Muscle activity estimation device, muscle activity estimation method, and muscle activity estimation program

Abstract

Provided is a muscle activity estimation device comprising a model generation unit and a muscle activity estimation unit. In the present invention, the model generation unit generates a decoding model for each muscle activity pattern and a decoding model for each muscle activity trend. The muscle activity estimation unit uses the decoding model for each muscle activity pattern and the decoding model for each muscle activity trend to calculate a muscle activity from the perspective of the muscle activity pattern and a muscle activity from the perspective of the muscle activity trend on the basis of a brain activity and estimates a muscle activity on the basis of the muscle activity from the perspective of the muscle activity pattern and the muscle activity from the perspective of the muscle activity trend.

Description

[Technical Field]

[0001] The present invention relates to a muscle activity estimation device, a muscle activity estimation method, and a muscle activity estimation program. [Background technology]

[0002] BACKGROUND ART There is known a technology for assisting a user's exercise through an assistive device such as an exoskeleton robot by decoding the user's exercise information, such as muscle activity, based on brain activity measured using an external device.

[0003] In order to decode muscle activity from brain activity, Non-Patent Document 1 proposes a method for calculating a linear regression model using multiple feature quantities calculated from brain activity signals measured from the brain as explanatory variables and electromyograms as the target variable. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Shin, D., Watanabe, H., Kambara, H., Nambu, A., Isa, T., Nishimura, Y., & Koike, Y. (2012). Prediction of Muscle Activities from Electrocorticograms in Primary Motor Cortex of Primates. PLOS ONE, 7(10), e47992. https: / / doi.org / 10.1371 / JOURNAL.PONE.0047992 Summary of the Invention [Problem to be solved by the invention]

[0005] The human body has many muscles, and movement can be achieved by coordinating their movements. It is assumed that the relationship between brain activity and muscle activity will differ depending on the spatial combination of muscles (muscle activation pattern) when multiple muscles are coordinating their movements.

[0006] Furthermore, precise motor movements are achieved by controlling the strength of each muscle's contraction over time, but even in such a state where muscle activity increases, remains constant, and decreases (muscle activity trend), it is assumed that the relationship between brain activity and muscle activity cannot necessarily be expressed by a single decoding model.

[0007] Thus, it is necessary to construct a decoding model that takes into account the spatial and temporal activity states of muscle activity. However, Non-Patent Document 1 only estimates muscle activity based on brain activity using a single decoding model.

[0008] The present invention has been made in light of the above circumstances, and its purpose is to provide a muscle activity estimation device, a muscle activity estimation method, and a muscle activity estimation program that estimate muscle activity based on brain activity taking into account the spatial and temporal activity state of muscle activity. [Means for solving the problem]

[0009] One aspect of the present invention is a muscle activity estimation device. The muscle activity estimation device includes a model generation unit and a muscle activity estimation unit. The model generation unit Calculating muscle activity for each muscle activity pattern Decoding model for each muscle activity pattern ,Calculate muscle activity for each muscle activity trend A decoding model for each muscle activity trend is generated. The muscle activity estimation unit calculates muscle activity of the muscle activity pattern and muscle activity of the muscle activity trend based on brain activity using the decoding model for each muscle activity pattern and the decoding model for each muscle activity trend, and estimates muscle activity based on the muscle activity of the muscle activity pattern and muscle activity of the muscle activity trend. The model generation unit includes a brain activity measurement unit that divides brain activity signals of multiple channels at each time into multiple frequency bands and calculates brain activity features; a muscle activity measurement unit that calculates muscle activity features from muscle activity signals of multiple channels at each time; a muscle activity pattern calculation unit that calculates a muscle activity pattern using the muscle activity features; a first decoding model generation unit that generates a decoding model for each muscle activity pattern using the brain activity features, the muscle activity features, and the muscle activity pattern; a muscle activity trend calculation unit that calculates multiple types of muscle activity trends for each muscle using the muscle activity features; and a second decoding model generation unit that generates a decoding model for each muscle activity trend using the brain activity features, the muscle activity features, and the muscle activity trend. The muscle activity estimation unit includes a brain activity measurement unit that divides brain activity signals of multiple channels at each time into multiple frequency bands and calculates brain activity feature quantities; a muscle activity calculation unit for each muscle activity pattern that calculates muscle activity for each muscle activity pattern from the brain activity feature quantities using a decoding model for each muscle activity pattern; a muscle activity pattern probability calculation unit that calculates a probability distribution of muscle activity patterns at each time from the brain activity feature quantities; a muscle activity pattern perspective muscle activity calculation unit that calculates muscle activity from the muscle activity for each muscle activity pattern calculated by the muscle activity calculation unit for each muscle activity pattern and the probability distribution of the muscle activity pattern calculated by the muscle activity pattern probability calculation unit; a muscle activity trend probability calculation unit that calculates a probability distribution of the muscle activity trend at each time from the brain activity feature amount; a muscle activity trend-perspective muscle activity calculation unit that calculates muscle activity from the muscle activity trend perspective based on the muscle activity for each muscle activity trend calculated by the muscle activity trend-perspective muscle activity calculation unit and the probability distribution of the muscle activity trend calculated by the muscle activity trend probability calculation unit; and an estimation result calculation unit that calculates an average of the muscle activity from the muscle activity pattern perspective calculated by the muscle activity pattern-perspective muscle activity calculation unit and the muscle activity from the muscle activity trend perspective calculated by the muscle activity trend calculation unit, and sets the calculated average as an estimation result of muscle activity.

[0010] One aspect of the present invention is The computer runs A muscle activity estimation method. The computera generation step of generating a decoding model for each muscle activity pattern that calculates muscle activity for each muscle activity pattern and a decoding model for each muscle activity trend that calculates muscle activity for each muscle activity trend; The computer and an estimation step of calculating muscle activities of muscle activity patterns and muscle activity trends based on brain activity using a decoding model for each muscle activity pattern and a decoding model for each muscle activity trend, and estimating muscle activities based on the muscle activities of the muscle activity patterns and muscle activity trends. The generation step includes a step of dividing brain activity signals of multiple channels at each time into multiple frequency bands to calculate brain activity features, a step of calculating muscle activity features from the muscle activity signals of the multiple channels at each time, a step of calculating a muscle activity pattern using the muscle activity features, a step of generating a decoding model for each muscle activity pattern using the brain activity features, the muscle activity features, and the muscle activity pattern, a step of calculating multiple types of muscle activity trends for each muscle using the muscle activity features, and a step of generating a decoding model for each muscle activity trend using the brain activity features, the muscle activity features, and the muscle activity trend. The estimation step includes the steps of: dividing the brain activity signals of multiple channels at each time into multiple frequency bands to calculate brain activity features; calculating muscle activity for each muscle activity pattern from the brain activity features using a decoding model for each muscle activity pattern; calculating a probability distribution of muscle activity patterns at each time from the brain activity features; calculating muscle activity from a muscle activity pattern perspective from muscle activity for each muscle activity pattern and the probability distribution of the muscle activity patterns; calculating muscle activity for each muscle activity trend from the brain activity features using a decoding model for each muscle activity trend; calculating a probability distribution of the muscle activity trend at each time from the brain activity features; calculating muscle activity from a muscle activity trend perspective from muscle activity for each muscle activity trend and the probability distribution of the muscle activity trend; and calculating an average of the muscle activity from the muscle activity pattern perspective and the muscle activity trend perspective, and setting the calculated average as an estimated muscle activity result.

[0011] One aspect of the present invention is a muscle activity estimation program that causes a computer to execute the functions of the components of the muscle activity estimation device described above. [Effects of the Invention]

[0012] According to the present invention, there are provided a muscle activity estimation device, a muscle activity estimation method, and a muscle activity estimation program that estimate muscle activity based on brain activity taking into account the spatial and temporal activity state of muscle activity. [Brief explanation of the drawings]

[0013] [Figure 1] FIG. 1 is a block diagram illustrating an example of a functional configuration of a muscle activity estimation device according to an embodiment. [Figure 2] FIG. 2 is a block diagram illustrating an example of a hardware configuration of the muscle activity estimation device according to the embodiment. [Figure 3] FIG. 3 is a flowchart showing an example of a processing procedure and processing content for generating a decoding model executed by the model generating unit of the muscle activity estimation device according to the embodiment. [Figure 4] FIG. 4 is a flowchart showing an example of a processing procedure and processing content for muscle activity trend calculation executed by the muscle activity trend calculation unit of the model generation unit of the muscle activity estimation device according to the embodiment. [Figure 5] FIG. 5 is a flowchart showing an example of a processing procedure and processing content for estimating muscle activity based on brain activity, which is executed by the muscle activity estimation unit of the muscle activity estimation device according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0014] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0015] <Configuration example> (Functional configuration) First, the functional configuration of a muscle activity estimation device according to an embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the functional configuration of a muscle activity estimation device 10 according to an embodiment.

[0016] The muscle activity estimation device 10 estimates a user's muscle activity from brain activity signals, which are bioelectric potential signals generated in association with the user's brain activity. The estimation result signals output by the muscle activity estimation device 10 are used in assistive devices such as exoskeleton robots that assist the user in exercising.

[0017] The muscle activity estimation device 10 includes a model generation unit 20 and a muscle activity estimation unit 30.

[0018] The model generation unit 20 uses pre-measured data, i.e., brain activity signals and muscle activity signals, to generate and store a decoding model for each muscle activity pattern and a decoding model for each muscle activity trend.

[0019] The muscle activity estimation unit 30 calculates the muscle activity of the muscle activity pattern and the muscle activity of the muscle activity trend based on brain activity using a decoding model for each muscle activity pattern and a decoding model for each muscle activity trend stored in the model generation unit 20, and estimates the muscle activity based on the muscle activity of the muscle activity pattern and the muscle activity of the muscle activity trend.

[0020] The model generation unit 20 includes a brain activity measurement unit 21, a muscle activity measurement unit 22, a muscle activity pattern calculation unit 23, a muscle activity trend calculation unit 24, a decoding model generation unit 25 for each muscle activity pattern, and a decoding model generation unit 26 for each muscle activity trend.

[0021] Brain activity signals are input to the brain activity measurement unit 21. For example, the brain activity signals are measured by N electrodes attached to the scalp. In other words, the brain activity signals are N-channel signals. In the following, the N-channel brain activity signals at each time t are referred to as S i This is written as (t)(i=1,…,N).

[0022] The brain activity measurement unit 21 receives the N-channel brain activity signal S at each time t. i For example, the brain activity measurement unit 21 divides the brain activity signal S (t) into multiple frequency bands to calculate the brain activity feature amount.i A band-pass filter is applied to (t) to divide it into nine frequency bands: δ (1.5 to 4 Hz), θ (4 to 8 Hz), α (8 to 14 Hz), β1 (14 to 20 Hz), β2 (20 to 30 Hz), γ1 (30 to 50 Hz), γ2 (50 to 90 Hz), γ3 (90 to 120 Hz), and γ4 (120 to 150 Hz). As a result, the brain activity measurement unit 21 calculates a 9N-dimensional (N channels × 9 frequency bands) brain activity feature value v (i,j) Calculate (t) (i=1,…,N) (j=δ,…,γ4).

[0023] The brain activity measurement unit 21 calculates the brain activity feature value v (i,j) (t) (i=1, ..., N) (j=δ, ..., γ4) is output to a decoding model generation unit 25 for each muscle activity pattern and a decoding model generation unit 26 for each muscle activity trend.

[0024] Muscle activity signals are input to the muscle activity measuring unit 22. For example, the muscle activity signals are measured by M electrodes attached to the skin directly above the muscles. In other words, the muscle activity signals are M-channel signals. In the following, the k-th muscle activity signal at each time t is referred to as m k (t) (i=1,...,M). The muscle corresponding to the k-th electrode is called the k-th muscle.

[0025] The muscle activity measuring unit 22 receives the muscle activity signal m at each time t. k For example, the muscle activity measuring unit 22 first calculates the muscle activity feature amount from the muscle activity signal m k (t) is filtered through a bandpass filter that passes through 20-500 Hz to remove noise during exercise. The muscle activity measuring unit 22 then outputs the noise-removed muscle activity signal m k (t) is rectified and a low-pass filter of 10 Hz or less is applied to calculate muscle activity features (t) (k=1,...,M). Muscle activity features can also be considered amplitude features.

[0026] The muscle activity measuring unit 22 outputs the calculated muscle activity feature amount (t) (k=1, . . . , M) to the muscle activity pattern calculating unit 23 and the muscle activity trend calculating unit 24.

[0027] The muscle activity pattern calculation unit 23 calculates muscle activity patterns using the muscle activity feature amount (t) input from the muscle activity measurement unit 22. The M muscle activity patterns at each time t are expressed as a vector [a1(t),...,a M (t)]. Hereinafter, this vector will be referred to as a muscle activity vector. The muscle activity pattern calculation unit 23 assigns a muscle activity pattern label q=1,...,Q to the muscle activity vector at each time t. Hereinafter, the muscle activity pattern label group at time t will be represented as q(t). Any method can be used to assign the muscle activity pattern label group q(t). For example, the muscle activity pattern calculation unit 23 assigns the muscle activity pattern label group q(t) using the "k-means method."

[0028] The muscle activity pattern calculation unit 23 outputs the muscle activity pattern label group q(t) and the muscle activity feature amount (t) at each time t to the decoding model generation unit 25 for each muscle activity pattern.

[0029] The decoding model generation unit 25 for each muscle activity pattern generates a brain activity feature value v (i,j) (t) and the muscle activity feature amount (t) and muscle activity pattern label group q(t) at each time t input from the muscle activity pattern calculation unit 23, a decoding model for each muscle activity pattern is generated and stored.

[0030] The decoding model generation unit 25 for each muscle activity pattern first calculates the time group in which the muscle activity pattern label is labeled as 1 from the muscle activity pattern label group q(t), and then calculates the muscle activity vector group [a1(τ), ..., a M (τ)] and the brain activity vectors [v (i,j) (τ)] (i = 1, ..., N; j = δ, ..., γ4).

[0031] The decoding model generation unit 25 for each muscle activity pattern extracts the extracted brain activity vectors [v (i,j) (τ)] and the muscle activity vectors [a1(τ),…,a M (τ)] is used to generate a decoding model for muscle activity pattern label 1.

[0032] Any method may be used to generate a decoding model. For example, the decoding model generation unit 25 for each muscle activity pattern uses a linear regression method to generate a decoding model for the muscle activity pattern label 1 shown below.

[0033]

number

[0034] The muscle activity trend calculation unit 24 calculates multiple types of muscle activity trends for each muscle using the muscle activity feature values ​​(t) (k=1, ..., M) input from the muscle activity measurement unit 22. For example, the muscle activity trend calculation unit 24 calculates three types of muscle activity trends: increasing, maintaining, and decreasing.

[0035] The muscle activity trend calculation unit 24 calculates a result of first-order differentiation of the muscle activity feature value (t) of the k-th muscle at each time t as a' k (t) and a' k (t) is compared with a threshold c (c>0), and the muscle activity trend label r of the kth muscle at time t is obtained. k Set (t)(=1,2,3).

[0036] The muscle activity trend calculation unit 24 calculates a' k When (t) is greater than the threshold c, the muscle activity of the kth muscle is determined to be on an upward trend, and the muscle activity trend label r k Set (t)=1.

[0037] The muscle activity trend calculation unit 24 calculates a' kWhen (t) is smaller than the threshold c, the muscle activity of the k-th muscle is determined to be in a downward trend, and the muscle activity trend label r k Set (t)=3.

[0038] The muscle activity trend calculation unit 24 calculates a' k When (t) is equal to the threshold c, the muscle activity of the k-th muscle is determined to be in a maintenance trend, and the muscle activity trend label r k Set (t)=2.

[0039] The muscle activity trend calculation unit 24 calculates the muscle activity trend label for each muscle at each time t, and as a result, a muscle activity trend label group r k Calculate (t) (k=1,…,M).

[0040] The muscle activity trend calculation unit 24 calculates the muscle activity trend label set r k (t) and the muscle activity feature quantity (t) are output to the decoding model generation unit 26 for each muscle activity trend.

[0041] The decoding model generation unit 26 for each muscle activity trend calculates the brain activity feature amount input from the brain activity measurement unit 21 and v (i,j) (t), and the muscle activity feature (t) and muscle activity trend label group r input from the muscle activity trend calculation unit 24. k (t) is used to generate and store a decoding model for each muscle activity trend for each muscle.

[0042] The muscle activity trend decoding model generation unit 26 generates a muscle activity trend label group r for the time group in which the muscle activity trend of the k-th muscle is labeled as 1. k (t), and the corresponding time group τ(τ∈r k (t=1) M (τ)] and the brain activity vectors [v (i,j) (τ)] (i = 1, ..., N; j = δ, ..., γ4).

[0043] The decoding model generation unit 26 for each muscle activity trend extracts the extracted brain activity vector group [v (i,j) (τ)] and the muscle activity vectors [a1(τ),…,a M (τ)] to decode the trend label r for the k-th muscle. k is generated as a decoding model with a value of 1.

[0044] Any method may be used to generate a decoding model. For example, the muscle activity trend-specific decoding model generation unit 26 uses a linear regression method to generate a decoding model for muscle activity trend label 1 shown below.

[0045]

number

[0046] The muscle activity estimation unit 30 includes a brain activity measurement unit 31, a muscle activity calculation unit 32 for each muscle activity pattern, a muscle activity pattern probability calculation unit 33, a muscle activity calculation unit 34 for each muscle activity trend, a muscle activity trend probability calculation unit 35, a muscle activity calculation unit 36 ​​from the perspective of muscle activity pattern, a muscle activity calculation unit 37 from the perspective of muscle activity trend, and an estimation result calculation unit 38.

[0047] Similarly to the brain activity measurement unit 21, the brain activity measurement unit 31 receives the N-channel brain activity signal S i (t) (i=1,...,N) is input. N-channel brain activity signal S i The measurement method for (t) is as described above.

[0048] The brain activity measurement unit 31 receives the input brain activity signal S at each time t. i (t) performs the same processing as the brain activity measurement unit 21. That is, the brain activity measurement unit 31 divides the N-channel brain activity signals at each input time t into nine frequency bands to calculate brain activity feature quantities. As described above in the explanation of the brain activity measurement unit 21, as a result, the brain activity measurement unit 31 calculates 9N (N channels × 9 frequency bands)-dimensional brain activity feature quantities v (i,j)Calculate (t) (i=1,…,N) (j=δ,…,γ4).

[0049] The brain activity measurement unit 31 calculates the brain activity feature value v (i,j) (t)(i=1,...,N)(j=δ,...,γ4) is output to the muscle activity calculation unit 32 for each muscle activity pattern, the muscle activity pattern probability calculation unit 33, the muscle activity calculation unit 34 for each muscle activity trend, and the muscle activity trend probability calculation unit 35.

[0050] The muscle activity calculation unit 32 for each muscle activity pattern calculates the brain activity feature value v input from the brain activity measurement unit 31. (i,j) From (t), the muscle activity for each muscle activity pattern is calculated using a decoding model for each muscle activity pattern.

[0051] For this reason, the muscle activity calculation unit 32 for each muscle activity pattern first obtains the weight coefficient w for each muscle activity pattern label from the decoding model generation unit 25 for each muscle activity pattern. k,q and the bias term b k,q Load.

[0052]

number

[0053]

number

[0054]

number

[0055]

number

[0056]

number

[0057]

number

[0058]

number

[0059]

number

[0060]

number

[0061] The estimation result signal is input to an assistive device such as an exoskeleton robot and is used to assist the user's exercise.

[0062] (Hardware configuration) Next, the hardware configuration of the muscle activity estimation device 10 will be described with reference to Fig. 2. The muscle activity estimation device 10 is configured by a computer. For example, the muscle activity estimation device 10 is configured by a personal computer or the like.

[0063] 2 is a block diagram showing an example of a hardware configuration of the muscle activity estimation device 10 according to the embodiment. As shown in FIG. 2, the muscle activity estimation device 10 includes a processor 41, a read-only memory (ROM) 42, a random access memory (RAM) 43, an auxiliary storage device 44, and an input / output interface 45.

[0064] The processor 41, ROM 42, RAM 43, auxiliary storage device 44, and input / output interface 45 are electrically connected to one another via a bus 46, and exchange data via the bus 46.

[0065] The processor 41 is configured by a general-purpose hardware processor including, for example, a CPU (Central Processing Unit), a GPU (Graphical Processing Unit), etc. The processor 41 controls the ROM 42, the RAM 43, the auxiliary storage device 44, and the input / output interface 45 as a whole.

[0066] The ROM 42 is a non-volatile memory. The ROM 42 non-temporarily stores a startup program required when the processor 41 is started. The processor 41 starts up by executing the program in the ROM 42. The ROM 42 is configured, for example, with an EPROM (Erasable Programmable Read Only Memory), and stores various startup settings in addition to the startup program.

[0067] The RAM 43 is a volatile memory that temporarily stores programs and the like required for processing by the processor 41. The RAM 43 is used by the processor 41 as a working memory.

[0068] The auxiliary storage device 44 is configured by a non-volatile memory such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The auxiliary storage device 44 non-temporarily stores programs to be executed by the processor 41 and data required for executing the programs. The processor 41 executes various functions by loading the programs and data in the auxiliary storage device 44 into the RAM 43 and executing them.

[0069] The input / output interface 45 is an interface that is connected to external input devices, output devices, etc., and enables the input of information from the input devices and the output of information to the output devices. For example, the input / output interface 45 includes a wireless communication interface. For example, the input devices are composed of a keyboard, a mouse, a touch panel, a receiving device, various drives, etc. The input devices are not limited to these, and may be composed of any other input devices. For example, the output devices are composed of a display, a transmitting device, etc. The output devices are not limited to these, and may be composed of any other output devices. The input and output devices may be composed of an input / output device that has the functions of both. For example, the input / output device is composed of a tablet, etc.

[0070] The program non-temporarily stored in the auxiliary storage device 44 is provided to the computer in a state where it is stored in a computer-readable recording medium. For example, the recording medium may be an optical disk (CD-ROM, CD-R, DVD-ROM, DVD-R, etc.), a magneto-optical disk (MO, etc.), a semiconductor memory, etc. In this case, the program in the recording medium is read via an input device such as a drive and non-temporarily stored in the auxiliary storage device 44. Alternatively, the program may be stored in a server on a network, downloaded from the server, and non-temporarily stored in the auxiliary storage device 44.

[0071] At startup, processor 41 executes a program in ROM 42 and loads and starts the OS into RAM 43. Under control of the OS, processor 41 monitors input instructions, connections to external devices, and the like. Also, under control of the OS, processor 41 sets up a program area and a data area in RAM 43. In response to an input instruction to start muscle activity estimation device 10, processor 41 loads a muscle activity estimation program and data from auxiliary storage device 44 into the program area and data area of ​​RAM 43, respectively. Processor 41 calculates the data in the data area according to the muscle activity estimation program and writes the calculation results to the data area. Through these operations, processor 41 and RAM work together to implement the functions of model generation unit 20 and muscle activity estimation unit 30 of muscle activity estimation device 10.

[0072] <Example of operation> (Decoding model generation process) Next, a decoding model generation process executed by the model generation unit 20 of the muscle activity estimation device 10 will be described with reference to Fig. 3. Fig. 3 is a flowchart showing an example of the processing procedure and processing content for generating a decoding model executed by the model generation unit of the muscle activity estimation device according to the embodiment.

[0073] In step S11, the brain activity measurement unit 21 receives the N-channel brain activity signals S at each time t. i For example, the brain activity measurement unit 21 divides the brain activity signal S (t) of each channel into nine frequency bands. i (t) into nine frequency bands: δ (1.5 to 4 Hz), θ (4 to 8 Hz), α (8 to 14 Hz), β1 (14 to 20 Hz), β2 (20 to 30 Hz), γ1 (30 to 50 Hz), γ2 (50 to 90 Hz), γ3 (90 to 120 Hz), and γ4 (120 to 150 Hz). As a result, the brain activity measurement unit 21 calculates a 9N-dimensional EEG feature value v (i,j) Calculate (t) (i=1,…,N) (j=δ,…,γ4).

[0074] In step S12, the muscle activity signals m of M channels at each time t are input. kCalculate muscle activity feature (t) (k=1,...,M) from (t) (i=1,...,M). Brain activity signal S i (t) is divided into nine frequency bands. The calculation method is as described above.

[0075] In step S13, the muscle activity pattern calculation unit 23 calculates the muscle activity pattern [a1(t), ..., a M The muscle activity pattern calculation unit 23 further calculates the muscle activity pattern [a1(t), ..., a M (t)] is assigned a set of muscle activity pattern labels q(t) using, for example, the k-means algorithm.

[0076] In step S14, the decoding model generation unit 25 for each muscle activity pattern generates the brain activity feature value v calculated in step S11. (i,j) (t), the muscle activity feature (t) calculated in step S12, and the muscle activity pattern label group q(t) calculated in step S13 are used to generate and store a decoding model for each muscle activity pattern. The method for generating the decoding model is as described above.

[0077] In step S15, the muscle activity trend calculation unit 24 calculates three types of muscle activity trends, increasing, maintaining, and decreasing, for each muscle using the muscle activity feature (t) (k=1, ..., M) calculated in step S12.

[0078] Here, the muscle activity trend calculation process executed by the muscle activity trend calculation unit 24 of the model generation unit 20 of the muscle activity estimation device 10 will be described with reference to Fig. 4. Fig. 4 is a flowchart showing an example of the processing procedure and processing content of muscle activity trend calculation executed by the muscle activity trend calculation unit 24 of the model generation unit 20 of the muscle activity estimation device 10 according to the embodiment.

[0079] In step S21, the muscle activity trend calculation unit 24 performs first-order differentiation on the muscle activity feature amount (t) of the k-th muscle at each time t.

[0080] In step S22, the muscle activity trend calculation unit 24 calculates the first-order differential result a' of the muscle activity feature value (t). k (t) is compared with a threshold c (c>0).

[0081] As a result of the comparison in step S22, a' k If (t) is greater than the threshold value c, in step S23, the muscle activity trend calculation unit 24 determines that the muscle activity of the k-th muscle is on an upward trend, and assigns the muscle activity trend label r k Set (t)=1.

[0082] As a result of the comparison in step S22, a' k If (t) is smaller than the threshold value c, in step S24, the muscle activity trend calculation unit 24 determines that the muscle activity of the k-th muscle is on a downward trend, and assigns the muscle activity trend label r k Set (t)=3.

[0083] As a result of the comparison in step S22, a' k If (t) is equal to the threshold value c, in step S25, the muscle activity trend calculation unit 24 determines that the muscle activity of the k-th muscle is in a maintenance trend, and assigns the muscle activity trend label r k Set (t)=2.

[0084] The muscle activity trend calculation unit 24 performs the above process for each muscle, and as a result, a muscle activity trend label set r k (t) (k=1,…,M) is calculated. k After calculating (t), the process proceeds to step S16 in FIG.

[0085] In step S16, the decoding model generation unit 26 for each muscle activity trend calculates the brain activity feature amount calculated in step S11 and v (i,j) (t), the muscle activity feature (t) calculated in step S12, and the muscle activity trend label group r calculated in step S15.k Using (t), a decoding model for each muscle activity trend is generated and stored for each muscle. The method for generating the decoding model is as described above.

[0086] (Estimation of muscle activity based on brain activity) Next, the process of estimating muscle activity based on brain activity executed by the muscle activity estimation unit 30 of the muscle activity estimation device 10 will be described with reference to Fig. 5. Fig. 5 is a flowchart showing the process of estimating muscle activity based on brain activity executed by the muscle activity estimation unit 30 of the muscle activity estimation device 10 according to the embodiment.

[0087] In step S31, the brain activity measurement unit 31 receives the N-channel brain activity signals S at each time t. i For example, the brain activity measurement unit 31 divides the brain activity signal S (t) of each channel into nine frequency bands. i (t) into nine frequency bands: δ (1.5 to 4 Hz), θ (4 to 8 Hz), α (8 to 14 Hz), β1 (14 to 20 Hz), β2 (20 to 30 Hz), γ1 (30 to 50 Hz), γ2 (50 to 90 Hz), γ3 (90 to 120 Hz), and γ4 (120 to 150 Hz). As a result, the brain activity measurement unit 31 calculates a 9N-dimensional EEG feature value v (i,j) Calculate (t) (i=1,…,N) (j=δ,…,γ4).

[0088]

number

[0089]

number

[0090]

number

[0091]

number

[0092]

number

[0093]

number

[0094]

number

[0095] <effect> In the muscle activity estimation device 10 according to the embodiment, a model generation unit 20 generates and stores a decoding model for each muscle activity pattern and a decoding model for each muscle activity trend using brain activity signals and muscle activity signals measured in advance. Furthermore, a muscle activity estimation unit 30 calculates the muscle activity of a muscle activity pattern and a muscle activity trend based on brain activity using the decoding model for each muscle activity pattern and the decoding model for each muscle activity trend stored in the model generation unit 20, and estimates the muscle activity based on the muscle activity of the muscle activity pattern and the muscle activity trend.

[0096] As described above, the muscle activity estimation device 10 according to the embodiment estimates muscle activity using two types of decoding models for muscle activity patterns that take into account the spatial and temporal activity states of muscle activity, and one decoding model for muscle activity trends. This makes it possible to estimate muscle activity based on brain activity that takes into account the spatial and temporal activity states of muscle activity.

[0097] Other Embodiments In the embodiment, an example has been described in which the generation of a decoding model and the estimation of muscle activities performed by the muscle activity estimation device 10 are all performed by having the processor 41 execute a program. However, some or all of these functions may be performed using an integrated circuit configured for a specific application, such as an ASIC (Application Specific Integrated Circuit) or a DSP (Digital Signal Processor).

[0098] The present invention is not limited to the above-described embodiments, and various modifications can be made in the implementation stage without departing from the spirit of the invention. Furthermore, the embodiments may be implemented in appropriate combinations, in which case the combined effects can be obtained. Furthermore, the above-described embodiments include various inventions, and various inventions can be extracted by combining selected elements from the disclosed elements. For example, if the problem can be solved and the desired effect can be obtained even if some elements are deleted from all elements shown in the embodiments, the configuration from which these elements are deleted can be extracted as an invention. [Explanation of symbols]

[0099] 10…Muscle activity estimation device 20...Model generation section 21...Brain activity measurement unit 22...Muscle activity measurement unit 23...Muscle activity pattern calculation unit 24...Muscle activity trend calculation section 25...Decoding model generation unit for each muscle activity pattern 26...Decoding model generation unit for each muscle activity trend 30...Muscle activity estimation section 31...Brain activity measurement unit 32...Muscle activity calculation unit for each muscle activity pattern 33...Muscle activity pattern probability calculation unit 34...Muscle activity calculation section for each muscle activity trend 35...Muscle activity trend probability calculation unit 36...Muscle activity calculation unit from the perspective of muscle activity pattern 37...Muscle activity calculation section from the perspective of muscle activity trend 38…Estimation result calculation unit 41...Processor 42...ROM 43...RAM 44…Auxiliary storage device 45...Input / output interface 46...Bus

Claims

1. a model generation unit that generates a decoding model for each muscle activity pattern that calculates muscle activity for each muscle activity pattern and a decoding model for each muscle activity trend that calculates muscle activity for each muscle activity trend; a muscle activity estimation unit that calculates muscle activity of a muscle activity pattern and a muscle activity of a muscle activity trend based on brain activity using a decoding model for each muscle activity pattern and a decoding model for each muscle activity trend, and estimates muscle activity based on the muscle activity of the muscle activity pattern and the muscle activity of the muscle activity trend; and The model generation unit a brain activity measurement unit that divides brain activity signals of multiple channels at each time into multiple frequency bands and calculates brain activity feature amounts; a muscle activity measurement unit that calculates muscle activity feature values ​​from muscle activity signals of multiple channels at each time; a muscle activity pattern calculation unit that calculates a muscle activity pattern using the muscle activity feature amount; a first decoding model generation unit that generates a decoding model for each of the muscle activity patterns using the brain activity feature amount, the muscle activity feature amount, and the muscle activity pattern; a muscle activity trend calculation unit that calculates multiple types of muscle activity trends for each muscle using the muscle activity feature amount; a second decoding model generation unit that generates a decoding model for each muscle activity trend using the brain activity feature amount, the muscle activity feature amount, and the muscle activity trend; and The muscle activity estimation unit a brain activity measurement unit that divides brain activity signals of multiple channels at each time into multiple frequency bands and calculates brain activity feature amounts; a muscle activity calculation unit for each muscle activity pattern that calculates muscle activity for each muscle activity pattern from the brain activity feature amount using a decoding model for each muscle activity pattern; a muscle activity pattern probability calculation unit that calculates a probability distribution of muscle activity patterns at each time from the brain activity feature amount; a muscle activity pattern-based muscle activity calculation unit that calculates muscle activity from the perspective of muscle activity patterns from the muscle activity for each muscle activity pattern calculated by the muscle activity calculation unit for each muscle activity pattern and the probability distribution of the muscle activity patterns calculated by the muscle activity pattern probability calculation unit; a muscle activity calculation unit for each muscle activity trend that calculates muscle activity for each muscle activity trend from the brain activity feature amount using a decoding model for each muscle activity trend; a muscle activity trend probability calculation unit that calculates a probability distribution of muscle activity trends at each time point from the brain activity feature amount; a muscle activity trend-based muscle activity calculation unit that calculates muscle activity from the perspective of the muscle activity trend based on the muscle activity for each muscle activity trend calculated by the muscle activity trend-based muscle activity calculation unit and the probability distribution of the muscle activity trend calculated by the muscle activity trend probability calculation unit; an estimation result calculation unit that calculates an average of the muscle activity from the perspective of the muscle activity pattern calculated by the muscle activity pattern calculation unit and the muscle activity from the perspective of the muscle activity trend calculated by the muscle activity trend calculation unit, and sets the calculated average as an estimation result of the muscle activity; A muscle activity estimation device having the above.

2. The muscle activity trend calculation unit compares a differentiation result obtained by first-order differentiation of the muscle activity feature amount with a threshold value, and determines that the muscle activity is in an upward trend when the differentiation result is greater than the threshold value, determines that the muscle activity is in a downward trend when the differentiation result is smaller than the threshold value, and determines that the muscle activity is in a maintenance trend when the differentiation result is equal to the threshold value. The muscle activity estimation device according to claim 1 .

3. the muscle activity pattern calculation unit assigns a group of muscle activity pattern labels to muscle activity vectors that represent multiple muscle activity patterns at each time point; the first decoding model generation unit extracts muscle activity vectors and brain activity vectors at time groups of each muscle activity pattern label from the muscle activity pattern label group, and generates a decoding model for each muscle activity pattern using the brain activity vectors and the muscle activity vector group; the muscle activity trend calculation unit calculates a group of muscle activity trend labels corresponding to a plurality of channels; the second decoding model generation unit extracts muscle activity vectors and brain activity vectors at time groups of each muscle activity trend label from the muscle activity trend label group, and generates a decoding model for each muscle activity trend using the brain activity vectors and the muscle activity vector group. The muscle activity estimation device according to claim 1 .

4. the muscle activity calculation unit for each muscle activity pattern reads a first weighting coefficient and a first bias term for each muscle activity pattern label from the first decoding model generation unit, and calculates muscle activity for each muscle activity pattern using a decoding model corresponding to each muscle activity pattern label using the first weighting coefficient and the first bias term for the brain activity feature; the muscle activity pattern probability calculation unit calculates a probability distribution of the muscle activity pattern at each time point using a calculation model that calculates a probability distribution of the muscle activity pattern label from the brain activity feature amount; and the muscle activity calculation unit for each muscle activity trend reads a second weighting coefficient and a second bias term for each muscle activity trend label from the second decoding model generation unit, and calculates muscle activity for each muscle activity trend using a decoding model corresponding to each muscle activity trend label using the second weighting coefficient and the second bias term for the brain activity feature; the muscle activity trend probability calculation unit calculates a probability distribution of the muscle activity trend at each time point using a calculation model that calculates a probability distribution of the muscle activity trend label from the brain activity feature amount; The muscle activity estimation device according to claim 3 .

5. A computer-implemented muscle activity estimation method, comprising: a generation step in which the computer generates a decoding model for each muscle activity pattern that calculates muscle activity for each muscle activity pattern and a decoding model for each muscle activity trend that calculates muscle activity for each muscle activity trend; an estimation step in which the computer calculates muscle activity of a muscle activity pattern and a muscle activity of a muscle activity trend based on brain activity using a decoding model for each muscle activity pattern and a decoding model for each muscle activity trend, and estimates muscle activity based on the muscle activity of the muscle activity pattern and the muscle activity of the muscle activity trend; and The generating step includes: A step of dividing brain activity signals of a plurality of channels at each time into a plurality of frequency bands and calculating brain activity feature amounts; calculating muscle activity feature values ​​from muscle activity signals of multiple channels at each time; calculating a muscle activity pattern using the muscle activity feature amount; generating a decoding model for each of the muscle activity patterns using the brain activity feature amount, the muscle activity feature amount, and the muscle activity pattern; calculating multiple types of muscle activity trends for each muscle using the muscle activity feature values; generating a decoding model for each of the muscle activity trends using the brain activity feature amount, the muscle activity feature amount, and the muscle activity trend; and The estimation step A step of dividing brain activity signals of a plurality of channels at each time into a plurality of frequency bands and calculating brain activity feature amounts; calculating muscle activity for each muscle activity pattern from the brain activity feature amount using a decoding model for each muscle activity pattern; calculating a probability distribution of muscle activity patterns at each time from the brain activity feature amount; Calculating muscle activity from the perspective of muscle activity patterns from the muscle activity for each muscle activity pattern and a probability distribution of the muscle activity patterns; calculating muscle activity for each muscle activity trend from the brain activity feature amount using a decoding model for each muscle activity trend; calculating a probability distribution of muscle activity trends at each time from the brain activity feature amount; Calculating muscle activity from the perspective of muscle activity trends from the muscle activity for each muscle activity trend and a probability distribution of the muscle activity trends; calculating an average of the muscle activity from the perspective of the muscle activity pattern and the muscle activity from the perspective of the muscle activity trend, and setting the calculated average as an estimation result of the muscle activity; A muscle activity estimation method comprising:

6. A muscle activity estimation program that causes a computer included in the muscle activity estimation device to execute the functions of each component of the muscle activity estimation device according to claim 1.

Images