JP6515070B2 - Muscle activity estimation device, method and program - Google Patents

Description

  The present invention relates to a muscle activity estimation apparatus, method and program for analyzing muscle activity during exercise to estimate the state of exercise.

  In recent years, many studies have been conducted to quantitatively analyze complex movements such as sports. For example, for simple exercises, there are studies to manually find features of exercises from electromyography (EMG) based on existing knowledge. However, pure comparison of a large number of muscle activities takes enormous time, and muscle activities are too complicated to be understood by humans. In particular, it is not easy to quantitatively extract features from complex movements such as golf.

  Therefore, methods of automatically extracting features of exercise based on measurement data of electromyography have been studied. For example, by performing nonnegative matrix factorization (NMF) on electromyographic data obtained by a plurality of myoelectric sensors, a method is proposed that decomposes muscle activity into a pattern of muscles that are active together and their activity. It is done. The pattern and amount of activity of muscles working together are called muscle synergy and activation, respectively (see, for example, non-patent document 1).

VCK Cheunga, A. Turollab, M. Agostinib, S. Silvonib, C. Bennisc, P. Kasic, S. Paganonic, P. Bonatoc and E. Bizzia, “Muscle synergy patterns as physiological markers of motor cortical damage,” PNAS, vol. 109, no. 2012, pp. 14652-14656.

  However, the method described in Non-Patent Document 1 can identify the type of muscle synergy involved in exercise and the amount of activity of each muscle synergy, but is the kinematic chain (Kinematic Chain) occurring at the transition of muscle synergy? It can not be grasped whether or not exercise which works in order while multiple muscles of the body work together is performed. In addition, technology to estimate the state of exercise from muscle activity patterns has not been established yet.

  The present invention has been made in view of the above circumstances, and it is possible to grasp the existence and transition of exercise linkage of muscle synergy due to exercise, and to estimate the state of exercise further, muscle activity estimation apparatus, method and program Is to provide.

In order to solve the above-mentioned subject, the 1st mode of this invention receives the electromyography data showing the activity of a plurality of muscles from a plurality of myoelectric sensors attached to a plurality of muscles used for a subject's exercise , And means for extracting the muscle synergy and its activation characteristic of each of the plurality of muscles based on the received electromyogram data, and the extracted muscle of each muscle Means for rearranging the synergy in order of the timing at which the maximum value of the activation appears, and data representing the muscle synergy of each muscle after the rearrangement and its activation are approximated to a predetermined function, A means for obtaining at least one of the parameter of the approximated function and the information representing the degree of the approximation, and a function of the function obtained for each of a plurality of movements performed in the past At least one of a means for storing at least one of a parameter and information indicating the degree of approximation in a storage unit, and at least one of parameters stored in the storage unit and showing parameters of a function obtained for a plurality of past motions and the degree of approximation And means for estimating the state of exercise newly performed.

  According to a second aspect of the present invention, a parameter of a function determined for a newly performed motion is determined for a plurality of past motions stored in the storage unit by the means for estimating the motion state. The newly performed motion state is estimated based on the past motion state in which the parameter is similar to the parameters of the function.

  According to a third aspect of the present invention, a new means is provided for referring to information representing the degree of approximation of a function obtained for a plurality of past motions stored in the storage unit by the means for estimating the motion state. The change in information representing the degree of approximation of the function obtained for the exercise performed is calculated, and the state of the newly performed exercise is estimated based on the calculation result.

  According to a fourth aspect of the present invention, the apparatus further comprises means for outputting information representing the state of the newly performed motion estimated by the means for estimating the state of motion to an external device. It is a thing.

  According to a fifth aspect of the present invention, as information representing the estimated state of exercise newly performed by the means for outputting to the external device, information representing a subjective evaluation result of the state of the exercise; At least one of the information representing the stability of the state of exercise is output.

According to the first aspect of the present invention, when trying to extract muscle synergy from electromyographic data received from an electromyographic sensor to find a feature of exercise, the muscle synergy of each muscle extracted is Sorted in order of the timing at which the maximum value of activation appears. For this reason, it is possible to determine the presence or absence of an exercise chain from data representing muscle synergy and activation of muscles rearranged in time series. Also, data representing muscle synergy and activation of the muscle after sorting is approximated to a predetermined function using regression analysis. For this reason, it becomes possible to explain the feature of muscle activity using a function. Furthermore, the state of the newly performed motion is estimated with reference to the parameter of the function obtained for the plurality of motions in the past, or the information indicating the degree of approximation. Therefore, it is possible to objectively estimate the state of exercise newly performed from the analysis result of the muscle activity at that time.

  According to the second aspect of the present invention, the simple operation can be performed simply by comparing the newly performed state of motion with the parameter of the function that approximates the feature of muscle activity obtained in each of a plurality of past movements. It can be estimated by

  According to the third aspect of the present invention, using information representing the degree of approximation of the function approximating the muscle activity characteristic, which is obtained for each exercise, the change in the degree of approximation is newly calculated. It is possible to estimate the state of the injured movement.

  According to the fourth aspect of the present invention, the information indicating the state of exercise performed anew is output to the external device, so that, for example, in a terminal used by a portable terminal, an instructor, etc. possessed by the target user, the target The target user himself or the instructor can confirm the estimation result of the user's exercise state.

  According to the fifth aspect of the present invention, the target user, the instructor, or the like can estimate the state of motion of the target user, the information indicating the result of subjective evaluation of the state of motion, or the stability of the state of motion. The information to be represented makes it possible to confirm in a form easy to understand.

  That is, according to an aspect of the present invention, it is possible to provide a muscle activity estimation device, method and program which make it possible to grasp the existence and transition of exercise linkage of muscle synergy by exercise and to estimate the state of exercise. it can.

1 is a functional block diagram showing an embodiment of a muscle activity estimation device according to the present invention. The flowchart which shows the procedure and process content of the muscle activity analysis process and estimation process by the muscle activity estimation apparatus shown in FIG. 7 is a flowchart showing a first embodiment of the muscle activity estimation process shown in FIG. 2; 7 is a flowchart showing a second embodiment of the muscle activity estimation process shown in FIG. 2; FIG. 9 is a view for explaining an operation example of muscle activity extraction processing in the muscle activity analysis processing shown in FIG. 2; The figure for demonstrating the operation example of muscle synergy analysis processing among the muscle activity analysis processing shown in FIG. The figure which shows an example of the active mass (activation) of the muscle synergy obtained by the muscle synergy analysis process shown in FIG. The figure for demonstrating the operation example of the sort process among the muscle activity analysis processes shown in FIG. The figure for demonstrating the operation example of the fitting process among the muscle activity analysis processes shown in FIG. The figure which shows the example of the sigmoid function obtained by a fitting process. The figure for demonstrating the operation | movement of the muscle activity estimation process shown in FIG. The figure for demonstrating the operation | movement of the muscle activity estimation process shown in FIG.

Hereinafter, embodiments according to the present invention will be described with reference to the drawings.
[One embodiment]
(Constitution)
FIG. 1 is a functional block diagram showing an embodiment of a muscle activity estimation apparatus according to the present invention, in which US is a target user, 2 is a camera, 3 is the muscle activity estimation apparatus, and 4 is a target user US. The user terminals are shown respectively. The camera 2 comprises, for example, a web camera. As the user terminal 4, for example, a smartphone, a wearable terminal or a tablet type terminal is used.

  In the target user US, myoelectric sensors 11 to 1 m are attached to a plurality of muscles operating when performing a golf swing. The myoelectric sensors 11 to 1 m each measure surface myoelectric potentials representing the movement of muscles of the target user US, and use an electromyogram (also called electromyographic data) representing the measured surface myoelectric potentials using a wireless interface. To send. As a wireless interface, a wireless interface adopting a short distance wireless data communication standard such as a wireless local area network (LAN) or Bluetooth (registered trademark) is used.

  Moreover, the camera 2 is arrange | positioned at the side position of object user US. The camera 2 captures an image of the motion (golf swing) of the target user US from its side position, and transmits the captured image data as motion data using a wireless interface. A wireless interface adopting a short distance wireless data communication standard as in the case of the myoelectric sensors 11 to 1m is also used for this wireless interface.

The muscle activity estimation device 3 is, for example, a personal computer, and includes a communication unit 31, a control unit 32, and a storage unit 33.
The communication unit 31 includes a first reception unit 311, a second reception unit 312, and a transmission unit 313. The first reception unit 311 receives motion data transmitted from the camera 2. The second reception unit 312 receives the myoelectric data transmitted from the myoelectric sensors 11 to 1 m. The transmission unit 313 is used to transmit analysis data to an external terminal such as the user terminal 4 or the like and an information processing apparatus. The first and second receiving units 311 and 312 may be shared.

  The storage unit 33 uses a hard disk drive (HDD) or a solid state drive (SSD) as a storage medium, and a motion data storage unit 331 and a storage area necessary for practicing the embodiment of the present invention. EMG data storage unit 332, muscle activity data storage unit 333, muscle synergy data storage unit 334, sort data storage unit 335, fitting data storage unit 336, and estimation result storage unit 337 There is.

  The motion data storage unit 331 is used to store motion data received by the first receiving unit 311. The myoelectric data storage unit 332 is used to store the myoelectric data of each of the myoelectric sensors 11 to 1m received by the second receiving unit 312.

  The muscle activity data storage unit 333 is used to store data representing muscle activity extracted by the control unit 32 described later. The muscle synergy data storage unit 334 is used to store muscle synergy data calculated by the control unit 32 described later. The sort data storage unit 335 is used to store muscle synergy data after sort processing obtained by the control unit 32 described later.

  The fitting data storage unit 336 stores parameters corresponding to the states of a plurality of assumed swings and information representing a subjective evaluation of the swings. The fitting data storage unit 336 is also used to sequentially store data after the fitting process calculated by the control unit 32 described later. The estimation result storage unit 337 is used to store information indicating the estimation result of the motion state calculated by the control unit 32 described later.

  The control unit 32 has a processor such as a CPU (Central Processing Unit), for example, and as a processing function necessary to carry out one embodiment of the present invention, a motion data acquisition processing unit 321 and an myoelectric data acquisition processing unit A muscle activity extraction processing unit 323, a muscle synergy analysis processing unit 324, a sort processing unit 325, a fitting processing unit 326, a muscle activity estimation processing unit 327, and an output processing unit 328 are provided.

  The motion data acquisition processing unit 321 associates the motion data received by the first reception unit 311 with the identification information of the target user US in a state in which the information representing the imaging time is attached, and Perform processing to store.

  The myoelectric data acquisition processing unit 322 sets each of the myoelectric data received by the second receiving unit 312 to the identification information of the myoelectric sensor 11 to 1 m (hereinafter also referred to as a channel) of the transmission source and the measurement date and time. A process of storing the information in the myoelectric data storage unit 332 is performed in a state associated with the information to be represented.

  The muscle activity amount extraction processing unit 323 is stored in the myoelectric data storage unit on the basis of the motion data stored in the motion data storage unit 331 as pre-processing to facilitate muscle synergy analysis processing. The myoelectric data of each channel is trimmed and the myoelectric activity is extracted. Then, processing is performed to associate data representing the extracted electromyographic activity with the channel and store the data in the muscle activity data storage unit 333.

  The muscle synergy analysis processing unit 324 performs nonnegative matrix factorization (NMF) on the data representing the myoelectric activity amount extracted for each channel by the muscle activity amount extraction processing unit 323, whereby the above-mentioned each channel is The myoelectric activity is decomposed into a characteristic quantity of muscle activity (also called muscle synergy) and its activity (also called activation). Then, processing for storing data representing the muscle synergy and its activation in the muscle synergy data storage unit 334 in association with the channel identification information is performed.

  The sort processing unit 325 sorts muscle synergy obtained for each channel by the muscle synergy analysis processing unit 324 in the order of earlier timing of appearance of the peak of the activity amount. Then, processing for storing data representing the sorted muscle synergy and its activation in the sort data storage unit 335 is performed.

  The fitting processing unit 326 fits the data representing the muscle synergy after the sorting and the activation thereof into a function using regression analysis. For example, it fits by regression to this sigmoid function by the least squares method using a sigmoid function as a function. Then, each parameter of the sigmoid function or the evaluation index of the degree of fitting of the sigmoid function is stored in the fitting data storage unit 336 in association with the identification information of the target user US and the information indicating the measurement date.

The muscle activity estimation processing unit 327 has the following two estimation processing functions.
(1) The parameters of the sigmoid function newly obtained by today's swing are compared with the parameters of the sigmoid function stored in the fitting data storage unit 336 corresponding to the state of the swing in a plurality of days in the past, Read out the information representing the subjective evaluation for the most similar past swing. Then, a process of storing the read information representing the subjective evaluation in the estimation result storage unit 337 together with the information representing the date.

  (2) The evaluation index of the plurality of swings of the target user US stored in time series in the fitting data storage unit 336 is referenced, the stability of the swing is determined from the change of the evaluation index, and the determination result is indicated. A process of storing information in the estimation result storage unit 337.

  The output processing unit 327 reads the information representing the subjective evaluation of the swing of the target user US from the estimation result storage unit 337 or the information representing the determination result of the stability of the swing, for example, according to the request of the target user US. To the user terminal 4.

(Operation)
Next, the operation of the device configured as described above will be described. FIG. 2 is a flowchart showing the processing procedure and processing contents of the muscle activity estimation apparatus 3.
(1) Data Acquisition The target user US who is a subject performs a golf swing in a state where the myoelectric sensors 11 to 1 m are attached to a plurality of muscles, respectively. Further, the situation is imaged by the camera 2 from the side position.

  When the operator performs a measurement start operation with the measurement button (not shown), the control unit 32 of the muscle activity estimation device 3 detects the measurement start operation in step S10. Then, first, under the control of the motion data acquisition processing unit 321, the motion data of the target user US transmitted from the camera 2 in step S11 is received through the first receiving unit 311, and the received motion data is received. The motion data storage unit 331 stores the identification information of the target user US and the information indicating the imaging time.

  At the same time, under the control of the myoelectric data acquisition unit 322, the respective myoelectric data transmitted from the myoelectric sensors 11 to 1 m in step S12 are received via the second reception unit 312 and received. The respective myoelectric data are stored in the myoelectric data storage unit 332 in a state of being associated with the channel identification information and the information indicating the measurement date and time. The control unit 32 repeats the above processing operations until the operator performs an end operation with the measurement button, and this end operation is detected in step S13.

  Specifically, the target user US performs a golf swing a predetermined number of times, and in the state in which the motion data captured by the camera 2 and the myoelectric data measured by the myoelectric sensors 11 to 1m are mutually synchronized. get.

(2) Pre-processing Subsequently, the control unit 32 extracts muscle activity amounts from each of the myoelectric data in step S14 under the control of the muscle activity amount extraction processing unit 323 in order to facilitate muscle synergy analysis described later. Execute the process

  First, in order to identify a time region to be an analysis target of the golf swing from each electromyography data stored in the myoelectric data storage unit 332, from the motion data stored in the motion data storage unit 331, the golf swing start time and The end time is specified, and the electromyographic data is trimmed based on the specified start time and end time of the golf swing. The start time and the end time of the golf swing are specified by extracting an image pattern at the start of the swing and an image pattern at the end from the motion data by, for example, pattern recognition. The start time and the end time of the golf swing may be designated by the operator by visually observing the motion data and designating and inputting the time.

  Next, in order to remove noise and artifacts from the myoelectric data, the myoelectric data of each channel after the trimming is passed through a band pass Butterworth filter of, for example, 5 to 100 Hz. Subsequently, the myoelectric waveform is normalized for each channel in order to calculate the myoelectric waveform width meaning muscle activity. Normalization is performed, for example, by subtracting the median from the waveform and dividing by standard. Finally, to calculate muscle activity, the root mean square (RMS) of the normalized myoelectric waveform is calculated for each channel using a sliding window.

The RMS of the electromyogram obtained from the ith (i = 1 to m) channel calculated by the above processing is the muscle activity of the muscle associated with the channel
Define as Here, n is the sampling number of the myoelectric waveform from the start time to the end time.

In addition, when this muscle activity is indicated by a matrix E connected in rows,
It is expressed as Where c is the number of channels.

  The data representing the muscle activity is stored in the muscle activity data storage unit 333 in association with the channel identification information. FIG. 3 shows an example of the waveforms of myoelectric data of the channels ch1, ch2, ch3,... Before the pre-processing and the muscle activity data MA1, MA2, MA3,.

(3) Muscle Synergy Analysis Next, in step S15, the control unit 32 controls the muscle activity amount extraction processing unit 323 to control the channel ch1 under the control of the muscle synergy analysis processing unit 324 in order to extract the features of the muscle activity. , Non-negative matrix factorization (NMF) is performed on the data representing the amount of myoelectric activity extracted for each of the channels ch2, ch3,. As a result, assuming that the muscle activity obtained by the above pretreatment is E, the muscle synergy is S 1, and the activity (activation) is ^t A, the muscle activity E is calculated by the nonnegative matrix factorization (NMF)
It can be disassembled like

  FIG. 4 shows an example of muscle synergy obtained by the muscle synergy analysis process. In the figure, muscle synergies MB1, MB2 and MB3 respectively represent the amount of activity for each of the channels ch1 to chm in a bar graph. Further, FIG. 5 shows an example of activation of muscle synergies MB1, MB2 and MB3. In the figure, the amount of activity is displayed in shades of gray. Specifically, the larger the amount of activity, the lighter, and the smaller the amount of activity, the thicker the display. The horizontal axis shows the time course, the back swing activity in the left half time domain, and the downswing and follow-through activity in the right half time domain.

  In the above decomposition, the rank r of muscle synergy S 1, that is, the number of muscle synergy, can adjust the restoration rate of SA with respect to the amount of muscle activity E 2. The data representing the muscle synergy and the amount of activity thereof are stored in the muscle synergy data storage unit 334 in association with the channel identification information.

(4) Sorting Next, the control unit 32 executes the muscle synergy obtained for each channel in step S16 under the control of the sort processing unit 326 in order to examine the time change of muscle synergy in one golf swing, Reorder the timings at which the peaks of the activity appear. The peak of activity is identified by finding the time at which the maximum value is obtained after moving average filtering. Subsequently, the number j is reassigned to muscle synergy in the order of earlier appearance timing of peaks, and finally, the order of peaks is p _ik .

  The above processing is repeatedly performed for all the golf swings for each target user US. Data representing muscle synergies and their activation sorted for each golf swing is stored in the sort data storage unit 335. FIG. 8 shows the muscle synergies MB2, MB3, MB1,.

(5) Fitting Subsequently, under the control of the fitting processing unit 326, in step S17, the muscle synergy after sorting is regressed to the sigmoid function by the least squares method. At this time, the sigmoid function σ (t) is σ (t) = c / (1 + e ^{a (t−t0)} ) + y ₀
It is expressed as Here, c is a constant, a is a slope (gain), and y ₀ is an offset value. The result of regression to the above sigmoid function is
p _jk ∪σ (t _jk ) (5)
It is expressed as

Finally, in order to evaluate the degree of fitting of the sigmoid function, the mean square error (RMSE) and the determination coefficient (R ² ) are determined. These indicators are
It is expressed as

According to equation (6), it can be seen that the lower the RMSE, the better the fit. Also, according to the equation (7), it is understood that the larger the value of R ² , the better the fitting. Two indicators RMSE and R ² representing the degree of the fitting is stored in association with identification information of the target user US in fitting the data storage unit 336.

(6) Estimation of Muscle Activity Next, under the control of the muscle activity estimation processing unit 327, the control unit 32 executes a muscle activity estimation process in step S18. There are the following two examples of the estimation processing method of muscle activity.

(First embodiment)
In the fitting data storage unit 336, a parameter of a sigmoid function corresponding to the state of swing on a plurality of days in the past, a subjective evaluation value of the swing corresponding to the parameter, and information indicating the date at that time are mutually associated. Are stored. FIG. 11B shows an example of the stored data, and the subjective evaluation value is represented by three types of 1 = “good”, 0 = “normal”, −1 = “bad”. Note that FIG. 11A shows the number of swings subjected to fitting to the sigmoid function and the timing thereof, and in the figure, A ₁ , A ₂ , A ₃ ,... Correspond to swing groups for each date.

  As shown in the flowchart of FIG. 3, the muscle activity estimation processing unit 327 first reads the parameter of the sigmoid function corresponding to the swing state on a plurality of past days from the fitting data storage unit 336 in step S181. Next, in step S182, the parameters of the sigmoid function corresponding to the today's swing are sequentially compared with the parameters of the sigmoid function corresponding to the swings of the read past days. Then, in step S183, the parameter of the sigmoid function corresponding to today's swing and the past swing whose parameter is most similar is selected, and in step S184 the subjective evaluation value corresponding to the swing is read out, and this evaluation value is added to the date. The information is stored in the estimation result storage unit 337 as information representing the estimation result of the swing state.

Second Embodiment
In the fitting data storage unit 336, as shown in FIG. 12A, each time the target user US swings once, the fitting of the sigmoid function obtained from a predetermined number of swing groups in the past including the swing is performed. An index for evaluating the degree, that is, an index (stability index) indicating the stability of the swing is sequentially stored. FIG. 12B shows an example of the stored data, in which the determination coefficient (R ² ) is stored as an index for each of the swing groups A ₁ , A ₂ , A ₃ ,. A mean square error (RMSE) may be stored as an index indicating the stability of the swing.

Muscle activity estimation processing unit 327, as shown in the flowchart of FIG. 4, first step from the fitting data memory unit 336 by S191, reads the stability index R ² corresponding to the swing state until last. In step S192, the difference between the stability index corresponding to the current swing with stability index R ² for swinging up the read last, i.e. determine the change in stability of the swing. Then, in step S193, message information representing the change in the degree of stability is generated, and this message information is stored in the estimation result storage unit 337 as information representing the estimation result of the state of the swing along with information representing the current date and time.

(7) Output of Estimation Result The control unit 32 executes the process of outputting the estimation result in step S19 under the control of the output processing unit 328 as follows.
That is, it is assumed that the display request of the estimation result is set in advance by the target user US, or the display request of the estimation result is received from the user terminal 4 after the end of the estimation process. When determining that the display request is set in advance or received, the control unit 32 reads out information representing the estimation result of the state of the swing from the estimation result storage unit 337.

  For example, in the first embodiment, a subjective evaluation value for today's swing is read out, and information representing this subjective evaluation value is attached with a date, and the transmission unit 313 sends, for example, a mail to the user terminal 4 of the target user US. Send by Also, together with that, calculate the difference of the sigmoid function parameter between today's swing and past similar swing, convert the amount of the difference into the type of sound (eg timbre, volume, generation interval), and The sound data is transmitted from the transmission unit 313 to the user terminal 4 of the target user US.

On the other hand, in the second embodiment, message information representing a change in the degree of stability of the current swing is read out from the estimation result storage unit 337. Then, the message information is transmitted from the transmission unit 313 to the user terminal 4 by e.g. mail. The message information representing a change in stability of the swing, for example, stability index R ² is "has improved stability swing" if increased by more than a predetermined value, stability index R ² is equal to or higher than a predetermined value If it decreases, "the stability of the swing is reduced" is used.

(effect)
As described in detail above, in one embodiment of the present invention, the myoelectric data measured by the myoelectric sensor 11 to 1 m (channel) during the exercise (golf swing) of the target user US is acquired, and the myoelectric data is After trimming based on the motion data of the target user US obtained by the camera 2, a muscle activity amount is extracted. Then, after calculating muscle synergy and its activity amount (activation) from the muscle activity amount of each channel, the muscle synergy is sorted in the order in which the timing at which the peak of the activity amount appears appears. Next, the above-sorted muscle synergy obtained for all motions of the target user US is fitted to a sigmoid function by regression analysis, and based on the parameter of the sigmoid function or the index representing the degree of fitting at this time, The state of the swing is estimated and the result is transmitted to the user terminal 4.

  Therefore, according to one embodiment, when trying to extract muscle synergy from the myoelectric data obtained by each of the myoelectric sensors 11 to 1 m to find the feature of exercise, the muscle synergy of each channel has its activity amount The timing of the appearance of the peaks is sorted in the order of earlier. For this reason, it is possible to easily determine the presence or absence of the exercise chain from the muscle synergy after sorting.

  Furthermore, since the muscle synergy after sorting is determined for each of the plurality of exercises, and the muscle synergy after sorting obtained in each exercise is fitted with a sigmoid function, the transition of the exercise chain is It is possible to present in a form that is easy to evaluate. Therefore, it is possible to objectively grasp the transition of the muscle activity of a complex exercise such as a golf swing.

  More specifically, with respect to the state of today's swing, evaluation information on the state of the past swing in which the parameters of the sigmoid function are most similar is presented to the user. Therefore, the user can grasp the state of today's swing in comparison with the state of past swing. In addition, the degree of stability of the swing is determined from the change in the evaluation index obtained each time the swing is performed, and information indicating the determination result is presented to the user. For this reason, it becomes possible for the user to clearly grasp whether the stability of his or her swing is improved or decreased.

(Verification)
The inventor has verified the effectiveness of the muscle activity analysis method according to one embodiment described above. In this test, a total of four men, two males and two females, who have sufficient golf experience, are selected as subjects, and each person is playing a golf swing eight times. While measuring the electromyogram using a sensor, the state of each swing was imaged with a camera.

  The obtained myoelectric data were first preprocessed to exclude channel data of data loss, and muscle synergy analysis was performed on the remaining channel data. This time, we chose the number of muscle synergy so that the restoration rate of EMG is more than 79%. As a result, the number of muscle synergy was experimentally determined to be nine.

Among the measured electromyograms, the electromyograms of a certain subject taken before and after a certain swing are as shown in FIG. Pre-processing and muscle synergy analysis are performed on the electromyogram corresponding to FIG. 5, and muscle synergy and activation are extracted as shown in FIGS. 6 and 7. Verification requires muscle synergy and activation for each of the four subjects. In addition, in the figure which shows activation, a vertical axis | shaft is a test subject number (Subject 1-4), and a horizontal axis is the frequency | count of a swing. Finally, the activation peaks are fitted to the sigmoid function for every four subjects. FIG. 9 shows the result. RMSE of this time are each 0.295,0.281,0.283,0.168, R ² were respectively 0.793,0.827,0.841,0.940. The fitting results also showed that muscle synergy differs from subject to subject.

  The above verification uses 25 channel EMGs, and it is usually difficult to see the combination of EMGs with this number of channels. This is apparent from the point of view of the freedom of combination. However, when the analysis method of the embodiment of the present invention is adopted, the degree of freedom is reduced while maintaining a high recovery rate of the electromyogram, and then the muscle synergy is rearranged in the order in which the maximum value of the activity amount occurs. It has become possible to automatically extract the presence of synergy and the transition.

Interestingly, the peak of activation was found to be fitted with a high coefficient of determination R ² to the sigmoid function. This means that after applying power to muscles slowly, it applies to muscles in a chain and in a complex manner suddenly at a certain timing. The tendency of this chain in motion is also pointed out conventionally, and it is called motion chain. The movement chain is said to be the basis for increasing the speed of the golf head in the golf swing. Not only the motion but also the muscles can clearly show that the movement chain is occurring. By reducing the degree of freedom by nonnegative matrix factorization (NMF) in this way, even complex motions like golf may help discover features such as motion chains.

  In general, golf swing movement can be divided into four phases: address, backswing, downswing and follow through. Muscle synergy has different characteristics from phase to phase.

[Other embodiments]
In the above embodiment, although the case where a dedicated personal computer is provided as the muscle activity estimation device 3 has been described as an example, the muscle activity estimation processing function may be provided in the user terminal 4 or in a server computer or cloud computer on a network. Good. In this case, the myoelectric data and motion data are transmitted to the server computer or cloud computer via the user terminal or the alternative personal computer.

  In the above embodiment, although the case where the control unit 32 implements one of the first embodiment and the second embodiment has been described, the control unit 32 may be implemented by both the first and second embodiments. You may configure it.

  Furthermore, in the above-mentioned embodiment, although the case where muscle activity by a golf swing is analyzed and a swing state is estimated is described as an example, the present invention can be applied to complex movements such as batting and throwing motion of baseball. That is, by using nonnegative matrix factorization (NMF) for extraction of muscle synergy and activation as in the present invention, the degree of freedom can be reduced and important muscle synergy transitions can be easily found. Further, by sorting activation of muscle synergy in order of appearance timing of the peak, it is possible to clarify the presence or absence of a characteristic muscle activity exercise chain. Furthermore, by fitting the muscle synergy after sorting with a sigmoid curve, it is possible to present the transition of the movement chain in a form that can be easily evaluated.

  In the above embodiment, motion data is acquired in order to trim a time region required for analysis among myoelectric data, but if all myoelectric data are required for analysis, trimming using motion data is performed. Can be unnecessary.

  In addition, the type of exercise and the number of channels, the configuration of the muscle activity estimation apparatus, the procedure of analysis and estimation processing, and the processing contents can be variously modified without departing from the scope of the present invention.

  In short, the present invention is not limited to the above embodiment as it is, and at the implementation stage, the constituent elements can be modified and embodied without departing from the scope of the invention. In addition, various inventions can be formed by appropriate combinations of a plurality of components disclosed in the above embodiments. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, components in different embodiments may be combined as appropriate.

  11 to 1 m: myoelectric sensor, 2: camera, 3: muscle activity estimation device, 4: user terminal, 31: communication unit, 32: control unit, 33: storage unit, 311, 312: reception unit, 313: transmission unit , 321: motion data acquisition processing unit, 322: myoelectric data acquisition processing unit, 323: muscle activity amount extraction processing unit, 324: muscle synergy analysis processing unit, 325: sort processing unit, 326: fitting processing unit, 327: muscle Activity estimation processing unit 328 Output processing unit 331 Motion data storage unit 332 Electromyographic data storage unit 333 Muscle activity data storage unit 334 Muscle synergy data storage unit 335 Sort data storage unit 336 ... fitting data storage unit, 337 ... estimation result storage unit.

Claims (7)

Means for receiving electromyography data representing activity of said plurality of muscles from a plurality of electromyographic sensors mounted on a plurality of muscles used for exercise of a subject;
Means for extracting, based on the received electromyography data, muscle synergy characteristic of activity of each of the plurality of muscles and activation thereof ;
A means for rearranging the extracted muscle synergy in the order of the timing at which the maximum value of the activation appears;
Means for approximating data representing muscle synergy and activation of each muscle after the reordering to a predetermined function, and obtaining at least one of information representing the parameter of the approximated function and the degree of approximation When,
A storage unit storing at least one of a parameter of the function obtained for each of a plurality of motions performed in the past and information indicating the degree of approximation;
Means for estimating the state of a newly performed motion with reference to at least one of parameters of a function obtained for a plurality of motions in the past and information representing the degree of approximation stored in the storage unit A muscle activity estimation device characterized in that.   The means for estimating the state of motion compares the parameters of the function determined for the newly performed motion with the parameters of the function determined for a plurality of past motions stored in the storage unit; The muscle activity estimation device according to claim 1, wherein the newly performed motion state is estimated based on a past motion state in which parameters are similar.   The means for estimating the motion state is obtained for the newly performed motion with reference to the information stored in the storage unit that represents the degree of approximation of the function obtained for a plurality of motions in the past. The muscle activity estimation apparatus according to claim 1, wherein a change in information representing the degree of approximation of a function is calculated, and the state of the newly performed exercise is estimated based on the calculation result.   The apparatus according to any one of claims 1 to 3, further comprising: means for outputting information representing a newly performed state of motion estimated by the means for estimating the state of motion to an external device. A muscle activity estimation device according to any one of the preceding claims.   The means for outputting to the external device includes, as information representing the estimated newly performed state of motion, information representing a result of subjective evaluation of the state of motion and information representing the degree of stability of the state of motion The apparatus for estimating muscle activity according to claim 4, wherein at least one of them is output. A muscle activity estimation method executed by a muscle activity estimation apparatus capable of performing data communication with a plurality of myoelectric sensors attached to a plurality of muscles used for exercise of a subject, comprising:
Receiving electromyography data representing activity of the plurality of muscles from the plurality of myoelectric sensors;
Extracting, based on the received electromyography data, muscle synergy indicating the characteristics of the activity and activation of each of the plurality of muscles;
Reordering the extracted muscle synergy in the order of the timing at which the maximum value of the activation appears;
A process of approximating data representing muscle synergy and activation of each muscle after the reordering to a predetermined function, and obtaining at least one of information representing the parameter of the approximated function and the degree of approximation When,
Storing in the storage unit at least one of the parameter of the function obtained for each of a plurality of movements performed in the past and information indicating the degree of approximation;
Estimating at least one of a parameter of a function obtained for a plurality of motions in the past and information representing the degree of approximation stored in the storage unit, and estimating a state of the newly performed motion. A method of estimating muscle activity characterized by:   A program that causes a computer to function as each means of the muscle activity estimation device according to any one of claims 1 to 5.