• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The present invention relates to a method for analyzing a periodic movement. The method comprises receiving (210) time series data from an inertial sensor used for sensing the periodic movement. The method further comprises partitioning (220) the received time series data into partitions, each partition corresponding to a period of the movement, and transforming (230) each partition into a data representation of a defined size. The method also comprises analyzing (240) the data representation of the periodic movement.
    公开号:SE1250065A1
    申请号:SE1250065
    申请日:2012-01-31
    公开日:2013-08-01
    发明作者:Christer Norstroem；Anders Holst
    申请人:Sics Swedish Ict Ab；
    IPC主号:
    专利说明:

    METHOD ÅSND NODE ÉOR ANALYSING PERIODIC MOVENIENTSTECHNICAL FIELDThe disclosure relates to analysis of movements. More particularly, the disclosurerelates to a system and a method for analyzing periodic movements of e.g. across-country skier.
    BACKGROUNDMany professional athletes, but also joggers or persons performing other physicalexercises use pulse watches as a training aid today. The pulse watches may alsoincorporate positioning systems for collecting more sophisticated informationduring exercise. However, the pulse watches still provide a limited amount ofinformation, and does not provide any information about how a physical exerciseor movement is performed. Different tools based on video analysis are used inmany sports for post hoc analysis of movements and physical exercises. Suchtools may comprise everything from simple systems based on consumer cameraswith graphlc annotation interfaces, to considerably more complex and expensivemotion-capture systems for more precise three dimensional modeling ofmovements. A drawback of such tools is that the feedback is provided only afterthe exercise, and that it may be difficult to observe small differences betweenmovements performed by the athlete. Furthermore, the tools based on videoanalysis are difficult to use in a sport such as cross-country skiing, as the athlete ismoving over a large area during the exercise, and it may be difficult to captureevery movement with a camera.
    Another type of tools for analysis of an athlete's movements is based on the use ofsensors placed on »different parts of the athlete's body, where the sensors typicallycommunicate wirelessly with a computer. The sensors may capture data whichmay be difficult for a trainer to see with his own eyes. The computer may thuscapture data fromlthe sensors on the athlete's body during the exercise, and thecaptured data may then be used in an analysis tool on the computer to analyzethe athlete's performance and movements. The analysis can be used e.g. as atraining aid to improve the performance of the athlete. However, a great number of2sensors may be needed to get a useful result, and they may interfere with theathletes movements during the exercise. Furthermore, even if the data Capture isperformed duringthe exercise, it may still be complex and time consuming toanalyze the large amount of data that is collected and to provide feedback in real-time during the exercise.
    SUMMARYlt is therefore an object to address some of the problems outlined above, and toprovide a solution for a detailed analysis of a periodic movement of e.g. an athletein real-time, without hlndering or restricting the athletes movements. This objectand others are achieved by the method and the system according to theindependent claims, and by the embodiments according to the dependent claims.ln accordance with a first embodiment, a method for analyzing a periodicmovement is provided. The method comprises receiving time series data from aninertial sensor used for sensing the periodic movement. The method furthercomprises partitioning the received time series data into partitions, each partitioncorresponding to period of the movement, and transforming each partition into adata representation of a defined size. The method also comprises analyzing thedata representation of the periodic movement.ln accordance with a second embodiment, a system for analyzing a periodicmovement is provided. The system comprises an inertial sensor for sensing theperiodic movement. The system is adapted to receive time series data from theinertial sensor. The system is further adapted to partition the received time seriesdata into partitions,; each partition corresponding to a period of the movement, andto transform eachípartition intoa data representation of a defined size. The systemis also adapted to analyze the data representation of the periodic movement.
    An advantage of embodiments 'is that only one sensor is needed for analyzing theperiodic movement performed by e.g. an athlete, which minimizes the risk ofdisturbance of the“'athlete's movements.
    A further advantage is that the analysis may be performed in real-time, such that3the result from the analysis may be fed back to the athlete or to a trainer duringthe exercise.
    Other objects, advantages and features of embodiments will be explained in thefollowing detailed description when considered in conjunction with theaccompanying drawings and claims.
    BRIEIL: DESCRlPflÖN OF THÉ DRAWINGSFigure 1 is a schematic illustration of a system according to one embodiment oftheinvenfion.Figures 2a-b are flowcharts illustrating the method according to embodiments.
    Figures 3a-b are block diagrams schematically illustrating the system accordingto embodiments. _Figures 4a-c are plot diagrams of raw data collected from three different skiersusing skating gear 2.
    Figures 5a-c areïplot diagrams of raw data collected from three different skiersusing skating gearlß.DVETAILED DESCRIPTIONin the following, different aspects will be described in more detail with referencesto certain embodiments and to accompanying drawings. For purposes ofexplanation and not limitation, specific details are set forth, such as particularscenarios and techniques, in order to provide a thorough understanding of thedifferent embodiments. However, other embodiments that depart from thesespecific details may also exist.lVloreover, those skilled in the art will appreciate that the functions and meansexplained herein' below may be implemented using software functioning inconjunction with a programmed microprocessor or general purpose computer,and/or; using anapplication specific integrated circuit (ASlC). lt will also beappreciated that while the embodiments are primarily described in the form of a4method and a system, they may also be embodied in a computer program productas well as in a system comprising a computer processor and a memory coupled tothe processor, wherein the memory is encoded with one or more programs thatmay perform the functions disclosed herein.
    Embodiments are described in a non-limiting general context in relation to anexample scenario where the periodic movement is a movement of a person that iscross-country skiing using a ski skating technique, and with a system according tothe one illustratedin Figure 1; However, lt should be noted that the embodimentsmay also be applied to other types of periodic movements, such as the periodicmovement of a trotting horse, of a person jogging, or of a person cross-countryskiing using a classic technique. lt may also be a periodic movement of a non-living object, such as the pedals of a bicycle, or moving parts of a robot.
    The object of analyzing a periodic movement of a cross-country skier in real~timewithout hindering the skiers movements has been solved by a method wherein aninertlal sensor, such as an accelerometer or a gyroscope, is placed on the body ofthe skier to sensethe periodical ski skating movement of the skler. Streaming datafrom the sensor is then collected and partitioned into the periods of the skier'speriodical movement. Each partition is transformed into a data representation of adefined size, in order to represent each partition in a normalized way. Thepartitioning and the transformation may be referred to as a pre-processing of thetime series data from the sensor. The advantage of the pre-processing is that itallows for a data representation in real-time of the periodical movement based onstreaming data front an inertlal sensor. The data representation may then be usedfor a subsequentíanalysis of” the periodic movement., The analysis may e.g.comprise a Classification of the periodic movement, which may thus give as aresult the class of the movement that the skier is performing during a time period.ln the example of ski-skating, different limb-movement patterns are used fordifferent terrain and speedsnA skier may thus e.g. be using different skatingtechniques depending on if~.- she is skiing uphill or downhill. There is nointernationally accepted naming convention for these different skating techniques.
    The following aresome examples of the naming of five well defined skating5techniques, followed by a brief description of the technique:1. Gear 1, diagonal V, or single-poling: Similar to the classic herringbone butwith a short glide on eachski. This technique is used in very steep uphill.2. Gear 2, V1, _or offset skate: Slightly off-set double-pole on every other leg.3. Gear 3, V2, or 1-skate: Double-pole on every leg. Used on the flat foráacceleratingand on moderate uphill.4. Gear 4, V2 alternate, or 2-Skate: Double-pole on every other leg. Used onthe flat, while climbing and on gentle downhill.5. Gear 5, V skating, or free-skate: Skating without using the poles. Used ondownhill at very high speed.ln the following the naming convention gear 1-5 will be used. Each defined skatinggear may correspond to a movement class, and the periodic movement of theskier may thus be classified into one of the defined skating gears in the analysis.
    Figure 1 illustrates a system 1.0 according to one embodiment of the invention. Aunit 100 comprising the inertial sensor 105 connected to a mobile terminal 101, isplaced on the skier. The unit 100 may e.g. be placed on the chest or on the backof the skier. The mobile terminal 101 comprises a processing circuit 102, and atransmitter 103 for transmitting sensor data to an internet server 110 via a mobilenetwork. The connection between the sensor 105 and the mobile terminal 101may be a Bluetooth connection, thus making it possible to place the sensor at adistance from thefmobile terminal. Furthermore, more than one sensor 105 maybe connected to the terminal"101. For example both an accelerometer and agyroscope may be connected and synchronized with each other. Alternatively twoaccelerometers placed at different parts of the body may be connected. Today'ssmart phones have integratedaccelerometers and could thus be used as the unit100 placed on theskier providing a combined sensor 105 and mobile terminal101. The server 110 may also' be connected to a database (DB) 120 used e.g. forstoring data collected from the inertial sensor in the form of raw data or pre-6processed data, or for storing results from the analysis of the movement.
    Time series data» received from the sensor 105 may be pre-processed andanalyzed locally inthe unit 100 carried by the skier, e.g. in the processing circuit102 of the terminal 101. Time series data may be referred to as a sequence ofdata received from the sensor-105 over time. An advantage of this embodiment oflocally analyzed data is that there is no need for the connection to the internetserver 110, so the system is self-contained. Alternatively, the raw time series datareceived from the sensor 105 may be transmitted to the server 110, and the pre-processing and analysis may be performed in the server 110. ln the case when noprocessing of the data is performed in the unit 100 carried by the skier, the unit100 will only need to comprise the sensor 105 and the transmitter 103 adapted tofonNard the raw sensor data to the server 110. An advantage of this embodimentis that the unit 100 carried by the skier is reduced to a minimum. A combination isalso possible, wherein the pre-processing of the time series data is performed inthe unit 100 carried by the skier, and wherein the pre-processed data istransmitted to the server 110 for further analysis. An advantage of thisembodiment is that the amount of data transmitted between the unit 100 on theskier and the server 110 is reduced compared to if raw sensor data is transmittedto the server 110. iFurthermore, a client or user terminal 130 may be connected to the internet server110 for visualizing the results from the analysis. Also the visualization could bedone in real-time. »The client may be connected wirelessly, and used e.g. by a skitrainer following the skier's performance on site by visualizing the analysis result inthe client user interface. Correspondingly, the unit 100 carried by the skier mayalso provide a user interface for visualizing the results from the movementanalysis in real-time during the exercise. lt could also be envisaged to use a soundinterface for providing feed-back sounds to the skier based on the analysis. ln thisway the skier mayulzget feed-back in a more convenient way while skiing.
    Figure 2a is a flowchart illustrating the method for analyzing a periodic movementaccording to an embodiment of the invention. The periodic movement may e.g. beperformed by a human or an animal, typically when performing a sport. ln one7embodiment, the periodic movement is a movement of a cross-country skiingperson. The method comprises:o 210: Receiving time series data from an lnertial sensor used for senslng theperiodic movement. The sensor may be an accelerometer or a gyroscope.An accelerometer may provide data from the sensed movement in differentdirections or axes. A three dimensional accelerometer would e.g. providedata in three perpendicular directions: The lateral movement direction, thevertical movement direction, and the horizontal backward and forwardmovement direction. The received time series data is thus threedimensional, and may provide support for a rather detailed movementanalysis. However, a one dimensional accelerometer may be enough forthe purpose of the invention, depending on the type of periodic movementthat is to be__ analyzed. lf the movement is primarily in the vertical direction,_a one dimensional accelerometer may be enough for a good analysis. Thesampling rate of the sensor has to be high enough to allow for a usefulanalysis. ln' one example embodiment suitable for the analysis of a skiskating movement, the sampling rate is in the order of one sample every 10m8.220: »Partitiorning the received time series data into partitions, each partitioncorresponding to a period of the movement. The purpose of the partitioningis thus to get a data representation of the sensor data for each period of themovement, such that each movement period may be analyzed. For ski-skating, a period is typically of the length of O.8-1.6 seconds.230: Transforming each partition into a data representation of a definedsize. This step is together with the partitioning of the data into periods alsoreferred toas the pre-processing of the sensor data, crucial for thepossibility 'to do a Classification or any other analysis of the datarepresentation. Furthermore, the pre-processing may be performed in real-time basedon streaming data.240: Analyzing the data representation of the periodic movement.8Figure 2b is a flowchart illustrating the method for analyzing the periodicmovement according to another embodiment of the invention. The methodcomprises:~ 205: Building up a statistical model based on the data representationderived from pre-collected time series data associated with at least onemovement class. The statistical model is used for the Classification of themovement. By pre-collecting time series data from skier's when they areperforming a certain type of skating technique, the model may be built up inbeforehand. ln one embodiment, wherein the pre-processing and theanalysis is performed in the unit 100 carried by the skier, the model may bedownloaded from a server 110 in order to perform the Classification.However, another possible embodiment is that the statistical model is builtup based on the skier's own movements. lt may be enough to build themodel on only ten periods from a periodic movement of a certain movementclass. The model may then be used in the unit 100 for further analysis ofthe skier's movements. ln this scenario, the step of building up the modelmay be performed after, the step 210 of receiving time series data from thesensor. The pre-processing of the sensor data is thus also useful for thebuild-up of a' powerful statistical model.210: Receiving time series data from an inertial sensor used for sensing theperiodic movement. This corresponds to step 210 described above withreference to Figure 2a.220: The partitioning of the received time series data into partitions,comprises the following steps:- 221: Low-pass filtering the received time series data to identify limitsbetween each period of the movement. By low-pass filtering the datareceived with a low cut-off frequency, the periods of the movementwill stand out clearer from the filtered data. This allows to pin-pointwhen in time one movement period stops and the next one beginsbased on the data. ln alternative example embodiments, the limits9between the periods may be identified as the point in time when thefiltered data has a peak, or when it passes zero in one direction(upward or downwards). in the example of ski skating, a period istypically of a length of O.8-1.6 seconds as already mentioned above,and a low enough cut-off frequency for the low-pass filtering maythen be 1-2 Hz.- 222: Partitioning the data according to the identified limits. As the limitsbetween the periods have been identified in step 221, the receivedsensor data may be cut up or partitioned into periods accordingly.~ 230: The transforming of each partition comprises the following steps:-231: Low-pass filtering data of each partition to reduce noise. ln thisway, the trajectory of the movement will be accentuated, and onlyrelevant information is kept. Typically a higher cut-off frequency isused 'than when identifying the periods. ln the example of ski skatingwithperiods slower than 1 Hz, a suitable cut-off frequency for thelow-pass filteringïrmight be around 5 Hz.-232: Re-sampling the low-pass filtered data to form a datarepresentation of a defined size. As a time period of the skier'smovement is not constant, due to Variations in the skier'sperformance e.g. because of the steepness of a ski track slope, thedata corresponding to a period needs to be re-sampled in order toget a data representation of a defined size. Another reason for there-sampling is that the sensor's sampling rate may present avariation over time such that the time between samples may beslightly differing.,o 240: The analyzing ofthe data representation of the periodic movement,comprises the followingristeps:-2411 Classifying the periodic movement into a movement class basedon the statisticalmodel of at least one movement class. lf only onemovement class is modeled, a statistical anomaly detection methodcan fbe used to determine whether a movement period belongs tothat class. With one movement class for each type of movement, astatistical machine learning method can be used to classify themovement period,--2422 Determining a parameter characterizing the periodic movementbased on the data representation. The parameter may comprise alevel- of symmetryof the movement, a variability of the movement, oran amplitude of the movement. For a skating skier, it may be veryimportant to know if she is making the skating movement in anasymmetric way, as that may affect her performance. The level ofsymmetry may therefore be interesting. A certain amount ofvariability of the movement is often wanted, as a very staticmovement may give problems with lactic acid in the muscles.However, the variability should be controlled, and an analysisproviding information regarding the variability is thereforeadvantageous. A simple analysis of the amplitude of a movementmaybe very interesting in order to find a relation between amovement amplitude and performance. The Classification of themovement may be needed before determining a parametercharacterizing the movement. A parameter such as the variability ofthe movement is e.g. coupled to the movement class, while theamplitude and the level of symmetry of the movement may bedetermined regardless of the movement class. ln alternativeembodiments, the steps of classifying in 241 and of determining acharacterizing parameter in 242 may be performed 'uncombined ltmay ~e.g. be interesting to perform the analysis of the symmetry leveldirectly in the unit 100 carried by the skier, in order to feed-back thatanalysis result immediately to the skier. The Classification analysismay-on the other hand be performed on the server based on pre-processed data transmitted from the unit 100 on the skier, as theClassification requires the statistical model(s) which are typically11stored on the server 110, or on the database 120 connected to theSefVèf.ø 250: Storing at least one of the received time series data, the datarepresentation, or information related to the analysis of the periodicmovement. Typically, the storing will be done in a database 120 connectedto a server 110. By storing the raw data, any analysis is possible to performin the future. However, 'storing only the pre-processed data representation-diminishes the requirement on storing capacity. lt may also be interesting tostore the results from the analysis, such as a level of symmetry in order totrack performance changes for a certain skier. Classification results for acertain skier and a certain ski track may also be useful to compare the styleof different skiers, or of a same skier over time.
    There are several 'alternative embodiments possible for the transformation 230 ofa partition. Alternatives may be to use a Fourier transform method or a PrincipalComponent Analysis (PCA) method. The embodiment described above withreference to steps: 231 and 2132 has been proved to be suitable for a ski-skatingperiodic movement.ln embodiments ofthe invention, the statistical model comprises a Markov chain ofmultivariate Gaussian distributions for each of the at least one movement class. lnmore detail this means that the' probability distribution over the data representationX of a movement period, consisting of k*n elements where k is the number ofsensors and n the number of 'time steps in each period after the transformation,comprising e.g. the filtering and the re-sampling, is for each movement class Cmodeled as:PtXlC) = P(folC)H =1P'(frlf¿~1,C) [11where each i,- contains the ksensor values corresponding to the izth time step,and each factor P(a2,-l5¿,-_1, C) is represented by a 2*k dimensional multivariateGaussian distribution. The parameters of the Gaussian distributions are estimatedfrom previously collected sensor data from each movement class.12When the statistical model comprising more than one movement class is used forClassification, Bayes theorem may be used to calculate the probability that a datarepresentation X of a movement period belongs to a certain movement class C.
    The probability is calculated for each movement class C using:Pccixï' °< Pcxioißtc) [21where P(C) is the prior probability of each class which is typically assumed to bethe same for allclasses, and oc stands for ”is proportional to”. The movementclass with the highest probability is then selected.
    When the statistical model comprises only one class, the data representation X ofthe movement period is said to belong to the class of the model when:P(X|C)3 > s [3]where s is a suitably selected threshold.
    An advantage of using a »statistical model comprising a Markov chain ofmultivariate Gaussian distributions is that the model is robust, as the Markov chainallows to break up the very high dimensional input space into a number of smallerdimensional input spaces, and each Gaussian distribution is not as sensitive tonoise and deviating values as a more complex non-linear model may be.Furthermore, the build-up of the model, as well as the Classification based on themodel, is computationally efficient. This means that it is possible to analyze datafrom a three-dimensional accelerometer without problems with regards tocomputation performance.
    Figures 4a-c and figures 5a-c illustrate plot diagrams produced from raw datareceived from a three-dimensional accelerometer fixed on the chest of skiersperforming the periodical movement of a defined ski skating gear. Figures 4a-cillustrate the plotldiagrams for three different skiers (one skier per figure) usingskating gear 2, and figures Sa-c illustrate the plot diagrams for the three differentskiers using skating gear 3. ln each figure, the left most plot diagram shows theacceleration of the lateral movement on the x-axis of the diagram, and the13acceleration of the vertical movement on the y-axis of the diagram. The middleplot shows the acceleration of the lateral movement on the x-axis, and theacceleration of the 'horizontal backward and forward movement on the y-axis. Theright most plot diagram shows 'the acceleration of the vertical movement on the x-axis, and the acceleration of the horizontal movement on the y-axis. Saidmovement directions, i.e. thelateral, the vertical, and the horizontal movementdirections are perpendicular to' each other.ln any of the previously described embodiments, the method may comprisevisualizing a result from analyzing the data representation of the periodicmovement in a user interface. Either the skier herself, or the trainer following theskier's performance, may want to visualize the results from the analysis during theexercise. The result from the analysis may e.g. be visualized in a graphical webuser interface. External analysis programs such as Excel or Mathlab may also beused to view the results.ln one use case of the present invention, the skier wants to view and store whatskating technique she uses at a given time during an exercise. She fastens theunit 100 on her chest with an elastic ribbon, Which may e.g. be a smart phonecomprising an accelerometer, and an application adapted to receive the sensordata and to perform the pre~processing and movement analysis. The application isstarted and the skier carries out the skating exercise in the track. When theexercise is finished, the resultmay be viewed on the smart phones display, e.g. ina graphical view. The result may also be viewed by the skier's trainer by accessingthe result via the internet server. More detailed analysis may also be performed onthe server, either in real-time, or at a later stage.
    An embodiment :of a system 30 for analyzing a periodic movement isschematically illustrated in the block diagram in Figure 3a. The system 30comprises an inertial sensor 305a for sensing the periodic movement. The sensormay be an accelerometer and/or a gyroscope. The system is adapted to receivetime series data from the inertial sensor 305a, and to partition the received timeseries data into partitions, each partition corresponding to a period of themovement. The system 30 is further configured to transform each partition into a'l4data representation of a defined size, and to analyze the data representation ofthe periodic movement. The system may in this embodiment correspond to theunit 100 in Figure 1 that the skier is carrying. The system 30 may thus comprise aprocessing circuit'302 configured to receive the sensor data, perform the pre-processing of theidata, and analyze the preprocessed data. The system mayoptionally comprise a communication interface 304 such as a Bluetooth interfacefor communicating with an additional sensor node 305b, as well as a transmitter303 connected to one or more antennas 308 for transmitting sensor data oranalysis results to e.g. an external sen/er. The system may in embodiments alsocomprise a receiver for recelving data from the external server.ln an alternative way to describe the embodiment in Figure 3a, the system 30comprises a Central Processing Unit (CPU) which may be a single unit or aplurality of units. Furthermore, thesystem comprises at least one computerprogram product (CPP) in the form of a non-volatile memory, e.g. an EEPROlVl(Electrically Erasable Programmable Read-Only Memory), a flash memory or adisk drive. The CPP comprises a computer program, which comprises codemeans which when run on the system 30 causes the CPU to perform steps of theprocedure described earlier in: conjunction with Figure 2a. ln other words, whensaid code means :are run on the CPU, they correspond to the processing circuit302 of Figure 3a. iAnother embodiment of the system 30 is schematically illustrated in the blockdiagram in Figure 3b. The system may in this embodiment correspond to thewhole of system in Figure The system 30 comprises an inertial sensor 305afor sensing the periodic movement, comprised in the unit 301 carried by the skierperforming the periodic movement. The unit 301 also comprises a communicationinterface 304 for communicating with an additional sensor node 305b, The system30 further comprises a server310, and the unit 301 comprises a transmitter 303connected to an ajntenna 308,! for transmitting sensor data, data representations,and/or analysis results to theuserver 310, eg. via a mobile network. The systemmay in embodiments also comprise a receiver for recelving data, such as thedownloaded statistical model from the external server. A processing unit 302 onthe unit 301 is adapted to receive time series data from the inertial sensor 305aand possibly also from an external sensor 305b connected via the communicationinterface 304. Thereceived sensor data may in one embodiment be transmitted tothe server 310. A processing circuit 312 on the server 310 is adapted to partitionthe received time series data into partitions, to transform each partition into a datarepresentation of _-a defined size, and to analyze the data representation of theperiodic movement. However, as already described previously, in an alternativeembodiment, the pre-processing of the sensor data may be performed by theprocessing circuit 302 in the unit 301, and the pre-processed data may betransmitted to the server 310 for analysis. The system is in embodiments furtheradapted to store at least one of the received time series data, the datarepresentation, or information related to the analysis of the periodic movement.The system 30 may thus further comprise a database 320 connected to the server310, adapted to store sensor data or analysis results. The database 320 may inembodiments be integrated with the server 310.ln any of the embodiments described above with reference to Figures 3a-b, thesystem 30, may be adapted to analyze the data representation of the periodicmovement by classifying the periodic movement into a movement class based ona statistical model of at least one movement class. The statistical model maycomprise a Markov chain of multivariate Gaussian distributions for each of the atleast one movement class. The system may be further adapted to build up thestatistical model based on the _data representation derived from pre-collected timeseries data associated with the at least one movement class.ln one embodiment, the system is adapted to analyze the data representation ofthe periodic movement by determining a parameter characterizing the periodicmovement based on the data representation. The parameter may comprise a levelof symmetry of the movement, a variability of the movement, and/or an amplitudeof the movement.ln a further embodiment, the system 30 is further adapted to visualize the resultfrom analyzing the data repfesentation of the periodic movement in a userinterface. ln the embodiment described in Figure 3b, the system 30 may also16comprise a client 330 connected to the server, to reaiize the visuaiization.ln anyof the embodiments described above with reference to Figures 3a-b, thesystem 30, may be adapted to partition the received time series data by low-passfilterind the received time series data to identify limits between each period of themovement, and partitioning the data according to the identified limits. Furthermore,the system may be adapted to transform each partition by low-pass filtering dataof each partition to reduce noise, and re-sampling the low-pass filtered data toform a data representation of a defined size. The system may be adapted toperform the partitioning and the transforming in real-time.
    The units and circuits described above with reference to Figures 3a-b may belogical units, separate physical units or a combination of both logical and physicalunits.
    The above mentioned and described embodiments are only given as examplesand should not be limiting. Other solutions, uses, objectives, and functions withinthe scope of the accompanying patent claims may be possible.
    权利要求:
    Claims (25)
    [1] . A method for anaiyzing a periodic movement, the method comprising: - receiving (210) time series data from an inertial sensor used for sensing theperiodic movement, - partitioning (220) the received time series data into partitions, each partitioncorresponding to a period of the movement, - transforming (230) each partition into a data representation of a definedsize, and - anaiyzing (240) the data representation of the periodic movement.
    [2] . The method according to claim 1, wherein anaiyzing (240) the data representation of the periodic movement comprises:- classifying (241) the periodic movement into a movement class based on a statistical model of at least one movement class.
    [3] . The method according to claim 2, wherein the statistical model comprises a
    [4] Markov chain of multivariate Gaussian distributions for each of the at least one movement class.
    [5] . The method according to any of claims 2-3, further comprising: - building up (205) the statistical model based on the data representationderived from pre-collected time series data associated with the at least one movement class.
    [6] . The method according to any of the preceding claims, wherein anaiyzing (240) the data representation of the periodic movement comprises:- determining (242) a parameter characterizing the periodic movement based on the data representation.
    [7] . The method according to claim 5, wherein the parameter comprises at least one of: a level of symmetry of the movement, a variability of the movement, and an amplitude of the movement. 18 _ The method according to any of the preceding claims, further comprising visualizing a result from analyzing the data representation of the periodic movement in a user interface.
    [8] . The method according to any of the preceding claims, wherein partitioning (220) the received time series data comprises low-pass filtering (221) thereceived time series data toidentify limits between each period of the movement, andpartitioning (222) the data according to the identified limits.
    [9] . The method according to any of the preceding claims, wherein transforming (230) each partition comprises low-pass filtering (231) data of each partition toreduce noise, and re-sampling (232) the low-pass filtered data to form a data representation of a defined size.
    [10] 10.The method according to any of the preceding claims, wherein the partitioning and the transforming is performed in real-time.
    [11] 11.The method according to any of the preceding claims, further comprising storing (250) at least one of the received time series data, the data representationer information related to the analysis of the periodic movement.
    [12] 12.The method according to any of the preceding claims, wherein a human or an animal is performing the periodic movement.
    [13] 13.The method according to any of the preceding claims, wherein the periodic movement is a »movement of a cross-country skiing human.
    [14] 14.A system for analyzing a periodic movement, the system comprising an inertial sensor for sens-ing the periodic movement, wherein the system is adapted to:~ receive time series data from the inertial sensor,- partition the received time series data into partitions, each partition corresponding to a period of the movement, 19 - transform each partition into a data representation of a defined size, and - analyze theldata representation of the periodic movement.
    [15] 15.The system according to ciaim 14, wherein the system is adapted to analyzethe data representation of the periodic movement by classifying the periodicmovement into 'a movement class based on a statistical model of at least one movement class.
    [16] 16.The system according to ciaim 15, wherein the statistical model comprises aMarkov chain of multivariate Gaussian distributions for each of the at least one movement class.
    [17] 17.The system according to any of claims 15-16, wherein the system is furtheradapted to build up the statistical model based on the data representationderived from pre-collected time series data associated with the at least one movement class.
    [18] 18.The system according to any of claims 14-17, wherein the system is adapted toanalyze the data representation of the periodic movement by determining aparameter characterizing the periodic movement based on the data representation.
    [19] 19.The system according to ciaim 18, wherein the parameter comprises at leastone of: a level of symmetry of the movement, a variability of the movement, and an amplitude of the movement.
    [20] 20.The system according to any of claims 14-19, wherein the system is furtheradapted to visualize the result from analyzing the data representation of the periodic movement in a user interface.
    [21] 21.The system according to any of claims 14-20, wherein the system is adapted to partition the received time series data by low-pass filtering the received time series data to identify Iimits between each period of the movement, andpartitioning the data according to the identified limits. i
    [22] 22.The system according to any of claims 14-21, wherein the system is adapted to5 transform each _partition by Iow-pass filtering data of each partition to reducenoise, and re-sampling the low-pass filtered data to form a data representation of a defined size.
    [23] 23.The system according to any of claims 14-22, wherein the system is adapted to 10 perform the partitioning and the transforming in real-time.
    [24] 24.The system according to any of claims 14-23, wherein the system is furtheradapted to store at least one of the received time series data, the datarepresentation, or information related to the analysis of the periodic movement.
    [25] 25.The system according to any of the claims 14-24, wherein the inertial sensor is an accelerometer and/or a gyroscope.
    类似技术:
    公开号 | 公开日 | 专利标题
    SE1250065A1|2013-08-01|Method and apparatus for analysis of periodic motion
    CN105411593B|2019-12-24|Method and apparatus for identifying gait task
    Jalal et al.2019|A wrist worn acceleration based human motion analysis and classification for ambient smart home system
    Jalal et al.2019|A Triaxial acceleration-based human motion detection for ambient smart home system
    JP2018506773A5|2019-01-31|
    CN103699795B|2018-01-23|A kind of motor behavior recognition methods, device and exercise intensity monitoring system
    CN105229666A|2016-01-06|Motion analysis in 3D rendering
    CN107754225A|2018-03-06|A kind of intelligent body-building coaching system
    CN107862270A|2018-03-30|Face classification device training method, method for detecting human face and device, electronic equipment
    CN106999748B|2020-10-30|Systems, devices and methods relating to athletic data
    US11047706B2|2021-06-29|Pedometer with accelerometer and foot motion distinguishing method
    CN106123911A|2016-11-16|A kind of based on acceleration sensor with the step recording method of angular-rate sensor
    CN108245880A|2018-07-06|Body-sensing detection method for visualizing and system based on more wearing annulus sensor fusions
    CN105068657B|2019-01-15|The recognition methods of gesture and device
    CN107016411B|2020-09-29|Data processing method and device
    CN110532996A|2019-12-03|The method of visual classification, the method for information processing and server
    CN106651978A|2017-05-10|Face image prediction method and system
    US10271775B2|2019-04-30|Determining performance indicators for periodic movements
    BR102017026251A2|2019-06-25|METHOD AND SYSTEM OF RECOGNITION OF SENSOR DATA USING THE ENRICHMENT OF DATA FOR THE LEARNING PROCESS
    CN104881647B|2019-03-01|Information processing method, information processing system and information processing unit
    US20200334451A1|2020-10-22|Motion behavior pattern classification method, system and device
    CN111288986A|2020-06-16|Motion recognition method and motion recognition device
    CN109902565B|2020-05-05|Multi-feature fusion human behavior recognition method
    CN106970705A|2017-07-21|Motion capture method, device and electronic equipment
    CN111318009A|2020-06-23|Somatosensory health entertainment system based on wearable inertial sensing and working method thereof
    同族专利:
    公开号 | 公开日
    SE537695C2|2015-09-29|
    EP2624171B1|2020-09-02|
    EP2624171A2|2013-08-07|
    EP2624171A3|2014-12-10|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US8612181B2|2010-03-04|2013-12-17|Ipcomm|Wireless system for monitoring and analysis of skiing|SE537845C2|2013-07-18|2015-11-03|Wememove Ab|Method and system for determining performance indicators for periodic movements|
    US9934259B2|2013-08-15|2018-04-03|Sas Institute Inc.|In-memory time series database and processing in a distributed environment|
    US10169720B2|2014-04-17|2019-01-01|Sas Institute Inc.|Systems and methods for machine learning using classifying, clustering, and grouping time series data|
    US9892370B2|2014-06-12|2018-02-13|Sas Institute Inc.|Systems and methods for resolving over multiple hierarchies|
    US9208209B1|2014-10-02|2015-12-08|Sas Institute Inc.|Techniques for monitoring transformation techniques using control charts|
    US9418339B1|2015-01-26|2016-08-16|Sas Institute, Inc.|Systems and methods for time series analysis techniques utilizing count data sets|
    PL235133B1|2017-09-15|2020-06-01|Przemyslowy Inst Automatyki I Pomiarow Piap|Method and system for supporting the ski instructor during the students' skiing lessons|
    法律状态:
    优先权:
    申请号 | 申请日 | 专利标题
    SE1250065A|SE537695C2|2012-01-31|2012-01-31|Method and apparatus for analysis of periodic motion|SE1250065A| SE537695C2|2012-01-31|2012-01-31|Method and apparatus for analysis of periodic motion|
    EP13153032.1A| EP2624171B1|2012-01-31|2013-01-29|Method and node for analysing periodic movements|
    [返回顶部]