence between the scale of the similarity coefficient values ​​and that of the colors.

[0050] This correspondence may or may not be proportional.

[0051] Preferably, each value is represented by a color located in a graduated color scale to allow the similarity between the two walking or running cycles i and j to be visualized with the naked eye (which amounts to studying the similarity between the two signals Ci and Cj).

[0052] Advantageously, the scale used by the display means 4 is chosen without thresholds or continuous as illustrated in the Figures 3 to 6b .

[0053] Here the scale is [0; 1] for the shape similarity coefficient, the amplitude similarity coefficient and the duration similarity coefficient allowing a quick and easy comparison of the different similarity coefficients with each other.

[0054] Each box in the matrix can be colored to correspond to the degree of similarity / resemblance of two steps to each other.

[0055] A warm color (example red) is close to 1, and reflects a great resemblance (similarity).

[0056] A cold color (example blue) is close to 0, and reflects a weak resemblance (similarity).

[0057] Note that this scale is reversed between the Figure 3 on one side and the Figures 4a to 6b on the other: on the Figure 3 dark colors indicate a resemblance, which is the opposite on Figures 4a to 6b or light colors indicate a resemblance.

[0058] The display means 4 display a visual representation of the similarity matrix such that: Each box represents a similarity coefficient between the signal Ci and Cj; the value of the similarity coefficient (fij, aij, dij) is represented in a graduated color scale.

[0059] In a first embodiment illustrated on the Figure 3, it is represented the coefficient of similarity between the signal Ci relating to the walking or running of a single foot (right foot, or respectively left foot) of the person, and the signal Cj relating to the walking or running of the same foot, and number of cycles N=Nd (respectively Ng) number of cycles of the right foot (respectively number of cycles of the left foot).

[0060] In a second embodiment illustrated in the Figures 4a to 6b , the display means 4 display: the similarity coefficients (fij, aij, dij) calculated for a single foot 3; also the similarity coefficient (fij, aij, dij) between the signal Ci relating to the walking or running of the person's right foot, and the signal Cj relating to the walking or running of the person's left foot, and N=Nd +Ng.

[0061] In this case, all the cycles studied with the right foot, then all the cycles studied with the left foot (or vice versa) are ordered chronologically, on the abscissa and ordinate of the matrix, as illustrated in the Figures 4a to 6b . They are separated here in the representation by a zone O, which gives 4 sub-squares.

[0062] So, the similarity matrix: “right foot / right foot” is represented by the lower left sub-square; “left foot / left foot” is represented by the upper right sub-square; “left foot / right foot” is represented by the lower right sub-square.

[0063] When the similarity coefficient is a coefficient of form fij, the processing and calculation unit 3 is arranged to: normalize in duration each time signal Ci with each time signal Cj so that the two signals Ci and Cj have the same duration; normalize in amplitude each time signal Ci; calculate a similarity coefficient of shape fij between each normalized signal C'i, with another normalized signal C'j, and this for all the walking cycles.

[0064] In other words, to study the similarity of the shape of the cycles between them, we must free ourselves from their difference in amplitude and their difference in duration.

[0065] For example, duration normalization is achieved by linear resampling, or by DTW, as illustrated in the Figure 10c .

[0066] For example, amplitude normalization is achieved by dividing the Signal Ci by the standard deviation or by the effective value ("Root Mean Square").

[0067] In a first embodiment, the shape similarity coefficient fij can be the Pearson coefficient, as illustrated in Figures 4a and 4b , the values ​​being brought back to 0 in the case of Pearson coefficient values ​​less than zero.

[0068] Other mathematical distances are possible, such as Spearman's correlation distance or the "dynamic time wrapping" technique.

[0069] If we are only interested in the shape, it is important to free ourselves from the amplitude and duration. To do this, the signal is re-normalized by resampling the signal to 100 samples (duration normalization) and dividing the signal by the standard deviation of the cycle (amplitude normalization). The Pearson correlation distance preserves the timeline and strictly compares the shape of the steps. The Spearman correlation distance and dynamic time wrapping distort the timeline and indicate whether there is a possible distortion so that the shape of two cycles resembles each other or whether the shapes are truly different.

[0070] In a second embodiment, the similarity coefficient may be a similarity coefficient of duration dij.

[0071] In this case, as illustrated in the Figures 5a and 5b , the processing and calculation unit 3 is arranged for example for: calculate a duration Di, Dj of the signal Ci and the duration of the signal Cj, calculate the duration similarity coefficient dij which is the ratio of the smallest of the two durations Di, Dj divided by the largest duration Di, Dj, in order to bring the value of the coefficients between 0 and 1.

[0072] The duration of the signal Ci is equal to the number of samples, each sample being taken regularly at a given frequency.

[0073] In a third embodiment, the coefficient may be an amplitude similarity coefficient.

[0074] In this case, as illustrated in the Figures 6a and 6b , the processing and calculation unit 3 is arranged for example for: calculate an amplitude Ai, Aj of the signal Ci and of the signal Cj, calculate the amplitude similarity coefficient aij which is the ratio of the smallest amplitude Ai, Aj divided by the largest amplitude Ai, Aj in order to bring the value of the coefficients between 0 and 1.

[0075] The amplitude Ai and Aj is for example the standard deviation or the effective value (“Root Mean Square”).

[0076] Once these calculations have been carried out to find the values ​​of the similarity coefficients, the processing and calculation unit 3 is arranged to calculate: parameters for assessing the regularity of the gait; parameters for assessing the symmetry of the gait; the number of cycles required to establish a gait sequence or sub-sequence.

[0077] Especially : the parameter evaluating the regularity of the gait: the average of all cells (i, j) of the matrix M(i,j) (A(i, j); D(i,j); F(i,j) ) excluding the cells i=j, the parameter evaluating the regularity of the gait: the standard deviation of all cells (i, j) of the matrix M(i,j) excluding the cells i=j, the parameter evaluating the symmetry of the gait: the average of the coefficients of the cells with i ranging from 1 to Nd (or from 1 to Ng), and j ranging from Nd+1 to Nd+Ng (j ranging from Ng+1 to Nd+Ng).

[0078] The present invention also relates to a method illustrated in the Figure 2 and using the device 1 described above.

[0079] The process, illustrated in the Figure 2 , includes: a step of detecting (i) the raw time signals; a step of processing and separating (ii) the raw time signals into distinct time signals Ci; a step of calculating (iii) the similarity coefficient; a step of arranging (iv) in a matrix M(i,j), the value of the similarity coefficient at row i and column j; with i and j natural integers varying from 1 to N, the N walking or running cycles being ordered chronologically; a display step (v) displaying the matrix M(i,j); a visualization step (vi) by an operator.

[0080] The operator visualization step allows to determine: the number of cycles required to complete a walking or running sequence or a walking subsequence, the walking or running subsequence being: a half-turn, the initiation of walking or running, an acceleration, a deceleration, a steady state, a stop, a turn; the number of cells having a color associated with a low coefficient value; whether a cycle has one or more cells having a color associated with a low or high coefficient value; whether the two consecutive walking cycles have one or more cells having a color associated with a low or high coefficient value; identify changes in step rate during the exercise by color heterogeneities; compare the color of the right-foot-right-foot cells with the right-foot-left-foot cells.

[0081] It may be provided that the processing and calculation unit 3 counts the number of cycles defined above, or only displays certain values ​​of similarity coefficients between the cycles in view of predefined threshold values. DESCRIPTION OF THE PROTOCOL FOR CARRYING OUT A WALKING EXERCISE AND DESCRIPTION OF THE POST-STROKE PARKINSONIAN COHORT 2. Step 1: Data Acquisition

[0082] A subject wearing a triaxial accelerometer on the dorsal surface of each foot performs a 10-meter round trip walking exercise from a standing start. Each of the sensors records a 3-dimensional time signal (along the three axes x, y and z). We call respectively acc right And acc LEFT the signals for the right and left foot, as shown on the curves of the Figure 7 . Step 2: Identify the beginnings of the walking cycles

[0083] We identify (manually or automatically) in the signals the times corresponding to the beginning of the walking cycles. These instants, according to the definition of Mariani et al. (2013), correspond to the "heel-strikes", that is to say the instants when the heel touches the ground. We define two sets τ right And τ LEFT corresponding respectively to the set of heel-strikes for the right foot and the left foot ( Figures 10A and 10B ). We call N d + 1 and N g + 1 respectively the number of heel-strikes detected for the right foot and for the left foot. The gait cycle i of the right foot is defined as the part of the signal starting at the instant τ right,i and ending at the beginning of the next walking cycle τ right,i +1, as shown by the Figures 10A and 10B . So, N d And N g are the gait cycle numbers of the right and left foot respectively. Step 3: Remove Gravity

[0084] At the beginning of the protocol, or in a preliminary stage, the subject was asked to remain standing and motionless. The (constant) value of the accelerations during this phase was recorded and stored in a vector accimmobil e droit , x accimmobil e droit , y accimmobil e droit , z As shown on the curves of the figure 8 , gravity is removed according to the following process: accfre e droit = ac c droit , x ac c droit , y ac c droit , z − accimmobil e droit , x accimmobil e droit , y accimmobil e droit , z

[0085] We repeat the same process for the left foot. Step 4: Calculation of the standard

[0086] As shown on the curve of the Figure 9 , the norm of the acceleration vector is calculated at each instant t: s droit t = accfre e droit , x t 2 + accfre e droit , y t 2 + accfre e droit , z t 2 the same for the left foot.

[0087] This fusion of the three axes of acceleration makes it possible to be independent of the position of the sensor which is a source of significant measurement inaccuracy in accelerometer measurement. Step 5: Creating the walk cycles

[0088] As shown on the curve of the Figures 10C to 10D, we note the set of walking cycles of the exercise in the following manner: For

[0089] ∀ i ϵ 1 , N d + N g c i = s droit τ droit , i τ droit , i + 1 si 1 ≤ i ≤ N d c i = s gauche τ gauche , i − N d τ gauche , i − N d + 1 si N d + 1 ≤ i ≤ N d + N g

[0090] We obtain { c 1 , c 2, ..., c Nd , c N d+1, c N d+2 , ... , c Nd+Ng} the set of walking cycles of a walking exercise with { c 1 , c 1, ..., c Nd} the cycles of the right foot and { c N d+1, c N d+2 , ... , c Nd+Ng} the left foot cycles. Step 6: Calculating Metrics

[0091] From all the steps present, a time signal Ci is determined comprising a series of points of the measured physical quantity as a function of time.

[0092] Each time signal Ci has a given shape, amplitude and duration, the series Ci being associated with a given cycle i, as represented in the Figure 2 . 6.1 Some preliminary notations

[0093] Given a vector x composed of N samples{ x 1 , x 2, ..., x N}, we define the following quantities: Average : x ¯ = 1 N ∑ i = 1 N x i Standard deviation: std x = 1 N ∑ i = 1 N x i − x ¯ 2

[0094] Given two vectors x And y , we define: Covariance: cov x y = 1 N ∑ i = 1 N x i − x ¯ y i − y ¯ 6.2 Duration Metrics

[0095] We define the duration metric by: ∀ i , j ϵ 1 , N d + N g 2 D c i c j = min c i c j c j c i or | c j | is the number of elements in the cycle j.

[0096] This is the ratio of the cycle durations, always putting the longest cycle in the denominator. 6.3 Amplitude metric

[0097] We define the amplitude metric by: ∀ i , j ϵ 1 , N d + N g 2 A c i c j = min std c i std c j std c j std c i

[0098] This is the ratio of standard deviations, always putting the cycle with the largest amplitude in the denominator. 6.4 Metric form

[0099] At this point, the cycles are not all the same length. We normalize the length of each cycle to 100 samples using the “resample” function in Matlab ®< (“MATLAB 2014a, The MathWorks, Natick, 2014.”), as shown in figures 10D . We call c ι ˜ the renormalized version of c i

[0100] We finally define the form metric by: F c i c j = max 0 , P c ι ˜ c j ˜ Or ∀ i , j ϵ 1 , N d + N g 2 P c ι ˜ c j ˜ = cov c ι ˜ c j ˜ std c ι ˜ std c j ˜ Step 7: Construction of the visualization matrices

[0101] So we get 3 matrices for a walking exercise D c i c j i = 1 .. N d + N g j = 1 .. N d + N g A c i c j i = 1 .. N d + N g j = 1 .. N d + N g F c i c j i = 1 .. N d + N g j = 1 .. N d + N g

[0102] We have the following form: Comparaison droite gauche Comparaison gauche Comparaison droit Comparaison droite gauche DETAILED DESCRIPTION OF THE MATRICES

[0103] Here are described the Figures 4A to 6B .

[0104] The figures presented in this document are for illustrative purposes only. According to the construction of the "locogram", the size does not differ between the shape, amplitude and duration. The differences in the number of squares observed here between the Figures 4A And 6Bare linked to a layout defect.

[0105] The healthy subject completed a total of 38 cycles compared to 54 for the Parkinson's subject for the same distance, which reflects pathological walking with small steps ( Figure 4A and 4B ). In the shape matrix of the healthy subject, the homogeneous light-colored tiles reflect a regular gait ( Figure 4A ). We observe a steady walking pattern reached in 1 cycle since only cycle no. 1 of the right foot is very different from the others.

[0106] For the Parkinsonian subject, the heterogeneity of the shape matrix reflects an irregular gait ( Figure 4B ). The steady state is reached in 2 cycles (cycle no. 1 of the left foot and cycle no. 1 of the right foot).

[0107] The homogeneity ranges along the diagonal of the "locogram" are signs of quality walking. In fact, this means that all the cycles in this homogeneity range are similar to each other and are regular.

[0108] On the Figure 4A (healthy subject), these ranges are long and omnipresent. We distinguish two for each foot which correspond to the outward and return of the walking exercise. The outward and return for each foot are separated by a black cycle (central cross in the squares at the bottom left and top right: cycle n° 10 right foot for example) which corresponds to the half-turn cycle and which normally does not resemble any other cycle.

[0109] On the Figure 4B (pathological subject) the homogeneity ranges are shorter. We identify several homogeneous ranges: cycles n° 2 to 6 and cycles n° 20 to 25 for the right foot, cycles n° 2 to 7, cycles n° 10 to 14 and cycles n° 15 to 26 for the left foot. It should be noted that these homogeneity ranges of relative quality. Indeed, they are homogeneous compared to the rest of the "locogram" of the Figure 4B but relatively heterogeneous compared to the “locogram” of the Figure 4A . On the Figure 4B, these ranges are at the beginning of the walk: the subject is not tired and can perform a quality walk. These ranges are also at the end of the exercise which can reflect a significant time needed to obtain a steady walking regime. In the middle of the exercise, the walk is degraded as evidenced by the absence of homogeneous ranges.

[0110] Furthermore, we observe erratic cycles reflecting irregularities that are well localized in time ('freezing', stumbling): cycle no. 19 right foot, cycles no. 8 and no. 9 left foot for example. Finally, we also observe more diffuse irregularities in time taking the form of heterogeneous ranges: cycles no. 15 to 20 right foot and cycles no. 15 to 20 left foot for example. It should be noted that these two heterogeneous ranges immediately follow the U-turn. This may reflect a difficulty in making the U-turn and resuming quality walking after it. There are no heterogeneous ranges or erratic cycles observable on the Figure 4A .

[0111] U-turns are also indicative of the quality of locomotion. On the Figure 4A (healthy subject), the half-turn is composed of only 1 cycle in a physiological manner: cycle n° 10 right foot. On the Figure 4B(pathological subject), the half-turn is composed of 2 cycles: cycle n° 14 right foot and cycle n° 15 left foot. We note that the half-turn on the Figure 4B has an impact on the surrounding cycles (cycles no. 14 to 17 left foot are darker) while we do not observe such an impact in the Figure 4A .

[0112] In the amplitude matrix of the healthy subject, the growth, then the decrease in the similarity of the amplitude of the acceleration between the cycles reflects the establishment of the cruising speed ( Figure 5A ). This growth and decline are less visible for the Parkinsonian subject ( Figure 5B ).

[0113] In the duration matrix of the control subject, the overall lighter color than that of the Parkinsonian subject reflects a greater resemblance between the gait cycles in terms of duration ( Figure 6A and 6B). The homogeneity of the color reflects less variability between walking cycles in terms of duration. DIGITAL ANALYSIS FROM THE “LOCOGRAM” ON 40 SUBJECTS

[0114] We calculate the standard deviation (SD) of the duration of the walking cycles on the established gait (all the walking cycles without the first 3 cycles, the 3 cycles preceding the U-turn, the cycles of the U-turn, the three cycles following the U-turn and the last three cycles of the exercise): P 0 = DS c i i = marche é tablie where | c i | is the number of samples of c i .

[0115] We calculate the average of the “locogram” on the established step: P 1 = moyenne F c i c j i = marche é tablie j = marche é tablie

[0116] We calculate the standard deviation of the “locogram” on the established step: P 2 = DS F c i c j i = marche é tablie j = marche é tablie

[0117] We calculate the number of hierarchical clusters on the obtained “locogram” with the same stopping rule: P 3 = cluster F c i c j i = 1 .. N r + N l j = 1 .. N r + N l

[0118] The results comparing the parameters {P1, P2, P3} for two groups of healthy subjects (young and elderly) and patients with Parkinson's disease are reported in Table 1. The results comparing the correlation of the parameters {P1, P2, P3} with the severity of clinical impairment of patients with Parkinson's disease (assessed with the UPDRS III score) and the quality of gait (assessed with gait speed and P0) are shown in Table 2.

[0119] The results show that the "locogram" of patients with Parkinson's disease is significantly more degraded than that of healthy elderly subjects and young subjects on the 3 parameters {P1, P2, P3} from the "locogram" (Table 1).

[0120] On the other hand, there is a correlation between the clinic (UPDRS III score) and the quality of walking assessed by the "locogram" according to the three parameters {P1, P2, P3} from the "locogram" (Table 2).

[0121] Finally, there is a correlation between the quality of walking assessed with the state-of-the-art parameters (walking speed and P0) and the quality of walking assessed by the “locogram” according to the three parameters {P1, P2, P3} (Table 2).

[0122] So, on a set of 40 subjects, the "locogram" allows us to numerically assess the quality of walking of a patient with Parkinson's disease (in addition to the visual point of view). ADVANTAGE OF MATRICES

[0123] The visualization matrix displayed on the Figures 4A And 6B calculated on a 10 meter round trip walking exercise from a standing start gives access to the following walking parameters: Number of cycles required to complete the exercise; Number of cycles required to turn around; Number of cycles required to initiate walking.

[0124] The visualization matrix presents these parameters more intuitively than in list form. Moreover, for the last two parameters, the definition in the literature is based on experimentally set thresholds. The advantage of a global representation rather than a single parameter is to avoid errors due to an arbitrary threshold.

[0125] The shape of the visualization matrix allows us to answer the following questions: Is the gait generally regular? Has the subject reached a steady gait pattern? When? Are there several different steady gait patterns? Are there erratic gait cycles ('freezing' in English)? When?

[0126] The visualization matrix allows us to have an idea of ​​the progress of the walking exercise unlike the parameters which do not have a time value. The visualization matrix gives access to the notion of walking regime which is a notion for which we do not have exploration tools apart from the parameter plot chronologically walking cycle after walking cycle.

[0127] Each subject has a particular walking style that causes a large inter-individual variability of the parameters. This style is reflected in the signals by a very reproducible signature from one walking cycle to another in the healthy subject. The parameters are not a suitable tool for describing this signature. The visualization matrix, with an adapted choice of distance (see the section "description of the invention") compares the shape of the steps and evaluates the resemblance of the signature of a cycle with respect to the other walking cycles. Thus, it evaluates the quality of the reproducibility of the gait and erases the personal style because in the distance between two cycles of the same person, the individual variabilities cancel each other out. The visualization matrix gives for each subject a table of cells bearing a number between 0 and 1. This allows the visualization matrices to be compared with each other reliably.

[0128] The visualization matrix provides access to new, unique walking parameters such as: The number of clusters with the same stopping rule: a pathological subject will have more. The variability of the gait calculated with the mean and the standard deviation of all the squares of the table while keeping or not the steps of U-turns or initiation of the gait.

[0129] This makes the visualization matrix a useful tool for measuring long-distance walking, which will show whether a subject is using two different walking regimes, which is an interesting clinical indicator (pain awakening in osteoarthritis or "freezing" in Parkinson's disease, for example).

[0130] Walking is a pseudoperiodic activity that is naturally subdivided into cycles modulated by physiological rhythm changes in walking (i.e., initiation, turnaround and stopping) or pathological irregularities in established walking. In a walking exercise, there are walking cycles of different nature such as walking initiation cycles (the first 4 cycles), established walking cycles, turnaround preparation cycles, turnaround cycles and stopping walking cycles. Based on this property, the visualization matrix allows to represent with the same process all the times of walking. This makes the visualization matrix a suitable tool for the representation of ambulatory walking which involves all the phases of walking.

[0131] The main application of the Visualization Matrix is ​​the visualization of a walking exercise. The Visualization Matrix can be used on any stepping exercise, including the 10-meter round trip as described above, the Time Up and Go test, and also on treadmills and ambulatory walking.

[0132] The visualization matrix is ​​suitable for visualizing a 10-meter round-trip walking exercise from a standing start, measured with the acceleration standard of an accelerometer placed on the dorsal surface of the foot, comparing gait cycles using Pearson's correlation distance. This protocol is suitable for routine clinical consultation and allows a synthetic representation for direct visualization in the clinic. Table 1: Inter-group comparison of P1, P2 and P3 for CJ, CA and MP. Settings CJ THAT MP P1 Average 0,96 0,94 0,88 I-III DS 0,02 0,04 0,05 II-III P2 Average 0,03 0,04 0,09 I-III DS 0,02 0,04 0,07 II-III P3 Average 5,09 5,08 6,71 I-III DS 2,30 2,59 2,72 II-III Statistical test: ANOVA (analysis of variance). II-III< = p-value < 0.05. DS standard deviation CJ (young controls) group of young adults (N=9; age = mean 38.3 SD 8.2 years) CA (elderly controls) group of older adults (N=11; age = mean 67.0 SD 10.1 years) PD (Parkinson's disease) group of patients with Parkinson's disease (N=20; age = mean 74.7 SD 11.0 years) P1 average parameter of the locogram on the established walk (unitless) P2 parameter standard deviation of the locogram on the established step (unitless) P3 parameter number of hierarchical clusters on the locomogram obtained with the same stopping rule (unitless) Table 2: Intra-group MP correlation between {P1, P2, P3} and {UPDRS III, walking speed, P0} Setting UPDRS III Walking speed P0 P1 Rho -0,51 * 0,67* 0,30* Pval 0,00 0,00 0,00 P2 Rho 0,39* -0,44* -0,16 Pval 0,00 0,00 0,06 P3 Rho 0,23* -0,47* -0,27* Pval 0,00 0,00 0,00 Correlations assessed with Pearson's coefficient. * = p-value < 0.05. Rho correlation coefficient UPDRS III clinical score used to quantify the severity of motor impairment in Parkinson's disease by the neurologist during consultation. MP (Parkinson's disease) group of patients with Parkinson's disease P0 parameter standard deviation of the duration of walking cycles on the established walk P1 average parameter of the locogram on the established walk P2 parameter standard deviation of the locogram on the established step P3 parameter number of hierarchical clusters on the locogram obtained with the same stopping rule

Claims

1. A method allowing, by a visual representation: • the temporal analysis of the regularity and symmetry of a sequence of N cycles of walking or running of the same person, each of said cycles being separated into two phases of which one is a support phase where the foot of said person is in contact with the ground and the other is an oscillating or pendulum phase where the heel of the person is in the air, by comparing each cycle considered individually with each of the other cycles considered individually, • and determining the presence and the number of erratic cycles as well as the number of cycles required to achieve the walking or running regimes, and at what moment in the sequence these erratic cycles and walking regimes are reached, including the following steps: • a step (i) of measuring raw inertial time signals or time signals relating to a physical displacement variable of at least one anatomical segment of the person, • a processing and calculation step, subdivided into: • a step (ii) of processing and separating the raw time signals into distinct time signals Ci in order to break down the walking or running into walking or running cycles, each distinct time signal Ci being a time series of points of the measured physical variable and having a given shape, amplitude and duration, the series Ci being associated with a given walking or running cycle i of the person, the separation of the raw time signals into distinct time signals Ci being implemented by means of a marking, in the raw time signals, of the time points when the heel of the person touches the ground; • a step (iii) of calculating at least one coefficient of similarity between the signal Ci associated with the walking or running cycle i, with another signal Cj associated with a walking or running cycle j of the same person, the coefficient of similarity being: a coefficient of similarity of the form fij of the two signals Ci and Cj which is independent of the amplitude and the duration, with fij=fji, or a coefficient of similarity of the amplitude aij of the two signals Ci and Cj which is independent of the form and the duration, with aij=aji, or a coefficient of similarity of the duration dij of the two signals Ci and Cj which is independent of the form and the duration, with dij=dji, a step (iv) of storing into a square matrix M(i,j), the value of the coefficient of similarity fij, aij or dij in the row i and in the column j; with i and j natural integers varying from 1 to N, the N walking or running cycles being chronologically ordered according to their order in the walking or running sequence, • a step (v) of displaying the square matrix M(i,j) with the values of the coefficient of similarity fij, aij or dij represented in the cells (i,j) of the square matrix M(i,j), by visual representation of the value of the coefficient of similarity fij, aij or dij located in an interval between two extreme values, continuously with no thresholds, to allow the visual and simultaneous determination of: • the similarity between all of the walking or running cycles i and j of the same person, by comparing each cycle considered individually with each of the other cycles considered individually, • the number of erratic cycles and the number of cycles required to achieve the regimes of the walking or running sequence, and at what moment in the sequence these erratic cycles and regimes are reached.

2. The method according to the preceding claim, characterized in that, in the processing and calculation step, the coefficient of similarity is determined between the signal Ci relating to the walking or running of the right foot (respectively of the left foot) of the person, and the signal Cj relating to the walking or running of the right foot (respectively of the left foot) of the same person, and in that N=Nd (respectively Ng) which is the number of cycles of the right foot (respectively the number of cycles of the left foot).

3. The method according to the preceding claim 2, characterized in that, in the processing and calculation step, the coefficient of similarity between the signal Ci relating to the walking or running of the right foot of the person, and the signal Cj relating to the walking or running of the left foot of the same person, is also determined, and in that the number of cycles N is equal to the sum of the number of cycles of the right foot and the number of cycles of the left foot: N=Nd+Ng.

4. The method according to one of the preceding claims, characterized in that the determined walking or running regimes are: one or more established regime(s), initiation of the walking or running sequence, or a half-turn.

5. The method according to one of the preceding claims, characterized in that, in the display step (v), each value of the coefficient of similarity is represented by a color positioned in a continuous graduated scale corresponding to the scale of the coefficients of similarity.

6. The method according to the preceding claim 5, characterized in that: • in the processing and calculation step, the three different coefficients of similarity fij, aij, dij are calculated, each coefficient of similarity fij, aij, dij being associated with the same color scale as that of the other coefficients of similarity, each color having the same meaning of similarity regardless of the coefficient of similarity used; • in the display step (v), the values of the coefficients of similarity fij, aij, dij are presented in different square matrices (F(i,j), A(i,j), D(i,j)); the coefficients of similarity being selected: - so that, once calculated, all of the values of the coefficients of similarity fij, aij, dij are in the same interval [a; b] and - so that the higher the values of the coefficients fij, aij, dij, the more similar the signals Ci and Cj will be.

7. The method according to one of the preceding claims, characterized in that the coefficient of similarity is a form coefficient fij, and in that it is, in the processing and calculation step: - normalize in duration each time signal Ci with each time signal Cj so that the two signals Ci and Cj have the same duration; - normalize in amplitude each time signal Ci; - calculate a coefficient of similarity of the form fij between each normalized signal C'i, with another normalized signal C'j, and that being so for all of the walking or running cycles.

8. The method according to the preceding claim, characterized in that the coefficient of similarity of the form fij is the Pearson coefficient, the values of the coefficient of similarity being reduced to 0 in the case of a Pearson coefficient value lower than zero.

9. The method according to one of the preceding claims, characterized in that - the coefficient of similarity is a coefficient of similarity of duration dij, - the processing and calculation unit (3) is arranged for » calculating a duration Di, of the signal Ci and of the duration Dj of the signal Cj, • calculating the coefficient of similarity of duration dij which is the ratio of the shortest of the two durations Di, Dj divided by the longest duration Di, Dj.

10. The method according to one of the preceding claims, characterized in that - the coefficient of similarity is a coefficient of similarity of amplitude aij, - and, the processing and calculation unit (3) is arranged for • calculating an amplitude Ai, Aj of the signal Ci and of the signal Cj, • calculating the coefficient of similarity of amplitude aij which is the ratio of the smallest amplitude Ai, Aj divided by the largest amplitude Ai, Aj.

11. The method according to the preceding claim, characterized in that the amplitude Ai, Aj is the standard deviation or the effective value.

12. The method according to one of the preceding claims, characterized in that the following is calculated: - parameters allowing assessing the regularity of the walking or running; - parameters allowing assessing the symmetry of the walking or running; - the number of cycles required for a walking or running sequence or sub-sequence.

13. The method according to the preceding claim, characterized in that, to determine the regularity of the walking or running, the following is calculated: - the average of all cells (i, j) of the square matrix M(i,j) excluding the cells i=j, - the standard deviation of all cells (i, j) of the square matrix M(i,j) excluding the cells i=j, the average of the cells i ranging from 1 to Nd (or from 1 to Ng), and j ranging from Nd+1 to Nd+Ng (or j ranging from Ng+1 to Nd+Ng).

14. The method according to one of the preceding claims, characterized in that the measured physical variable is selected from among the following list: the acceleration norm, the acceleration norm without the terrestrial gravity component, the speed, the angular speed, the displacement, the position or a force exerted by the person in motion on a device.

15. The method according to one of the preceding claims 1 to 14, wherein the raw time signals are processed and separated automatically or manually prior to the calculation of the coefficient of similarity automatically or manually in the calculation step (iii).

16. A device (1) allowing, by a visual representation: • the temporal analysis of the regularity and symmetry of a sequence of N walking or running cycles of the same person, each of said cycles being separated into two phases of which one is a support phase where the foot of said person is in contact with the ground and the other is an oscillating or pendulum phase where the heel of the person is in the air, by comparing each cycle considered individually with each of the other cycles considered individually, • and determining the presence and the number of erratic cycles as well as the number of cycles required to achieve the regimes of the walking or running sequence, and at what moment in the sequence these erratic cycles and regimes are reached, the device (1) comprising: • sensors (2) for measuring raw inertial time signals or time signals relating to a physical displacement variable of at least one anatomical segment of the person, • a processing and computing unit (3), connected to the measurement sensors (2), arranged for: ∘ processing and separating the raw time signals into distinct time signals Ci, each time signal Ci being a time series of points of the measured physical variable and having a given form, amplitude, and duration, the series Ci being associated with a given walking or running cycle i of the person, the separation of the raw time signals into distinct time signals Ci being implemented by marking, in the raw time signals, the time points when the heel of the person touches the ground; ∘ calculating at least one coefficient of similarity between the signal Ci associated with the walking or running cycle i, with another signal Cj associated with a walking or running cycle j of the same person, the coefficient of similarity being: ▪ a coefficient of similarity of the form fij of the two signals Ci and Cj which is independent of the amplitude and the duration, with fij=fji, or ▪ a coefficient of similarity of the amplitude aij of the two signals Ci and Cj which is independent of the form and the duration, with aij=aji, or ▪ a coefficient of similarity of the duration dij of the two signals Ci and Cj which is independent of the amplitude and the form, with dij=dji; ∘ storing in a square matrix M(i,j), the value of the coefficient of similarity fij, aij or dij in the row i and in the column j, with i and j natural integers varying from 1 to N, the N walking or running cycles being chronologically ordered according to their order in the walking or running sequence; • display means (4) connected to the processing and calculation unit (3), and displaying the square matrix M(i,j) with the values of the coefficient of similarity represented in the cells (i,j) of the square matrix M(i,j), by visual representation of the value of the coefficient of similarity fij, aij or dij located in an interval between two extreme values, continuously with no thresholds, to allow simultaneous visualization of: ∘ the similarity between all of the walking or running cycles i and j of the same person, ∘ the presence and the number of erratic cycles and the number of cycles required to achieve the regimes of the walking or running sequence, and at what moment in the sequence these erratic cycles and regimes are reached.

17. The device according to the preceding claim 16, characterized in that the processing and calculation means determine the coefficient of similarity between the signal Ci relating to the walking or running of the right foot (respectively of the left foot) of the person, and the signal Cj relating to the walking or running of the right foot (respectively of the left foot) of the same person, and in that N=Nd (respectively Ng) which is the number of cycles of the right foot (respectively the number of cycles of the left foot).

18. The device according to the preceding claim 17, characterized in that the processing and calculation means also determine the coefficient of similarity between the signal Ci relating to the walking or running of the right foot of the person, and the signal Cj relating to the walking or running of the left foot of the same person, and in that the number of cycles N is equal to the sum of the number of cycles of the right foot and the number of cycles of the left foot: N=Nd+Ng.

19. The device according to one of the preceding claims 15 to 18, characterized in that the determined walking or running regimes are: one or more established regime(s), initiation of the walking or running sequence, or a half-turn.

20. The device according to one of the preceding claims 15 to 19, characterized in that, in the display step (iv), each value of the coefficient of similarity is represented by a color positioned in a continuous graduated scale corresponding to the scale of the coefficients of similarity.

21. The device according to one of the preceding claims 15 to 20, characterized in that the measurement sensors (2) are: accelerometer, gyroscope, electromyography, pressure insoles, infrared kinematic acquisition device or force platform, and in that the measured physical variable is selected from among the following list: the acceleration norm, the acceleration norm without the terrestrial gravity component, the velocity, the angular velocity, the displacement, the position or a force exerted by the person in motion on a device.