JP2015192870A - Intention estimation method, intention estimation program, and intention estimation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an intention estimation device capable of easily and highly accurately estimating a selection intention from spontaneous brain waves of few sites.SOLUTION: An intention estimation device 1 sets, by a preliminary measurement, two situations (a base situation and an optimal intention) that can be created by an object person and brain waves to be measured in such situation, and defines one of them as a situation representing "Yes" and the other as a situation representing "No" (steps I and II). In the Yes/No determination, the intention estimation device makes the object person create a situation representing "Yes" or "No", compares brain waves measured at that time with the brain waves of the set two situations, and estimates the selection intension of "Yes" or "No" of the object person.

Description

  The present invention relates to an intention estimation method for estimating or discriminating a human selection intention by analyzing a measured human brain wave, and more particularly, a method for appropriately setting an electroencephalogram used for estimating or discriminating a selection intention. About.

  Several methods have been proposed for measuring human brain waves and estimating or discriminating human intentions from the measured brain waves. For example, there is a method of discriminating a subject's intention from an event-related potential (eg, a component called P300 or P3) that appears in an electroencephalogram when a stimulus is given. Patent Document 1 (International Publication No. 2009/122585, “Adjustment apparatus, method and computer program for electroencephalogram identification method”) shows a measurement method of P300. Patent Document 2 (International Publication No. 2011/105000, “Electroencephalogram interface system, electroencephalogram interface providing apparatus, electroencephalogram interface execution method, and program”) describes a method of increasing intention estimation accuracy in as short a time as possible. Yes. Non-Patent Document 1 (Fujioka, Amagasaki, Fujii's “Communication Device Using EEG for Visual Stimulation”; 12th Symposium on Human Interface, Oct.22-25, 1996, Yokohama) Yes / No is discriminated using the fact that the characteristic amount of the electroencephalogram induced in the occipital region of the subject when the stimulus is shown (latency: about 120 to 140 msec) is different.

  These methods using event-related potentials utilize the electroencephalogram response when a stimulus is applied, and a device for applying the stimulus and a complicated measurement procedure are required. In addition, since the response of the electroencephalogram when a stimulus is given is weak, it is generally necessary to perform a plurality of addition averaging. Therefore, a complicated algorithm for increasing the measurement time or shortening the measurement time is required. That is, problems such as the inability of the subject to cope with the measurement procedure and the inability of complicated algorithms to deal with individual differences and noise often occur and the intent cannot be accurately identified.

  Such a problem is solved by using a natural brain wave (spontaneous brain wave). In Patent Document 3 (Japanese Patent Laid-Open No. 10-262943, “EEG measurement method and apparatus and intention transmission method and apparatus using the measurement method”), predetermined frequency components are extracted from brain waves at a plurality of locations on the scalp. Then, the brain wave topography is created by calculating the difference, and the intention is estimated from the topography pattern. Further, in Patent Document 4 (Japanese Patent Laid-Open No. 2003-325467, “intention transmission support apparatus and intention transmission support software”), when thinking about operation preparation or operation, μ wave (slightly faster than α wave 7) Based on the decrease in the amplitude of a ~ 13 Hz μ-shaped wave) and β wave (13-30 Hz fast wave appearing when nervous), the brain wave power values at multiple locations on the scalp are used. The calculated input value is input to an artificial neural network (ANN) to estimate the intention. In addition, by using the increase in the β-wave power value of the spontaneous brain wave that appears when recalling a predetermined state or a predetermined matter, the subject is made to confirm whether the determination result of the spontaneous intention is correct. I am doing so.

  These methods that use spontaneous brain waves simultaneously use multiple brain waves at the same time.For example, if the brain waves at a certain location contain noise due to the brain electrode at some location floating from the scalp, etc. Intention estimation accuracy decreases. Moreover, it takes time and labor to install the electroencephalogram electrodes at a plurality of locations, and the calculation is complicated, so that a special analysis device is required. Furthermore, since it is difficult to verify complicated calculations and artificial neural networks, it is difficult to improve when the estimation accuracy is low due to individual differences or temporal changes.

"A communication device using brain waves for visual stimulation" (Fujioka, Amagasaki, Fujii; 12th Symposium on Human Interface, Oct.22-25, 1996, Yokohama)

International Publication No. 2009/122585 International Publication No. 2011/105000 Japanese Patent Laid-Open No. 10-262943 JP 2003-325467 A

  As described above, as a method of measuring a human brain wave and discriminating or estimating a human intention from the measured brain wave, a method using an event-related brain potential which is an evoked brain wave of an intention estimation target person (subject), and an intention Those using the spontaneous brain wave of the person to be estimated are known. In addition, in the conventional method, the target of discrimination or estimation is various intentions of the intention estimation target person or the intention estimation target person's selection intention (depending on the human Yes / No). When determining alternative decision making).

  However, as a method or apparatus for discriminating the selection intention of the intention estimation target person using the spontaneous brain wave targeted by the present invention, a method or apparatus that can easily and accurately discriminate or estimate the selection intention has been proposed in the past. Not.

  In other words, when the selection intention is estimated using spontaneous brain waves, it is necessary to voluntarily recall a specific state or a specific matter from the subject, and to measure the brain waves in the recall state. Spontaneous EEGs vary greatly from person to person, even if recalling the same state and matters, even if an EEG pattern that can be clearly identified is obtained from a certain intent estimation target, In many cases, an effective electroencephalogram pattern cannot be obtained from a person. Therefore, it is conceivable to set an electroencephalogram pattern for estimating the selection intention for each intention estimation target person. However, even if the subject is the same, if the subject's mental and physical conditions, location, time, etc. are different, the reproducibility of the same pattern EEG is low even if the same state and matter are recalled. There is. For this reason, it is difficult to determine or estimate the selection intention simply and accurately using the spontaneous electroencephalogram acquired from a small part.

  In view of these points, an object of the present invention is to provide an intention estimation method, an intention estimation program, and an intention estimation device that can easily and accurately estimate a selection intention from spontaneous brain waves of a small number of parts.

  Also, an object of the present invention is to provide a setting program and a setting device for an intention estimation electroencephalogram capable of appropriately setting an electroencephalogram used for intention estimation so that a selection intention can be easily and accurately estimated from spontaneous electroencephalograms of a small number of parts. There is.

In order to solve the above problems, the intention estimation method of the present invention is:
A first measurement step of acquiring the subject's brain waves measured when the subject is made to make the first state and the second state;
The first electroencephalogram measured in the first state and the first state is set as the state and the electroencephalogram when the subject selects one of the selection intentions, and the second state and the first electroencephalogram are set. A setting step for setting the second electroencephalogram measured in the state of 2 to be a state and an electroencephalogram when the subject selects the other of the selection intentions;
A second measurement step of acquiring the subject's brain wave measured when the subject who gave the question is asked to make either the first state or the second state;
An intention determination step of comparing the brain wave measured in the second measurement step with the first brain wave and the second brain wave to determine a selection intention of the subject with respect to the question;
It is characterized by including.

  In the setting step in the intention estimation method of the present invention, the subject can be relatively easily created in a first state, for example, a relaxed rest state (hereinafter sometimes referred to as a “base state”). )) And a second state, for example, a state recalling an intention to become an electroencephalogram pattern that can be most clearly distinguished from the base state (hereinafter, sometimes referred to as “optimum intention”). At the same time, the measurement position (electrode position) of the electroencephalogram used for intention estimation may be determined. The set first and second states and the corresponding first and second electroencephalograms are associated with one and the other of the selection intentions. In the following description, “resting state” or “resting” means a state of resting open eyes (for example, a state where a bedridden person has a half eye with thin eyes open) and a state of resting closed eyes ( For example, a person who is bedridden in a closed eye state) is included, and an appropriate rest state or rest is defined according to the subject.

  In the intention determination step of the intention estimation method of the present invention, the brain state of the subject measured at the set measurement position (electrode position) is compared with the first and second brain waves, and the base state is the first state. The optimum intention that is the second state is determined. For example, the subject is informed in advance that the base state is assigned to the selection intention “No” and the optimum intention is assigned to the selection intention “Yes”, and the subject is asked a question. The subject's intention of Yes / No can be confirmed (intention estimation) depending on whether the subject is kept in the base state or recalls the optimum intention.

  Here, it is desirable that one of the first and second states is a predetermined mind-body state that can be created by the subject, particularly a state that can be easily created by the subject. The other state is preferably a recall state that recalls a predetermined matter that can be recalled by the subject, and particularly a recall state that recalls a matter that can be easily recalled by the subject.

  For this purpose, in the first measurement step, a plurality of candidate states that are candidates for one state are used as the electroencephalogram measured in at least one of the first and second states. Each of the brain waves measured when the person makes and asks the person, and in the setting step, the candidate of the state where one brain wave extracted from the plurality of measured brain waves is measured It is desirable to set the state.

  For example, in the first measurement step, for each of the plurality of candidate states, the subject's brain waves measured when the candidate is asked to make each candidate state multiple times are obtained, In the setting step, it is desirable to set the candidate state having the highest degree of coincidence between the plurality of electroencephalograms measured for each candidate as the one state. In this way, the subject can easily create a state where a reproducible electroencephalogram can be obtained can be set to the first or second state, and the accuracy of intention determination can be increased.

  In the setting step, after determining one state of the first and second states, the other state of the first and second states is changed to the determined electroencephalogram of the one state. It is desirable to set a state in which an electroencephalogram having the largest degree of separation is obtained. If it does in this way, the state from which the brain wave which can be clearly identified among the set of states which a subject can make can be set to the 1st, 2nd state. Thereby, the precision of intention determination can be raised.

Since the electroencephalogram changes depending on the physical condition and mood of the subject, the intention estimation is performed after a predetermined time elapses after the determination of the first state (base state) and the second state (optimum intention), for example, It is desirable to do it within at least one day. In addition, it is possible to easily verify the estimation accuracy by repeatedly asking questions whose answers are reversed in the same meaning, or by asking questions that only the target person or related parties know. If the estimation accuracy is low, the estimation accuracy can be increased by resetting the first state (base state), the second state (optimum intention), and the like.

  Next, in the intention estimation method using an electroencephalogram of the present invention, in order to be able to perform intention estimation more easily, more accurately, and in a short time, the first state is set to the rest state, and the second state It is desirable that the state is a state in which the subject is reminded of a predetermined intention. For example, the rest state (base state) may be “No” as the selection intention, and the second state (optimum intention) may be “Yes”. In the second state, an item that the subject can easily recall may be selected from a plurality of candidates. In addition, it is easy to set the measurement position of the electroencephalogram to one place (one channel). For that purpose, the first electroencephalogram obtained in the rest state, which is the first state, and the optimum intention, which is the second state, are set. It is desirable to determine the measurement position having the greatest degree of separation from the second brain wave obtained in the recalled state as the optimum measurement position and measure the brain wave at the position to perform intention estimation.

  In the second and subsequent intention estimations for the same subject, it is desirable to select the optimal intention that has been set as the second state most frequently so far as candidates for the second state. Thereby, the electroencephalogram measurement time in a 1st measurement process can be shortened. Moreover, if the candidate is set as the second state in the setting step, the process can move to the second measurement step for intention estimation in a short time.

  Further, in the setting step of setting the second state, by setting the optimum intention to be adopted as the second state using a threshold value set based on the intention estimation accuracy predicted in advance, for example, When an estimation accuracy of 80% or more is expected, it is desirable to shift to the second measurement process and the intention determination process, which are intention estimation processes.

  Further, the measured electroencephalogram may include a noise component resulting from the subject's blinking. Therefore, in order to remove the noise caused by blinking and increase the accuracy of intention estimation, the frequency band of the brain wave used for intention estimation is set to, for example, 8 to 30 Hz, which is a frequency band in which blinking frequencies are not easily included. Is desirable. When there is no need to consider the noise caused by blinking, the frequency band of the brain wave used for intention estimation can be expanded and set to, for example, 4 to 30 Hz.

  In addition, in order to shorten the intention estimation time, it is desirable to execute the intention determination step using the measured electroencephalogram for each electroencephalogram measurement in the second measurement step or for each separately set period. .

  On the other hand, in the case of a bedridden target person or the like, it may not be possible to visually determine whether the target person is in an active state or a sleep state, and communication with the target person may not be achieved by intention estimation using brain waves. Therefore, it is desirable to continuously measure the brain wave of the subject, confirm that the subject is in an active state based on the measured brain wave pattern, and attempt communication by intent estimation using an electroencephalogram in the active state. In addition, when the activity state is not recognized from the electroencephalogram, the subject was given an external stimulus at approximately the same time every day, for example, to adjust the circadian rhythm of the subject and induced to become active around that time In addition, it is desirable to perform communication based on intention estimation based on brain waves.

  In the present invention, intention estimation is performed using brain waves. However, in some cases, physiological indices other than brain waves can be measured in addition to brain waves, and intention estimation can be performed based on the brain waves and physiological indices. It is. In this case, the measured electroencephalogram and physiological index can be treated in the same way, for example, each may be normalized to mean 0 and variance 1.

  In the present invention, first, for each subject, two types of spontaneous electroencephalograms used for intent estimation by measuring electroencephalograms, that is, subjects from which spontaneous electroencephalograms indicating one and the other of the selection intention (Yes / No determination) are obtained. First and second states (base state and optimum intention) of the person are determined. By appropriately setting the first and second states according to the subject, it becomes possible to accurately identify the spontaneous brain waves measured when these states are created by the subject. In particular, by setting a resting state as the first state and setting an optimal intention that is most easily recalled by the subject as the second state, it becomes possible to accurately identify the spontaneous brain waves. Therefore, the selection intention of the subject can be estimated with high accuracy by using two types of spontaneous brain waves set for each subject. Also, since two types of identifiable spontaneous electroencephalograms are used, the selection intention estimation operation with high reproducibility is guaranteed without being greatly affected by the subject's mind-body condition and the external environment.

  Further, according to the present invention, two types of spontaneous electroencephalograms that are most suitable for each target person can be used for estimation of selection intention, so intention estimation is performed with high accuracy using electroencephalograms measured at a small number of sites. be able to. Therefore, unlike the case where electrodes for electroencephalogram measurement are arranged at a large number of electroencephalogram measurement positions and a large number of wirings are performed, the time and labor for measurement preparation can be reduced. In addition, since it is not necessary to use a large number of electroencephalograms measured at a large number of sites, calculations and algorithms for intention estimation or discrimination can be simplified. Therefore, it is possible to realize an intention estimation device that can perform simple and high-precision intention estimation using a general-purpose PC.

  Furthermore, according to the present invention, high-precision intention estimation (intention determination) that is not affected by noise or the like can be performed. For example, even if there is a possibility that the electroencephalogram of a part of the plurality of electroencephalogram measurement parts floats from the scalp, the electroencephalogram of that part may contain noise, and there is one part that can correctly measure the electroencephalogram. However, there is a possibility that the intention can be estimated.

  In addition to this, it is possible to easily verify the intention estimation result by repeatedly asking questions with opposite answers in the same meaning, or by asking questions that only the target person or related parties understand. If the estimation accuracy is low, the estimation accuracy can be easily increased by resetting the first assumed content (base state), the second assumed content (optimum intention), and the electroencephalogram measurement site (electrode position). it can.

It is the schematic block diagram which shows the whole structure of the intention estimation apparatus which concerns on Embodiment 1 to which this invention is applied, and the schematic flowchart which shows operation | movement. It is a flowchart which shows the procedure until the data measurement start in the preliminary measurement by an intention estimation apparatus. It is a flowchart which shows the procedure of the data measurement in the preliminary measurement by an intention estimation apparatus. It is a flowchart which shows the procedure of the data analysis of the preliminary measurement by an intention estimation apparatus. It is a flowchart which shows the procedure of the data analysis of the preliminary measurement following FIG. 3A. It is a flowchart which shows the procedure of the data analysis of the preliminary measurement following FIG. 3B. It is a flowchart which shows the calculation procedure of the average of the power ratio of each frequency band of the base state of step ST31 of FIG. 3A. It is a flowchart which shows the procedure of the intention determination by the intention estimation apparatus. It is a flowchart which shows the procedure of the intention determination following FIG. 5A. It is a graph which shows an example of the verification experiment of intention estimation. It is a graph which shows an example of the verification experiment of intention estimation. It is a graph which shows an example of the verification experiment of intention estimation. It is a flowchart which shows the procedure of the preliminary measurement process in Embodiment 2 to which this invention is applied, and a setting process. It is a flowchart which shows the modification of the procedure shown in FIG. 10 is a flowchart illustrating a procedure of an intention estimation mode (main measurement process and intention determination process) in the second embodiment. It is a flowchart which shows the procedure in the case of performing intention estimation using an electroencephalogram and a physiological index. It is a graph which shows the relationship between RMSrest_max / SDall_ave and intention estimation precision. It is a graph which shows distribution of the frequency | count of the correct answer of intention estimation with respect to RMS (RMSno_i and RMSyes_i), and an incorrect answer. It is a block diagram which shows the example of a home medical treatment support system which can apply the intention estimation apparatus of this invention.

  Embodiments of an intention estimation apparatus to which the present invention is applied will be described below with reference to the drawings. The target person whose intention is estimated using the intention estimation apparatus according to the present embodiment is assumed to be in a situation where the perceptual functions such as vision and hearing are maintained, but the intention expression such as utterance and facial expression cannot be performed at all. doing. In other words, although instructions and questions can be given by voice or screen, the responses and answers cannot be obtained except for physiological responses such as brain waves.

[Embodiment 1]
(Intention estimation device)
FIG. 1A is a schematic block diagram showing the overall configuration of the intention estimation apparatus according to Embodiment 1, and FIG. 1B is a schematic flowchart showing the operation of the intention estimation apparatus. As shown in FIG. 1 (a), the intention estimation apparatus 1 also functions as an intention estimation brain wave setting apparatus, and an electroencephalograph 2 that is an electroencephalogram measurement unit for measuring the brain waves of the subject P, and a personal computer ( PC) 3, an input device 4 such as a keyboard and a mouse, and a display device 5 such as a monitor. The PC 3 includes a CPU, a ROM, a RAM, and the like, and executes an intention estimation program including an intention estimation brain wave setting program stored in the ROM or the like, whereby a preliminary measurement unit (first measurement unit) and a main measurement are performed. It functions as an electroencephalogram acquisition unit 6 as a unit (second measurement unit), an electroencephalogram analysis unit 7 as a setting unit and an intention determination unit, a display control unit 8 that drives and controls the display device 5, and the like. And the setting operation | movement which sets information, such as a brain wave for intention estimation, for every subject in the memory | storage part 9 is performed. Based on the information set in the storage unit 9, the brain wave analysis unit 7 performs an intention estimation operation (Yes / No determination operation) of the subject.

  As shown in FIG. 1 (b), in the setting operation of the intention estimation brain wave by the intention estimation apparatus 1, an electroencephalogram 2 measures an electroencephalogram generated when a subject is caused to recall a plurality of different recall contents, The measured electroencephalogram information is taken into the PC 3 via the electroencephalogram acquisition unit 6 (step I: preliminary measurement process). The electroencephalogram captured by the PC 3 is sent to the electroencephalogram analysis unit 7, and the electroencephalogram information is analyzed, and a pair of electroencephalograms in which changes that can be distinguished from each other occur are used for estimation of the selection intention. Two brain waves are extracted. Moreover, the state of the subject from which each of the extracted first and second electroencephalograms is measured is extracted as a first state and a second state used for intention estimation, respectively. In the storage unit 9, the extracted first electroencephalogram and the first state are set as information used for estimating one of the selection intentions (for example, Yes). Further, the extracted second electroencephalogram and the second state are set as information used for estimating the other second intention (for example, No) of the selection intention (step II: setting step).

In the intention estimation operation, a question is asked to the target person, and the selection intention (Yes / No intention) for the question is recalled to the target person in one of the first state and the second state corresponding to the intention to answer. Then, the electroencephalogram measured in this state is taken into the PC 3 (step STIII: main measurement process). In the electroencephalogram analysis unit 7, it is analyzed whether the captured electroencephalogram matches the first electroencephalogram or the second electroencephalogram (step IV: intention determination step). The estimation intention as an analysis result is displayed on the screen of the display device 5 via the display control unit 8.

  In the present embodiment, the brain wave of the subject is measured at two sites of Cz and Fz in the electrode arrangement for electroencephalogram measurement based on the international 10-20 method shown in FIG. Needless to say, electroencephalogram measurement at other sites, or electroencephalogram measurement at three or more sites. In addition, the electrode positions of A1 and A2 of the left and right earlobe are reference electrode parts, and the electroencephalogram of each part is obtained as a potential difference between the electrode attached to each electrode part and the reference electrode.

(Preliminary measurement / setting process)
FIG. 2A is a schematic flowchart showing a flow of operations up to the start of data measurement in preliminary measurement of the intention estimation apparatus 1. After attaching the electrodes of the electroencephalograph 2 to the subject, the intention estimation device 1 is activated (step ST1). When the intention estimation device 1 is activated, an initial setting screen (not shown) is displayed on the display device 5 (step ST2). On the initial setting screen, the brain waves of the subject measured by the electroencephalograph 2 are displayed in real time. The measurer can confirm that the electroencephalogram can be normally measured by looking at the initial setting screen. However, since there is noise that cannot be determined on the screen, the noise detection method described later is employed (see FIG. 3A, ST25 and ST27).

In the initial setting screen, the following three types of candidates are displayed as defaults as candidates for the base state, which is the first state serving as a reference for determination of intention estimation.
(1) Relaxed rest (eg, relaxed calmly without thinking)
(2) Excited high arousal state (eg, something very angry and frustrated)
(3) State in which calculation problem is concentrated (eg, repeat calculation by subtracting 7 from 100)

  The measurer selects any one of the candidates (1) to (3) (step ST3) and verbally conveys it to the subject. Normally, candidate (1) is selected. However, when it is desired to determine objectively and quantitatively, for example, when one candidate is presented a plurality of times and the fluctuation of the electroencephalogram is small (if it falls within a predetermined fluctuation range set in advance), it is decided. Instead of this, a plurality of candidates may be presented at a fixed number of times at random, and the candidate with the smallest fluctuation of the electroencephalogram may be selected. A method for determining whether this variation is large or a method for selecting a candidate having the smallest variation will be described later (described in paragraph number 0053).

  When any of candidates (1) to (3) is selected on the initial setting screen, the screen shifts to a preliminary measurement screen (not shown) for setting the brain wave for estimating the intention of the subject (step ST4). . In the case of the present embodiment, “Preliminary measurement is performed from now on, and intention estimation (Yes / No determination) is performed using the result” on the upper part of “Measurement start button” of preliminary measurement. Intent estimation is also performed with the eyes open, so never close your eyes. However, blinking does not matter. "" Preliminary measurement will be completed in about 10 minutes. " After the countdown, the instruction is displayed for 5 seconds.After that, please keep the state or recall the intention according to the instruction for 20 seconds. This cycle is repeated 20 times. " These are shown to the subject and explained verbally by the measurer. When the measurement start button displayed on the preliminary measurement screen is pressed (step ST5), data measurement (preliminary measurement) is started (step ST10).

In the case of the present embodiment, six types of intentions 1 to 6 are prepared as defaults as candidates for the second state recalled by the subject in the preliminary measurement. Then, the following 20 instructions are displayed in a random order.
(Instruction 1) Base state (Instruction example: Please relax and rest): 8 times (Instruction 2) Intent 1 (Instruction example: Please image sour taste): 2 times (Instruction 3) Intent 2 (instruction example: please insist on your thoughts): 2 times (instruction 4) Intent 3 (instruction example: please take a bathing and relaxing image): 2 times (instructions 5) Intent 4 (instruction example: please image the right hand moving): 2 times (instruction 6) Intent 5 (instruction example: please image that calm music is heard): 2 times (instruction 7) Intent 6 (Instruction example: Please image chewing): 2 times

  FIG. 2B is a schematic flowchart showing a data measurement processing procedure after the measurement start button on the preliminary measurement screen is pressed. When the measurement start button is pressed, first, a countdown “5, 4, 3, 2, 1” is displayed for one second (step ST11), and an instruction is displayed for the next five seconds (step ST12). The subject looks at this preliminary measurement screen, and at the same time, the measurer also gives an oral instruction. For the next 20 seconds, the screen disappears, and the subject is reminded of the recall contents in accordance with the instructions. The brain wave measured during this time is captured (step ST13). The total time of steps ST11 to ST13 is 30 seconds, which is one cycle of instruction. Thereafter, the instruction contents are sequentially switched in a random order and repeated 20 times in total (steps ST11 to ST14). When 20 instructions have been completed, that is, when the base state maintenance has been performed 8 times and 6 types of intentions have been recalled 2 times each, a total of 20 electroencephalogram measurements have been completed. (Step ST20) is started.

  The electroencephalogram is continuously recorded at a sampling frequency of 128 Hz from when the measurement start button is pressed. The data to be analyzed is 16 seconds (2048 data) in the central portion of 20 seconds (2560 data) in which the screen disappears and the subject maintains the base state in accordance with the instruction or performs intention recall.

(Setting process: data analysis)
3A, 3B, 3C and FIG. 4 are schematic flowcharts showing the flow of data analysis (electroencephalogram analysis) for analyzing the electroencephalogram acquired by the preliminary measurement. Data analysis is first performed from Cz (steps ST21 and ST22). Then, the eight base states are sequentially analyzed (steps ST23 to ST30).

  Here, when the analysis proceeds, if a part of the measured amplitude of the electroencephalogram includes a part that deviates from the preset amplitude width, the data measured at that part is not analyzed. For example, if there is a portion with a value of ± 50 μV or more, it is determined that noise is mixed in that portion, and data measured at that portion is not analyzed (step ST25).

In the data analysis, first, FFT (Fast Fourier Transform: Fast Fourier Transform) is performed on the measured electroencephalogram, a power spectrum (PS) is calculated, and a power value of each frequency band is calculated (step ST26). When a part of the power spectrum exceeds a preset value, for example, when the value is 2500 μV ² or more, it is determined that noise is mixed in the part, and data measured at the part Does not perform the subsequent analysis (step ST27).

In this embodiment, the frequency band of the measurement electroencephalogram used for analysis, that is, the entire frequency band to be analyzed is set to 0 to 30 Hz, the delta wave included in the entire frequency band is 0 to 3.5 Hz, and the theta wave is 3 .5 to 7.5 Hz, alpha wave was set to 7.5 to 13.5 Hz, and beta wave was set to 13.5 to 30 Hz. In addition, as a result of preliminary studies so far, there are many examples in which the power value of 8 to 11 Hz varies depending on the intention. Therefore, this band is named a new wave, and in this embodiment, a new wave of 8 to 11 Hz is also set. As a result, the band of the new wave overlaps with the alpha wave, and only this band is calculated twice. However, since it is a band that has a large response to the intention, it is assumed that the accuracy of intention estimation is improved. However, it is not always necessary to make the same setting as in this embodiment.

Next, the following index is calculated from the power value of each frequency band (step ST28).
A: Delta wave power ratio = Delta wave power value / Full frequency power value B: Theta wave power ratio = Theta wave power value / Full frequency power value C: Alpha wave power ratio = Alpha wave power value / Full frequency power value D: New wave power ratio = New wave power value / Full frequency power value E: Beta wave power ratio = Beta wave power value / Full frequency power value

  After all the power ratios (An, Bn, Cn, Dn, En: n = 1 to 8) have been calculated for the eight base states, the respective average values are calculated (steps ST29 → ST31). Each average value is described as B_Aave, B_Bave, B_Cave, B_Dave, B_Eave. However, this calculation method is not a simple average but performs the processing shown in FIG.

  A method for calculating the average value of the power ratio will be described with reference to the flowchart of FIG. First, when all power ratios are calculated for eight base states (steps ST32 and ST33), the average and variation of n (= 8) base states for each of A, B, C, D, and E. SD (Standard Deviation: Standard Deviation) is calculated (step ST34), then "average ± 2SD" is calculated, and even one of A, B, C, D, and E is out of this range. Is not used for calculating the average value (step ST35). Then, excluding the excluded base state, “average ± 2SD” is calculated again, and this process is repeated (steps ST34 to ST38).

  In general, it is assumed that the base state cannot be properly created due to external stimuli or arousal reduction, and at that time, all or one is significantly different from A, B, C, D, E when properly created. It is expected that. Therefore, if any one of A, B, C, D, and E deviates from the “average ± 2SD”, it is determined that the base state could not be properly created, and A, B, C, D, and E at that time Were all excluded from the mean calculation. In addition, when the number of base states remaining after such exclusion is 4 or less, it is determined that the fluctuation of the base state is too large, and the data measured at that part is analyzed after this Is not performed (YES in step ST38). Eventually, B_Aave, B_Bave, B_Cave, B_Dave, and B_Eave are calculated when there are five or more base states that survive the above processing (step ST39). However, when this calculation is performed, SD is calculated for each of A, B, C, D, and E, and if the average value is 0.05 or more, the fluctuation of the base state is too large. Even in this case, the data measured at that part is not analyzed thereafter (YES in step ST40).

Next, returning to the flowchart of FIG. 3B, for the intentions 1 to 6 recalled by the subject twice each time, if it is determined that noise is not mixed in the measured brain wave, the difference between the two times RMS (Root Mean Square) is calculated (steps ST41 to ST46). The RMS of intention 1 is calculated as follows.
I1_rms
= (((I1_1_A−I1_2_A) ² + (I1_1_B−I1_2_B) ²
+ (I1_1_C−I1_2_C) ² + (I1_1_D−I1_2_D) ²
+ (I1_1_E−I1_2_E) ² ) / 5) ^1/2

  Here, I1_rms is the RMS of intention 1, and I2_rms is the RMS of intention 2. I1_1_A is A when recalling intention 1 for the first time, and I1_2_A is A when recalling intention 1 for the second time. Similarly, I1_1_B is B when recalling intention 1 for the first time, and I1_2_B is B when recalling intention 1 for the second time. Similarly, I2_rms, I3_rms, I4_rms, I5_rms, and I6_rms are also calculated (steps ST42 to ST49).

Next, the average values of A, B, C, D, and E are calculated for intentions 1 to 6 recalled twice. Here, the average value of the power ratio in each frequency band of intention 1 is described as I1_Aave, I1_Bave, I1_Cave, I1_Dave, I1_Eave. Similarly, the average value in intention 2 is described as I2_Aave, I2_Bave, I2_Cave, I2_Dave, and I2_Eave. Then, the RMS of the difference between these and B_Aave, B_Bave, B_Cave, B_Dave, B_Eave is calculated. For example,
I1_B_rms
= (((I1_Aave−B_Aave) ² + (I1_Bave−B_Bave) ² + (I1_Cave−B_Cave) ² + (I1_Dave−B_Dave) ²
+ (I1_Eave−B_Eave) ² ) / 5) ^1/2
It becomes. Here, I1_B_rms is the RMS of the difference between the average value of intention 1 and the average value of the base state. Similarly, I2_B_rms, I3_B_rms, I4_B_rms, I5_B_rms, and I6_B_rms are also calculated (step ST50).

  Finally, the time when “In_B_rms−In_rms: n = 1 to 6” is the largest among all the intentions of all the sites is selected (step ST51). For example, a value obtained by subtracting the RMS “I1_rms” of the difference when recalling the intention 1 twice from the RMS “I1_B_rms” of the difference between the average value of the intention 1 and the average value of the base state is calculated. In this case, this value is Cz1. Similarly, Cz2, Cz3, Cz4, Cz5, and Cz6 are also calculated, and the maximum value among them is set to MAX_Cz. Similarly, MAX_Fz is also calculated (steps ST52, 53 → steps ST22 to ST51).

  After this, if MAX_Cz and MAX_Fz are compared and the larger one is selected, there is a possibility that the intention at that part can be used for Yes when determining Yes / No. Here, one or both of MAX_Cz and MAX_Fz may not be calculated due to noise mixing. Even if both MAX_Cz and MAX_Fz are present, if the larger of them is not a positive number, the fluctuation of the EEG when the intention is recalled several times in all intentions is Therefore, it is determined that it cannot be accurately distinguished from the base state no matter what intention is determined to be Yes. The same applies when only one of MAX_Cz and MAX_Fz exists and the value is not a positive number. In these cases, including the case where both MAX_Cz and MAX_Fz cannot be calculated, the site (electrode position) and intention at the time of Yes cannot be determined. In such a case, the initial setting and the preliminary measurement are performed again, or the Yes / No determination is abandoned.

  As described above, when the electrode position and intention of Yes can be determined, when both MAX_Cz and MAX_Fz exist and the larger one is a positive number (step ST54 → ST55 in FIG. 3C), or MAX_Cz and MAX_Fz Only one of them exists and the value is a positive number (step ST56 → ST57 in FIG. 3C). In other cases, the initial setting and the preliminary measurement are performed again (step ST59 → ST61 in FIG. 3C), or the Yes / No determination is abandoned and the process is terminated (step ST59 → ST60 in FIG. 3C).

In this embodiment, each intention is recalled twice in the preliminary measurement, but it may be recalled three or more times to check the fluctuation. In that case, the calculation similar to the above may be performed by setting the variation index to SD instead of RMS.

  When changing the base state at the time of initial setting, any state may be used, but a state expected to be easily created by the subject is selected. Further, in order to determine whether or not the fluctuation of the generated electroencephalogram in the base state is large, SD is calculated for each of A, B, C, D, and E measured multiple times, and the average value of SD is 0.05. If it is smaller, it may be determined that the fluctuation is small. Further, a plurality of types of base state candidates may be created a predetermined number of times, and the candidate having the smallest SD average value may be selected. This is a method for determining whether or not a candidate's electroencephalogram fluctuation is large, and a method for selecting a candidate having the smallest fluctuation.

  Moreover, when changing an electrode position, it is not necessarily based on the international 10-20 method. However, for example, when it is desired to detect the left-right difference of the electroencephalogram, such as Yes is an image of moving the left hand and No is an image of moving the right hand, it is desirable to include symmetrical electrode positions (for example, C3 and C4).

  Furthermore, when changing the optimal intention candidate, it is desirable to include an intention to activate consciousness and an intention to relax. Further, if the base state is a relaxed state, it is desirable to increase the intention to activate. Conversely, if the base state is active, it is desirable to increase the intention to relax. In this example, in addition to the above, being massaged, moving the left hand, smelling good, seeing beautiful flowers, observing images finely, having a sweet taste, gently Candidates can be displayed, such as talking to a person, performing mental arithmetic quickly by subtracting 7 from 100, hearing a loud sound continuously, urinating in the toilet, and the like.

  In this embodiment, when the target person's intention estimation is always performed, the initial setting and the preliminary measurement are performed once a day on the assumption that the electroencephalogram changes with the state change. Yes.

(Intent determination process: Yes / No determination)
Next, referring to the flowcharts of FIGS. 5A and 5B, the first state (base state indicating No) and the second state (Yes) determined in the above manner and set in the storage unit 9. The operation for determining the selection intention of the target person will be described.

  When the intention of Yes is successfully determined in the present embodiment, the time required until the intention of Yes is displayed on the screen of the display device 5 after the preliminary measurement is completed is about 3 seconds. The specific display contents (not shown) at that time are “Yes / No judgment button” and on the button, “The question will be spoken verbally. If the answer to the question is No, XXXXXX ( (Sentence requesting to create a base state) .If yes, YYYYY (sentence requesting to recall the intention of Yes determined by preliminary measurement). Please give me 20 seconds. "

  In the Yes / No determination process, first, the measurer asks the subject about any question (step ST71), and immediately presses the “Yes / No determination button” (step ST72). The subject then creates a base state if the question is No for 20 seconds, or recalls the intention determined as Yes if Yes (step ST73). During this time, the brain waves are continuously recorded at 128 Hz for 20 seconds (2560 data), but the central portion of 16 seconds (2048 data) is selected (step ST74), and the analysis is performed as described below for these. Is called.

First, when it is determined that noise is not mixed in the electroencephalogram of the site used for Yes / No determination (steps ST75, ST76, ST77), the power value of each frequency band is calculated and the power ratio (A, B, C, D, E) are calculated (step ST78). Then, the average power ratio (B_Aave, B_Bave, B_Cave, B_Dave, B_Eave) of each frequency band in the base state representing No, and the average power ratio of the intention to be recalled when Yes (tentative 1 is set to Yes) In this case, the RMS is calculated between I1_Aave, I1_Bave, I1_Cave, I1_Dave, and I1_Eave) and the power ratio (A, B, C, D, E) calculated this time (steps ST79 and ST80). .

That is, when the RMS of the difference between the base state and the brain wave measured this time is No_RMS, Yes, and the RMS of the difference between the brain wave measured this time and the intention measured at this time is Yes_RMS, the following equation is obtained.
No_RMS
= ((((B_Aave−A) ² + (B_Bave−B) ²
+ (B_Cave−C) ² + (B_Dave−D) ²
+ (B_Eave−E) ² ) / 5) ^1/2
Yes_RMS
= (((I1_Aave−A) ² + (I1_Bave−B) ²
+ (I1_Cave−C) ² + (I1_Dave−D) ²
+ (I1_Eave−E) ² ) / 5) ^1/2

  The average power ratio of the base state representing No and the average power ratio of the intention representing Yes in the first embodiment are indices corresponding to average patterns Tno and Ties in the second embodiment described later, respectively. Further, No_RMS and Yes_RMS in the first embodiment are indices corresponding to RMSno_i and RMSyes_i in the second embodiment described later, respectively.

  Next, if the difference between Yes_RMS and No_RMS is less than 0.01, it is determined that there is not a sufficient difference for performing Yes / No determination, and the determination is suspended (steps ST81 → ST85). If this difference is 0.01 or more, No_RMS and Yes_RMS are compared, and the smaller one is displayed as the determination result (steps ST81, ST82, ST83, ST84). If noise is mixed in the electroencephalogram, it is displayed that determination is impossible (step ST86). When the determination (estimation) is repeatedly performed, the process returns from step ST87 to step ST71, and otherwise, the process ends.

  In this embodiment, it is possible to easily verify the estimation accuracy by repeatedly asking questions having the same meaning and the opposite answers. For example, if the answer to the question “I am hungry?” Is “Yes”, it should be “No” to the question “I am not hungry?”. If the answers to these two questions are switched between Yes and No, it can be considered that the determination has been made accurately. However, since such negative questions are not frequently used on a daily basis, there is a possibility that the subject may answer by mistake. In such a case, it can be verified by asking a question that only the target person and his family know.

  If such verification is performed and it is determined that the determination accuracy is low, initial setting and preliminary measurement are performed again. However, if the determination accuracy is not improved even after performing the initial setting and the preliminary measurement again, it is expected that the subject is sleeping, has strong pain, etc., and cannot respond to the instruction or question. In such a case, it is desirable to once complete the Yes / No determination and start over from the initial setting after at least several hours have elapsed. In addition, when noise is mixed in an electroencephalogram, it is necessary to take measures such as removing surrounding noise sources and reattaching electrodes.

(Verification experiment)
FIGS. 6A, 6B, and 6C show experimental examples as Tables 1 to 3 in which Yes / No determinations of three subjects were performed using the intention estimation system developed by the inventors. For all three subjects, the resting state was set to the base state by default, and then preliminary measurement was performed. Tables 1 to 3 list not only the results but also the following values.
Average SD for each power ratio in the base state
RMS of differences when recalling each intention twice
RMS of the difference between the average value of each intention and the average value of the base state

  Moreover, in Yes / No determination, the intention recall of Yes was performed 6 times and No base state creation was performed 6 times at random. However, since this was carried out as a verification experiment, six intentions were recalled at the time of six Yes intention recalls, and the Yes determination results in the table were listed together in one column. Also, each of the six intentions is used for the determination of No. Therefore, in each part (electrode position), the result of performing Yes / No determination using each intention as Yes, the correct answer rate is determined by how many times the answer is correct among the seven times of execution.

  First, when the subject A in Table 1 shown in FIG. 6A makes a Yes / No determination according to the present method, the Fz electroencephalogram when moving the right hand is set to Yes. At this time, the number of correct answers was 7/7, and the correct answer rate was 100%. Next, when subject B in Table 2 of FIG. 6B makes a Yes / No determination according to the present method, the Cz electroencephalogram when taking an image of taking a bath and relaxing is set to Yes. At this time, the number of correct answers was 7/7, and the correct answer rate was 100%.

  Next, when subject C in Table 3 of FIG. 6C makes a Yes / No determination according to the present method, in Fz, the average SD after one out of the eight base states is 0. Since it is 084 and is 0.05 or more, it cannot be determined originally. It is considered that the base state was not stable for some reason. Moreover, noise was mixed in Cz and calculation was impossible. Here, for verification, Yes / No determination was performed by setting the Fz electroencephalogram to Yes when taking a bathing and relaxing image. At this time, the number of correct answers was 6/7, and the correct answer rate was 86%. Since the accuracy rate did not reach 100%, it was inferred that the base state was stable.

  As a result of the above verification, it has been clarified that when the Yes / No determination is performed by this method, the correct answer rate when the brain waves can be measured correctly is 100%. However, since the electroencephalogram fluctuates according to changes in mood and environment, it is not always possible to maintain a 100% accuracy rate. In the present invention, since the determination accuracy can be easily verified, it can be said that, as a result, it can be used without any problem in practice by advancing Yes / No determination while verifying the determination accuracy.

[Embodiment 2]
The intention estimation apparatus and method according to Embodiment 2 to which the present invention is applied will be described below. The overall configuration and procedure of the intention estimation apparatus and method of the second embodiment are the same as those of the first embodiment. That is, the overall configuration of the intention estimation device is the same as that of the intention estimation device 1 shown in FIG. 1A, the intention estimation procedure is the same as the procedure shown in FIG. 1B, and until the start of data measurement in the preliminary measurement process. The procedure is the same as the procedure shown in FIG. 2A.

In the second embodiment, the electroencephalogram measurement site is only one channel. As in the case of the first embodiment described above, it is possible to measure at two or more sites. When the measurement site is one channel, a channel that is predicted to have the highest estimation accuracy is selected from a plurality of channels at the time of preliminary measurement. A specific example of this selection method is the RMS of rest and intention (intention recall state), as in the case of selecting the intention max shown in FIG.
Select the channel with the largest.

  Further, in the second embodiment, the first state when the selection intention is No is set in advance to a state of resting and not recalling anything (hereinafter sometimes simply referred to as “resting”). The intention to be recalled at the time of Yes is determined through a preliminary measurement process and a setting process. Of course, as in the case of the first embodiment, it is possible to set the base state and the optimum intention through the preliminary measurement step and the setting step, and to assign No and Yes to the base state and the optimum intention.

  Further, also in the second embodiment, the coping method in the case of the first embodiment can be adopted as the coping method when the measured brain wave noise is mixed. In addition to this, in Embodiment 2, for noise caused by blinking of the subject, in order to remove the noise component due to blinking, a component (for example, 8 to 30 Hz) having a frequency band of 8 Hz or more in the measured electroencephalogram is used. I use it. Depending on the subject, it may not be necessary to consider noise caused by blinking. In this case, a component (for example, 4 to 30 Hz) including a frequency band of 8 Hz or less is used in the measured electroencephalogram.

(Preliminary measurement process and setting process)
FIG. 7 is a flowchart showing the flow of preliminary measurement, which is the procedure of the preliminary measurement process and the setting process in the second embodiment, and shows the first case. FIG. 8 is a flowchart showing the flow of preliminary measurement in the case of intention estimation for the second and subsequent times for the same target person.

  In the process of FIG. 7, a teacher signal (electroencephalogram pattern) representing rest No in the first state and a teacher signal (electroencephalogram pattern) representing Yes of the intention in the second state are set. The time spent for this setting process has been shortened by devising each procedure, and can be processed in about 6.5 minutes.

  Further, when setting the teacher signal for the second time or later for the same target person, as shown in steps ST120 and ST121 of FIG. 8, the intention set to “Yes” most frequently so far is a candidate intention representing “Yes”. The intention max is selected. Thereby, the setting operation of the intention max in the preliminary measurement (steps ST100 to ST106 in FIG. 7) can be omitted. For example, the processing time of about 6.5 minutes can be shortened to about 3.5 minutes.

  In the second and subsequent times, the process shown in steps ST107 to ST110 in FIG. 8 is omitted, and the teacher signal (an electroencephalogram pattern) of the intention set to Yes most so far, and the rest teacher set at that time The signal (electroencephalogram pattern) can be used as it is. In this case, the measurement and calculation processes in steps ST100 to ST110 in FIG. 7 are not necessary, and the processing time can be greatly shortened. For example, the processing time which takes about 6.5 minutes can be made substantially zero.

  Furthermore, in the second embodiment, the intention estimation accuracy is automatically predicted, and when an estimation accuracy of 80% or more is expected, the mode is shifted to the intention estimation mode (“RMSrest_max in step ST110 of FIGS. 7 and 8). / SDall_ave> TH3? ").

7 and 8 will be described. First, in steps ST101 to ST105, five intentions (i = 1 to 5) are presented as candidates to the subject and recalled. Defaults are prepared for the five intentions to be presented. However, it is only necessary that the settings can be freely changed to intentions closely related to the subject's hobbies and memories. In general, hobbies (eg: playing the piano) and memories (eg: climbing) are easy to recall. Examples of five intentions of the first state (rest) and the second state prepared as defaults are shown below.
No: Rest (do not think anything)
Yes candidate intention:
(1) An image that tastes sour (2) An image that insists on my thoughts (3) An image of bathing and relaxing (4) An image of moving the right hand (5) I hear my favorite music Image

  The brain wave to be analyzed when recalling the intention is 16 seconds (2048 data measured at 128 Hz). The measurement procedure per intention was a total of 30 seconds including presentation 5 seconds, intention recall 20 seconds, and break 5 seconds, and the analysis target was the central portion 16 seconds of intention recall 20 seconds. Note that, in the preliminary measurement process of the second embodiment, in addition to actually presenting the intention and measuring the electroencephalogram, about 30 seconds are expected as the time for screen switching and operation time.

  Further, since the noise due to blinking can be specified to be less than about 8 Hz, the frequency band of the measured electroencephalogram used for estimating the intention of the subject (subject) who can blink is set to 8 to 30 Hz. This frequency band of 8 to 30 Hz is the entire frequency band used for intention estimation. As will be described below, the power value of this entire frequency band is calculated as the total frequency power value. In addition, the entire frequency band is divided into three sections, a slow α wave (8 to 9.5 Hz), a fast α wave (9.5 to 13 Hz), and a β wave (13 to 30 Hz), and the power of each frequency band. The value is calculated. On the other hand, in the case of a subject who cannot blink, the frequency band of the electroencephalogram used for intention estimation is set to 4 to 30 Hz. Actually, each power value in the frequency band of θ wave (4 to 8 Hz), α wave (8 to 13 Hz), and β wave (13 to 30 Hz) is used. The total frequency power value in this case is the power value of the entire frequency band of 4 to 30 Hz of the measured electroencephalogram. According to the experiments by the present inventors, if the analysis processing is performed by dividing into three bands in this way, the intention estimation accuracy can be improved, and if the analysis processing is performed by dividing into more bands than this, the amount of calculation is increased. It was confirmed that the accuracy of intention estimation cannot be increased for a large number of.

  In step ST106 of FIG. 7 and FIG. 8, the calculation of RMSrest_i, which is the RMS of the difference between rest and candidate intention i (i = 1 to 5), is performed as follows.

First, during the preliminary measurement, the EEG of one channel is continuously measured. When a rest is instructed (step ST101) and each candidate intention i (= 1 to 5) is presented (step ST103), 16 seconds of the analysis target of the EEG of one channel being measured is recorded, and the following is calculated To do.
Prest = (SArest, FArest, BErest)
P_i = (SA_i, FA_i, BE_i), i = 1, 2, 3, 4, 5

Here, the respective symbols are as follows.
Prest: Resting electroencephalogram pattern SArest: SA (slow α wave) power value / total frequency power value calculated by performing FFT on resting electroencephalogram data FArest: Performing FFT on resting electroencephalogram data Calculated power value of FA (fast α wave) / power value of all frequencies BErest: Calculated by performing FFT on resting brain wave data, power value of BE (β wave) / full frequency power value P_i: Candidate intention Electroencephalogram pattern when recalling i SA_i: SA (slow α wave) power value / total frequency power value calculated by performing FFT on candidate intent i brain wave data FA_i: FFT on candidate intent i brain wave data The power value of FA (fast α wave) / the power value of all frequencies calculated by performing BE_i: calculated by performing FFT on the brain wave data of candidate intention i, Power value / total frequency power value of E (β wave)

Based on these values, RMSrest_i, which is the RMS of the difference between rest and candidate intention i, is calculated as follows.
RMSrest_i =
{(SArest−SAi) ² + (FArest−FAi) ²
+ (BErest−BEi) ² } ^1/2

  In step ST106, the candidate intention i that maximizes the calculated RMS (RMSrest_i) is selected as the intention max.

In step ST107, with the rest instructed three times,
SArest_j, FArest_j, BErest_j (j = 1, 2, 3)
Can be calculated. Similarly, with the intention max specified three times,
SAmax_j, FAmax_j, BEmax_j (j = 1, 2, 3)
Can be calculated.

Next, the average of the power values of the respective frequencies calculated in step ST108 is described as follows.
SArest_ave,
FArest_ave,
BErest_ave,
SAmax_ave,
FAmax_ave,
BEmax_ave

Moreover, each SD is described as follows.
SArest_sd,
FArest_sd,
BErest_sd,
SAmax_sd,
FAmax_sd,
BEmax_sd

SDrest_ave, which is the average SD of rest used in step ST109, and SDmax_ave, which is the average SD of the intended max, are calculated as follows.
SDrest_ave =
(SArest_sd + FArest_sd + BErest_sd) / 3
SDmax_ave =
(SAmax_sd + FAmax_sd + BEmax_sd) / 3

  In step ST109, when the calculated rest SDrest_ave is smaller than the preset first threshold value TH1 and the SDmax_ave at the intention max is smaller than the preset second threshold value TH2, The process proceeds to ST110 and the setting process is continued. These first and second threshold values TH1 and TH2 are set to 0.05, for example, based on empirical rules.

  If these conditions are not satisfied, the process proceeds from step ST109 to step ST112. If the setting is attempted again, the process returns to step S101. If the setting is abandoned, the process proceeds to ST113 and the process ends.

Next, SDall_ave, which is the average of the sum of SDrest_ave and SDmax_ave used in step ST110, and RMSrest_max, which is the RMS of the average value of the rest and the average value of the intended max, are respectively calculated as follows.
SDall_ave =
(SDrest_ave + SDmax_ave) / 2
RMSrest_max =
{(SArest_ave−SAmax_ave) ²
+ (FArest_ave−FAmax_ave) ²
+ (BErest_ave−BEmax_ave) ² } ^1/2

In step ST110, the ratio of RMSrest_max to SDall_ave is calculated as follows, and it is determined whether or not this value is larger than a preset third threshold value TH3.
RMSrest_max / SDall_ave> TH3?

  If larger, the process proceeds to step ST111, and the average value of the intention max is set to the brain wave pattern of Yes. That is, the average value of the power ratio of each frequency band in each of the electroencephalograms measured in the state of intention max three times in this example is set to the Yes average pattern Ties. Moreover, the average value of rest is set to the average pattern Tno which is an electroencephalogram pattern of No. That is, the average value of the power ratio of each frequency band in each of the electroencephalograms measured multiple times, in this example, three times in a resting state, is set to No average pattern Tno. Therefore, the average patterns Tno and Ties in the second embodiment are indices corresponding to the average power ratio of the base state representing No and the average power ratio of the intention representing Yes in the first embodiment described above, respectively. .

  On the other hand, when the ratio is equal to or smaller than the third threshold value TH3, the process proceeds to step ST112. When the setting is attempted again, the process returns to step ST101. When the setting is abandoned, the process returns to ST113. The process ends after the migration.

(Intent estimation accuracy prediction method: third threshold TH3)
As for the prediction of the intention estimation accuracy, it is expected that an estimation accuracy of 80% or more is expected when “RMSrest_max / SDall_ave> TH3?” Is satisfied in step ST110 of FIG. 7 and FIG. 9 is shifted to the intention estimation mode (main measurement process / intention determination process) shown in FIG.

  FIG. 11 is a graph showing the relationship between RMSrest_max / SDall_ave obtained by the inventor's estimation from the brain wave and the intention estimation accuracy so far. The electroencephalogram measurement sites were Cz and Fz, and the intention was estimated using the power values of θ wave (4 to 8 Hz), α wave (8 to 13 Hz), and β wave (13 to 30 Hz). The intention estimation in this figure is performed by the method of the first embodiment described above.

  From this figure, it can be seen that a significant correlation (P <0.001) is recognized between RMSrest_max / SDall_ave and the intention estimation accuracy. Therefore, it can be said that the higher the RMSrest_max / SDall_ave, the higher the intention estimation accuracy. However, if the third threshold value TH3 is set too high, it may be difficult to satisfy the condition in the preliminary measurement.

In the second embodiment, when processing is performed with TH3 = 2.0, the preliminary measurement time is about 6.5 to 12.5 minutes for the first time and 0 to 9 for the second time and thereafter. It was about 5 minutes. Moreover, the intention estimation accuracy in these cases was 80% or more.

(Intention estimation mode: main measurement process / intention determination process)
Next, the intention estimation mode (main measurement process / intention determination process) in the second embodiment will be described.

  FIG. 9 is a flowchart showing a procedure in the intention estimation mode in the second embodiment. In the intention estimation mode, the data measurement time required for intention estimation is shortened to about 8 seconds at the shortest. In addition, the accuracy is improved so that the intention estimation accuracy is guaranteed to be 80% or more (substantially 100%).

  Referring to the flowchart shown in FIG. 9, first, in the routine of steps ST200 to ST204, the brain waves are continuously measured at 128 Hz, and the brain waves of 8 seconds are subjected to FFT analysis every 1/128 seconds. Calculate the RMS. In the routine of FIG. 9, t is an elapsed time when the moment when the question is presented is set to 0 seconds, and increases with the passage of real time.

  If the RMS with the Yes or No teacher signal is equal to or less than a certain value, intention determination is performed at that time (steps ST205, 206, 210, and 211). However, if the RMS does not become a certain value or less after 24 seconds (step ST208), it is displayed that the determination is impossible (step ST209). In addition, when determination impossible continues for a fixed number of times, it restarts from preliminary measurement (not shown).

  Here, RMSno_i, which is the RMS with the No teacher signal used in steps ST205 and ST206, and RMSyes_i, which is the RMS with the Yes teacher signal, are calculated as follows.

As described above, in the preliminary measurement, an average pattern Tno and an average pattern Tyes, which are No and Yes brain wave patterns, are determined and stored in the storage unit 9 (see FIG. 1A) (FIG. 7, 8 step ST111). Each of these electroencephalogram patterns is described as follows.
Tno = (SAno, Fano, BEno)
Ties = (SAyes, FAyes, BEyes)
here,
Tno: electroencephalogram pattern when No (No teacher signal)
SAno: SArest_ave
Fano: FArest_ave
BEno: BErest_ave
Ties: Electroencephalogram pattern for Yes (Yes teacher signal)
SAyes: SAmax_ave
FAyes: FAmax_ave
BEyes: BEmax_ave

In the intention estimation mode, after the start of the mode (step ST200), when 8 seconds or more have passed after the question presentation (start intention recall and brain wave measurement: step ST201) (step ST202), the brain wave data for the past 8 seconds is converted into the brain wave data. FFT is performed, and the electroencephalogram pattern at that time is calculated. This electroencephalogram pattern is described as follows.
Di = (SAi, FAi, BEi)
The meaning of each code is as follows.
Di: Electroencephalogram pattern when the intention estimation mode is executed (0 ≦ i <2048)
SAi: SA (slow α wave) power value / total frequency power value calculated by performing FFT on EEG data for the past 8 seconds FAi: FA (calculated by performing FFT on EEG data for the past 8 seconds (Fast α wave) power value / all frequency power value BEi: BE (β wave) power value / all frequency power value calculated by performing FFT on brain wave data for the past 8 seconds

RMSno_i that is the RMS with the average pattern Tno that is the No teacher signal used in Steps ST205 and 206, and RMSyes_i that is the RMS with the average pattern Tyes that is the Yes teacher signal are calculated as follows.
RMSno_i =
{(SAno−SAi) ² + (FAno−FAi) ²
+ (BEno−BEi) ² } ^1/2
RMSyes_i =
{(SAyes-SAi) ² + (FAyes-FAi) ²
+ (BEyes−BEi) ² } ^1/2

  Therefore, RMSno_i and RMSyes_i in the second embodiment are indices corresponding to No_RMS and Yes_RMS in the first embodiment described above, respectively.

(Guaranteed accuracy of intention estimation: fourth and fifth thresholds TH4 and TH5)
The intention estimation accuracy in the second embodiment can guarantee an estimation accuracy of 80% or more when the condition “RMSno_i> TH4?” In step ST205 in FIG. 9 or the condition “RMSyes_i> TH5?” In step ST206 is satisfied. The estimation accuracy is substantially 100%. This will be described below.

  FIG. 12 is a diagram showing “RMS distribution at the time of correct answer and incorrect answer” obtained by the inventors of the present invention by performing intention estimation from brain waves. The correct answer is when the intention can be correctly estimated, and the incorrect answer is when the intention cannot be estimated correctly. In addition, since there was no difference in distribution of RMSno_i and RMSyes_i, in FIG. 12, these were not distinguished.

  The electroencephalogram measurement sites were Cz and Fz, and the intention was estimated using the power values of θ wave (4 to 8 Hz), α wave (8 to 13 Hz), and β wave (13 to 30 Hz). The intention estimation in this figure was performed by the method of the first embodiment described above.

  From this figure, it can be seen that the distribution at the time of correct answer and that at the time of incorrect answer are greatly different. In particular, when RMS is 0.06 or less, the correct answer is 100%.

  Assuming that the target estimation accuracy is 80% or more, from FIG. 12, if the fourth threshold value TH4 and the fifth threshold value TH5 are set to about 0.10, the estimation accuracy becomes about 80%. When TH4 and TH5 are set to about 0.08, the estimation accuracy is about 95%, and when TH4 and TH5 are set to about 0.06, the estimation accuracy is substantially 100%.

  Note that the expression “substantially 100%” has an estimation accuracy of 100% in the data obtained so far, but there are cases where it cannot be predicted in the future, so it can be said that 100% can always be maintained reliably. It means not.

As described above, in the second embodiment, the power spectrum obtained by performing FFT (Fast Fourier Transform) for each of the measured electroencephalograms in the setting step is a plurality of frequency bands that do not overlap each other. Or, it is divided into a plurality of frequency bands that overlap at least partially, the power value of each frequency band is calculated, and the value obtained by dividing each power value of the calculated frequency band by the total frequency power value is calculated for each frequency band. Power ratio, the power ratio of each frequency band for the electroencephalogram measured in the resting state, which is the first state, and the frequency band of each frequency band for the electroencephalogram measured in each state of each candidate in the second state Calculate RMS (root mean square) (RMSrest_i) of the difference from the power ratio, and obtain the largest value of RMS in the second state. When the complement intended max, it is set the intended max interim candidate in the second state.

  Following this, each of the first state of the rest state and the state of the intention max is made to be made multiple times by the subject, and the electroencephalogram in each state is measured, and is measured in the state of multiple times of rest. The average pattern Tno of the power ratio of each frequency band in each of the electroencephalograms and the average value SDrest_ave of the SD (standard deviation) of the power ratio are calculated, and each frequency in each of the electroencephalograms measured in a plurality of states of intention max An average pattern Ties of the power ratio of the band and an average value SDmax_ave of the SD of the power ratio are calculated, and the SD (standard deviation) of the power ratio of each frequency band in each of the electroencephalograms measured in a plurality of rest states and a plurality The total average value SDa of the SD of the power ratio of each frequency band in each of the electroencephalograms measured in the state of the intention max of times And it calculates the l_ave.

  In addition, the average value of the power ratio of each frequency band in each of the electroencephalograms measured in a plurality of times of rest and the power ratio of each frequency band in each of the electroencephalograms measured in a state of multiple intention max The RMS (root mean square) (RMSrest_max) of the difference from the average value is calculated, the SD (standard deviation) of the power ratio of each frequency band in each of the electroencephalograms measured in a plurality of rest states, and the plurality of intentions The total average value SDall_ave of the power ratio SD of each frequency band in each electroencephalogram measured in the state of max is calculated.

  When SDrest_ave is smaller than a predetermined first threshold TH1, SDmax_ave is smaller than a predetermined second threshold TH2, and the ratio of RMSrest_max to SDall_ave is larger than a predetermined third threshold TH3. In addition, the intention max is set to the second state. That is, the average pattern Tno is stored and held as a teacher signal that represents the state of rest, which is the first state, and the average pattern Tyes is stored and held as a teacher signal that represents the state of the intention max that is the second state.

  In the intention estimation mode, the brain wave of the subject is measured at a predetermined sampling period, and the measured brain wave is subjected to FFT (Fast Fourier Transform) at each sampling period, and each frequency band in each of the brain waves is measured. The power ratio is calculated, and RMSno_i and RMSyes_i (root mean square), which are RMS of the difference between the power ratio and the average pattern Tno and the average pattern Ties, which are teacher signals, are calculated. When it is smaller than the predetermined fourth threshold value, it is determined that the selection intention of the subject is the intention assigned to the first state, and the average pattern Ties is smaller than the predetermined fifth threshold value. In this case, it is determined that the selection position of the target person is the intention assigned to the second state.

  According to the second embodiment, the processing time in each process can be shortened, and intention estimation can be performed with high accuracy of 80% or more by appropriately setting the third to fifth thresholds TH3 to TH5. It can be carried out.

[Other embodiments]
(Use of physiological indices other than EEG)
In the first and second embodiments, intention estimation is performed using only brain waves. It is also possible to perform intention estimation using a physiological index other than the electroencephalogram together with the electroencephalogram. FIG. 10 is a flowchart showing a processing procedure added to the processing procedures of FIGS. 7 and 8 when a plurality of types of physiological indices are measured together with an electroencephalogram and used for intention estimation. For example, NIRS and autonomous system indices (HR, GSR, etc.) can also be used for intention estimation when measured simultaneously with an electroencephalogram. However, it is used only when the physiological index changes remarkably and stably when recalling the rest “No” in the first state and the optimum intention “Yes” in the second state.

  In this case, it is necessary to normalize and use the measured physiological index so that data processing can be performed similarly to the measured brain wave (step ST301 in FIG. 10).

  The physiological index is measured simultaneously with the electroencephalogram measurement from steps ST107 at the latest in FIGS. Thereafter, for example, after the data processing on the measured electroencephalogram, the data processing of the measurement physiological index shown in steps ST301 to ST307 in FIG. 10 is performed from step ST109 in FIGS. Thereafter, the process returns to step ST110 in FIGS. 7 and 8 to perform the subsequent processing.

  A method of using the measured physiological index i as one of the frequency powers of the electroencephalogram for RMSrest_max will be specifically described below with reference to FIG.

  First, it is assumed that n physiological indices other than the electroencephalogram are simultaneously measured at the time when the processes of steps ST107 and ST108 of FIGS. That is, the physiological index i can be measured three times at rest, and is described here as Brest_i_j (j = 1, 2, 3). In addition, it is measured three times when recalling the intention max, and is described as Bmax_i_j (j = 1, 2, 3) here.

Here, the symbols and the like used in steps ST303 and ST304 in FIG. 10 are as follows.
Brest_i_sd: SD of Brest_i_j (j = 1, 2, 3)
Bmax_i_sd: SD of Bmax_i_j (j = 1, 2, 3)
Brest_i_ave: Average of Brest_i_j (j = 1, 2, 3) Bmax_i_ave: Average of Bmax_i_j (j = 1, 2, 3) SDall_ave: Average value of Brest_i_sd and Bmax_i_sd

Di is calculated as follows.
Di = | Brest_i_ave−Bmax_i_ave |

In step ST305 of FIG. 10, the calculation of RMSrest_max including the physiological index i (i = 1 to n) is as follows. The following formula is an example including at least physiological indices i = 1 and i = 2, but in practice, any number of physiological indices may be included.
RMSrest_max =
{(SArest_ave−SAmax_ave) ²
+ (FArest_ave−FAmax_ave) ²
+ (BErest_ave−BEmax_ave) ²
+ (Brest_1_ave−Bmax_1_ave) ²
+ (Brest_2_ave−Bmax_2_ave) ² +...} ^1/2

  Further, in the intention estimation mode (see FIG. 9), when intention estimation including physiological indices other than brain waves is performed, the following is performed.

First, regarding the physiological index j (j = 1, 2, 3,..., N), the teacher signals “No” and “Yes” are determined by the method shown in FIGS. In this example, the determined teacher signal is
Brest_j_ave, and
Bmax_j_ave (j = 1, 2, 3,..., N)
It becomes.

The RMS (RMSno_i) with the No teacher signal and the RMS (RMSyes_i) with the Yes teacher signal in the intention estimation mode (FIG. 9) are calculated as follows. The following formula is an example including at least physiological indices j = 1 and j = 2, but in practice, any number of physiological indices may be included.
RMSno_i =
{(SAno−SAi) ² + (FAno−FAi) ²
+ (BEno-BEi) ² + (Brest_1_ave−B1_i) ²
+ (Brest_2_ave−B2_i) ² +. } ^1/2
RMSyes_i =
{(SAyes-SAi) ² + (FAyes-FAi) ²
+ (BEyes-BEi) ² + (Bmax — 1_ave−B1 — i) <sup>2
+ (Bmax_2_ave-B2_i) 2 +</sup> ··} 1/2

  Here, B1_i is an average value of the physiological index 1 for the past 8 seconds when t = 8 + i / 128 seconds have elapsed since the question presentation, and B2_i is a physiological index for the past 8 seconds when t = 8 + i / 128 seconds has elapsed since the question presentation. The average value of 2.

The following values can be used as the sixth threshold value TH6, the seventh threshold value TH7 used in step ST303, and the eighth threshold value TH8 used in step ST304.
TH6 = 0.05
TH7 = 0.05
TH8 = 2.0

(EEG cloud storage and information sharing)
The above explanation is based on the premise that intention estimation by brain waves is performed on a subject at a fixed place. However, the present invention provides each home care recipient at a remote place such as a hospital via a communication line such as the Internet. It is also possible to monitor the brain wave and communicate by intention estimation from a remote location. In this way, it is possible to share the electroencephalogram of each home caregiver, the communication result by intention estimation, etc. among the care team members and family members of each home caregiver.

  For example, FIG. 13 shows a configuration example of a home care support system using a cloud server. In this case, the electroencephalogram data is stored as follows, and communication records can be created and viewed by care team members and families of various occupations, and team care is performed by sharing information.

  Here, the continuously measured electroencephalogram usually has different waveforms at the time of activity and at rest (sleeping). Specifically, high frequency components such as β waves increase during activity, and low frequency components such as θ waves increase during rest (sleeping). However, if the rhythm of activity and sleep (the circadian rhythm) is disrupted, the fluctuation is reduced. In such a case, a rhythm of activity and rest (sleep) is created by giving various stimuli (light, smell, sound, etc.) at a certain time every day. It is desirable to perform intention estimation (communication) when active.

During intent estimation (communication) at home, voice, brain waves, and intent estimation results are saved simultaneously. For example, it is stored in the telecommunication information storage unit 124 as a communication record. During remote intention estimation (communication), both video, audio, brain waves, and intention estimation results are displayed simultaneously, stored as communication records in the telecommunication information storage unit 124, and brain wave data stored in the tele monitoring information storage unit 125 To do. When the intention is not estimated, the brain waves are continuously stored in the telemonitoring information storage unit 125, the abnormal value (seizure) is confirmed, and the circadian rhythm is confirmed.

  The home care support system 100 shown in FIG. 13 will be briefly described below. A home care support system 100 (hereinafter simply referred to as “system 100”) performs communication between a cloud server (management server) 102 that integrally manages information on patients to be treated at home, and the cloud server 102. A plurality of communication terminals 103 (a) (a = 1, 2, 3...), And a communication line connecting the cloud server 102 and each communication terminal 103 (a), for example, an Internet line 104 are configured. The communication terminal 103 (a) is a personal computer, a mobile terminal, a mobile phone or the like, and is a team member 106 (c) (c) of the home care support team 105 (b) (b = 1, 2, 3...). = 1, 2, 3 ...).

  An administrator of the system 100 or an authority (not shown) authorized by the system administrator designates a patient and a person who performs medical care for the patient as a team member 106 (c) of one team 105 (b). ) To the system 100. Team members 106 (c) include patients 106 (1), their families 106 (2), doctors 106 (3), nurses 106 (4), physiotherapists 106 (5), home helpers 106 (6), etc. Is included. The system administrator performs maintenance such as maintenance related to the system 100, but it is optional whether or not to include it in the team member. In addition, home care support by the system 100 is performed for a large number of patients, and one team 105 (b) is formed for each patient. Therefore, in one team 105 (b), the team member 106 (c) other than the patient 106 (1) and the family 106 (2) becomes team members of a plurality of teams 105 (b), and at the same time. In many cases, home care support is provided for a large number of patients.

  The cloud server 102 functions as various processing units such as the information processing unit 111, the screen display unit 112, the transmission / reception unit 113, and the authentication unit 114 by executing the home care support computer program. In addition, a plurality of databases 120 in which various types of information for home medical treatment are registered are provided. The database 120 includes a team information storage unit 121, a basic information storage unit 122, a care record storage unit 123, and telecommunications information. A storage unit 124, a telemonitoring information storage unit 125, and the like are included.

  The cloud server 102 implements an information sharing function, a telecommunications function, and a telemonitoring function between job types having the following contents using various information registered in the database 120.

(1) Information sharing function between job types The information sharing function between job types includes the following configurations and functions of 1-1) to 1-3).
1-1) In the team information storage unit 121, a person who performs medical care of each patient 106 (1) by a system administrator (or a person who has been given authority by the system administrator) is a team member. (1) The family 106 (2), the doctor 106 (3), the nurse 106 (4), the physical therapist 106 (5), the home helper 106 (6), etc. are registered including the person. It is optional whether or not to include a system administrator as a member of each team.
1-2) The patient record (facial expression, blood pressure) acquired by the visit member, hospitalization, telephone call, etc. by the team member who has logged in by accessing the cloud server 102 via the communication terminal 103 (a) in the care record storage unit 123 , Physical condition, etc.) and medical care (instructions, massage, medication, etc.) for the patient 106 (1) performed in response thereto is referred to as a care record (even if only a description of the patient situation is a care record) Can be registered. The team member includes the patient 106 (1). For example, when the patient 106 (1) himself / herself measures blood pressure, or when the blood pressure is measured and a hypotensive drug is taken by self-judgment, etc. The patient 106 (1) can also register a care record.
1-3) Each team member registered in the care record storage unit 123 by each team member can view the team member who has logged on by accessing the cloud server 102 via the communication terminal 103 (a). Here, when a team member registers a care record, all or a part of the care record can be set so that only a specific team member can view it.

(2) Telecommunication function The telecommunication function includes the following functions 2-1) to 2-4).
2-1) When a team member registers or views a care record, communication can be registered on a screen corresponding to the care record. This screen is a screen for team members to exchange opinions related to the care record, and can be viewed and input by the team members.
2-2) When a team member views a biological signal by the telemonitoring function described below, communication can be registered on a screen corresponding to the browsing. This screen is a screen for team members to exchange opinions related to the biological signal, and can be browsed and input by the team members.
2-3) Team members can register communications on a new screen regardless of care records or telemonitoring. This screen is a screen for team members to exchange opinions on an arbitrary topic and can be viewed and input by the team members.
2-4) Team members can perform telecommunications in real time through a WEB conference (or screen sharing).

(3) Telemonitoring function The telemonitoring function includes the following functions 3-1) and 3-2).
3-1) Team members view in real time the waveforms and numerical values output from devices that measure biological signals such as patient breathing, SpO2 (arterial oxygen saturation), electrocardiogram, electroencephalogram, body temperature, blood pressure, etc. it can. Alternatively, waveforms and values related to the state of devices such as oxygen supply devices and ventilators can be viewed in real time.
3-2) The above waveforms and numerical values are registered as raw data or summary information and can be viewed by team members at any time.

1 Intent Estimator 2 EEG 3 PC
4 Input device 5 Display device 6 EEG acquisition unit 7 EEG analysis unit 8 Display control unit 9 Storage unit

Claims (36)

A first measurement step of acquiring the subject's brain waves measured when the subject is made to make the first state and the second state;
The first electroencephalogram measured in the first state and the first state is set as the state and the electroencephalogram when the subject selects one of the selection intentions, and the second state and the first electroencephalogram are set. A setting step for setting the second electroencephalogram measured in the state of 2 to be a state and an electroencephalogram when the subject selects the other of the selection intentions;
A second measurement step of acquiring the subject's brain wave measured when the subject who gave the question is asked to make either the first state or the second state;
An intention determination step of comparing the brain wave measured in the second measurement step with the first brain wave and the second brain wave to determine a selection intention of the subject with respect to the question;
An intention estimation method using an electroencephalogram, comprising:   The intention estimation method using an electroencephalogram according to claim 1, wherein in the first measurement step, a spontaneous electroencephalogram measured at at least one measurement site on the subject's head is acquired.   One of the first and second states is a predetermined mind-body state that the subject can create, and the other state is a recall state that recalls a predetermined matter that the subject can recall. The intention estimation method by an electroencephalogram according to claim 1. In the first measurement step,
As an electroencephalogram measured in at least one of the first and second states, an electroencephalogram is measured when a plurality of candidate states that are candidates for the one state are created and given to the subject. Get each
In the setting step,
A candidate in which one electroencephalogram extracted from the plurality of electroencephalograms measured is set as the one state.
The intention estimation method by an electroencephalogram according to claim 1. In the first measurement step,
For each of the plurality of candidate states, obtain an electroencephalogram that is measured when the candidate is asked to make each candidate state multiple times,
In the setting step,
The candidate state having the highest degree of coincidence between the plurality of electroencephalograms measured for each candidate is set as the one state.
The intention estimation method by an electroencephalogram according to claim 4. In the setting step,
After determining one of the first and second states,
Setting the other state of the first and second states to a state in which an electroencephalogram having the largest degree of separation from the electroencephalogram of the one state that has been determined is obtained;
The intention estimation method by an electroencephalogram according to claim 5. In the first measurement step,
In a plurality of measurement parts of the subject's head, obtain an electroencephalogram that is measured when each of the plurality of candidate states is made by the subject.
The intention estimation method by an electroencephalogram according to claim 4. When the accuracy of the determination result of the selection intention in the intention determination step is equal to or lower than a predetermined value, or a predetermined time from the first and second states and the first and second brain waves set by the setting step Including a setting invalidating step of invalidating at least the setting of the first and second electroencephalograms among the first and second states and the first and second electroencephalograms when
The intention estimation method by an electroencephalogram according to claim 1. In the setting step,
If the measured amplitude of the electroencephalogram includes a portion that exceeds a predetermined setting range, or the setting is predetermined for a part of the power spectrum obtained by performing FFT (Fast Fourier Transform) on the electroencephalogram When there is a value exceeding the value, the brain wave is invalidated.
The intention estimation method by an electroencephalogram according to claim 1. Set a predetermined frequency band in the measured electroencephalogram to all frequency bands to be analyzed,
Performing the setting step and the intention determination step for the electroencephalogram components included in the entire frequency band,
The intention estimation method by an electroencephalogram according to any one of claims 1 to 9. In the first measurement step,
If there is one candidate for at least one of the first and second states, obtain a plurality of electroencephalograms measured when the subject makes the candidate state multiple times,
In the case where there are a plurality of candidates for the one state, for each of the candidates, a plurality of electroencephalograms that are measured when the subject is made a plurality of states of each candidate are acquired for each candidate,
In the setting step,
Set a predetermined frequency band in the measured electroencephalogram as the entire frequency band to be analyzed,
For the measured electroencephalogram, a power spectrum obtained by performing FFT (Fast Fourier Transform) on the entire frequency band is converted into a plurality of frequency bands that do not overlap each other or a plurality of frequency bands that overlap at least partially. Partition and
Calculate the power value of each of the frequency bands,
A value obtained by dividing the calculated power value of each frequency band by the total frequency power value is a power ratio of each frequency band,
When the one state is set, if there is one candidate for the one state, an average value of SD (standard deviation) of the power ratio of each frequency band in each of the measured electroencephalograms is determined in advance. If less than a set value, set the candidate to the one state,
When there are a plurality of candidates for the one state, an average value of the power ratio SD of each frequency band of each brain wave in each measured candidate is calculated, and the brain wave having the smallest average value is measured. Set the candidate to one of the states,
The intention estimation method by an electroencephalogram according to claim 4. In the setting step,
When setting at least one of the first and second states, and there is one candidate for the one state,
For each power ratio calculated from each of the measured electroencephalograms, an average value and variation are calculated for each power ratio, a predetermined variation range centered on the average value is set, and even one of the frequency bands If the brain wave is out of the range, the brain wave is excluded from the calculation of the power ratio, the average value and the dispersion are calculated again for each power ratio from the brain waves other than the excluded brain wave, the range is reset, and the frequency is reset. If even one of the bands is out of the range, repeat the process of excluding the electroencephalogram from the calculation of the average power ratio,
If the number of remaining electroencephalograms falls below a predetermined set number, give up setting the candidate to the one state,
If the number of remaining electroencephalograms is greater than or equal to the set number, if the average value of the power ratio variation of each frequency band is less than a preset value, the candidate is set to the one state.
The intention estimation method by an electroencephalogram according to claim 11. In the setting step,
Set a predetermined frequency band in the measured electroencephalogram as the entire frequency band to be analyzed,
For the measured electroencephalogram, a power spectrum obtained by performing FFT (Fast Fourier Transform) on the entire frequency band is converted into a plurality of frequency bands that do not overlap each other or a plurality of frequency bands that overlap at least partially. Partition and
Calculate the power value of each of the frequency bands,
A value obtained by dividing the calculated power value of each frequency band by the total frequency power value is a power ratio of each frequency band,
When one of the first and second states is determined, when determining the other state from a plurality of candidates,
The average value of the power ratio of each frequency band for each electroencephalogram measured by causing the subject to make the determined one state a plurality of times, and one candidate of the other state candidates to be determined from the mean value Calculate the RMS (root mean square) of the difference from the average value of the power ratio of each frequency band for each brain wave measured by making the subject multiple times,
From the RMS, a subtracted value obtained by subtracting an average SD (standard deviation) of the power ratio of each frequency band when the candidate is made three or more times in the other state, or from the RMS, Calculating a subtracted value obtained by subtracting the RMS of the average value of the power ratio when the candidate is made twice in one candidate of the other state;
In the case where all the candidates in the other state are made to be created by the subject, the measurement part and the other of the measurement parts that can obtain the electroencephalogram with the largest subtraction value among the electroencephalograms obtained from the plurality of measurement parts. The intention estimation method using an electroencephalogram according to claim 7, wherein the state candidates are set to the electroencephalogram measurement region and the other state in the second measurement step, respectively. In the intention determination step,
If the measured amplitude of the electroencephalogram includes a portion that exceeds a predetermined setting range, or the setting is predetermined for a part of the power spectrum obtained by performing FFT (Fast Fourier Transform) on the electroencephalogram When there is a value exceeding the value, the brain wave is invalidated.
The intention estimation method by an electroencephalogram according to claim 1. In the setting step,
Set a predetermined frequency band in the measured electroencephalogram as the entire frequency band to be analyzed,
For the measured electroencephalogram, a power spectrum obtained by performing FFT (Fast Fourier Transform) on the entire frequency band is converted into a plurality of frequency bands that do not overlap each other or a plurality of frequency bands that overlap at least partially. Partition and
Calculate the power value of each of the frequency bands,
A value obtained by dividing the calculated power value of each frequency band by the total frequency power value is a power ratio of each frequency band,
Calculating a first average power ratio and a second average power ratio that are average power ratios of respective frequency bands of the first brain wave and the second brain wave;
In the intention determination step,
Calculating the power ratio of each frequency band of the electroencephalogram measured in the second measurement step as a third power ratio;
Calculating the RMS (root mean square) of the difference between the first and third power ratios;
Calculating the RMS of the difference between the second and third power ratios;
The selection intention assigned to the electroencephalogram on the side where the smaller one of the two RMS values is obtained is determined as the intention of the subject.
The intention estimation method by an electroencephalogram according to claim 14. In the intention determination step, if the difference between the RMS of the difference between the first and third power ratios and the RMS of the difference between the second and third power ratios is less than a predetermined set value, the target Put the intention determination of the person on hold,
The intention estimation method by an electroencephalogram according to claim 15. The first state is a state of rest of the subject,
The second state is a recall state that recalls a predetermined matter that the subject can recall.
The intention estimation method by an electroencephalogram according to claim 1. In the plurality of measurement sites on the subject's head, the first brain wave and the second brain wave are measured in the rest state and the recall state, respectively, and the measurement site having the largest degree of separation is determined as the optimum measurement site. Select as
In the first and second measurement steps, an electroencephalogram is measured at one of the optimal measurement sites.
The intent estimation method by an electroencephalogram according to claim 17. In the first measurement step,
As the electroencephalograms measured in the second state, each of the electroencephalograms measured when the subject is made with a plurality of candidate states that are candidates for the second state are obtained,
In the setting step,
The second state is set to a state in which an electroencephalogram having the largest degree of separation from the electroencephalogram measured in the resting state is obtained among a plurality of candidates.
The intent estimation method by an electroencephalogram according to claim 17. In the first measurement step,
Of the plurality of candidate states for the target person, select the candidate set to the second state most frequently in the past as a provisional candidate,
Obtain an electroencephalogram that is measured when the subject is made up with the status of the provisional candidate,
In the setting step,
Comparing the measured electroencephalogram with an electroencephalogram measured in a resting state, which is the first state, and setting the provisional candidate as the second state when satisfying a predetermined fitness criterion;
The intention estimation method by an electroencephalogram according to claim 19. In the setting step and the intention determination step,
For each of the measured electroencephalograms,
The frequency band uses a band that does not include θ waves (4 to 8 Hz).
The intention estimation method by an electroencephalogram according to claim 1. Set a predetermined frequency band in the measured electroencephalogram to all frequency bands to be analyzed,
The intention estimation method using an electroencephalogram according to claim 17, wherein the setting step and the intention determination step are performed on an electroencephalogram component included in the entire frequency band. For each of the measured electroencephalograms, a power spectrum obtained by performing FFT (Fast Fourier Transform) for all the frequency bands, a plurality of frequency bands that do not overlap each other or three frequencies at least partially overlapping Partition into bands,
Based on the respective power values of the three frequency bands and the total frequency power values of the all frequency bands, the setting step and the intention determination step are performed.
23. The intention estimation method using an electroencephalogram according to claim 22. In the setting step,
Set a predetermined frequency band in the measured electroencephalogram to all frequency bands to be analyzed,
For each of the measured electroencephalograms, a power spectrum obtained by performing FFT (Fast Fourier Transform) on the entire frequency band, a plurality of frequency bands that do not overlap each other, or a plurality of frequencies that at least partially overlap each other. Partition into bands,
Calculate the power value of each of the frequency bands,
A value obtained by dividing the calculated power value of each frequency band by the total frequency power value of the entire frequency band is a power ratio of each frequency band,
The power ratio of each frequency band for the electroencephalogram measured in a resting state that is the first state, and the power ratio of each frequency band for the electroencephalogram measured in each state of each candidate of the second state RMSrest_i, which is the RMS (root mean square) of the difference between
20. The intention estimation method using an electroencephalogram according to claim 19, wherein the intention max is set to the second state when the candidate of the second state in which the RMS having the largest value is obtained is the intention max. In the setting step,
Each of the state of the first state and the state of the intention max is caused to be made multiple times by the subject, and an electroencephalogram in each state is measured,
An average pattern Tno of the power ratio of each frequency band in each of the electroencephalograms measured in a plurality of rest states and an average value SDrest_ave of the SD (standard deviation) of the power ratio,
An average pattern Ties of the power ratio of each frequency band in each of the electroencephalograms measured in the state of the intention max a plurality of times and an average value SDmax_ave of SD of the power ratio;
SD (standard deviation) of the power ratio of each frequency band in each of the electroencephalograms measured in a plurality of resting states and the power ratio of each frequency band in each of the electroencephalograms measured in a plurality of states of the intention max. Calculate the total average value SDall_ave of SD,
The average value of the power ratio of each frequency band in each of the electroencephalograms measured in a plurality of rest states and the average value of the power ratio in each frequency band for each of the electroencephalograms measured in a plurality of states of intention max. RMSrest_max, which is the RMS (root mean square) difference of
When SDrest_ave is smaller than a predetermined first threshold value TH1, SDmax_ave is smaller than a predetermined second threshold value TH2, and the ratio of RMSrest_max to SDall_ave is larger than a predetermined third threshold value TH3, Setting the intention max to the second state;
The average pattern Tno is stored and held as a teacher signal representing a rest state that is the first state, and the average pattern Ties is stored and held as a teacher signal that represents the state of the intention max that is the second state. The intention estimation method by an electroencephalogram according to claim 24. As the first and second threshold values TH1 and TH2, a value of 0.05 or smaller is adopted,
26. The intention estimation method using an electroencephalogram according to claim 25, wherein the third threshold TH3 is 2.0 or larger. In the second measurement step, the brain wave of the subject is measured at a predetermined sampling period,
In the intention determination step,
The FFT (Fast Fourier Transform) is performed on the entire frequency band in the measured brain wave for each sampling period or separately determined period, and the power ratio of each frequency band in each of the brain wave is calculated. ,
RMSno_i and RMSyes_i (root mean square) that are RMS of the difference between the power ratio and the average pattern Tno and the average pattern Ties that are teacher signals,
If the RMSno_i is smaller than a predetermined fourth threshold TH4, it is determined that the selection intention of the subject is an intention assigned to the first state;
If the RMSyes_i is smaller than a predetermined fifth threshold TH5, it is determined that the selection position of the subject is an intention assigned to the second state;
The intention estimation method by an electroencephalogram according to claim 25.   28. The intention estimation method according to claim 27, wherein the fourth and fifth threshold values TH4 and TH5 are 0.10 or smaller. Even if the second measurement step and the intention determination step are repeated over a predetermined time, the RMSno_i does not become smaller than a predetermined fourth threshold value, and the RMSyes_i is a predetermined fifth value. 28. The intention estimation method using an electroencephalogram according to claim 27, wherein the intention determination is terminated because the intention determination is impossible if the threshold value is not smaller than the threshold value. Continuously measure the subject's brain waves,
Determine the period of activity of the subject based on continuously measured brain waves,
The intention estimation method using an electroencephalogram according to claim 1, wherein the first measurement step, the setting step, the second measurement step, and the intention determination step are performed in the period of the active state. Applying an external stimulus to the subject every day at a fixed time to arrange the circadian rhythm of the subject, and after determining that the subject is in an active state based on the continuously measured brain waves, The intention estimation method using an electroencephalogram according to claim 30, wherein the intention estimation is performed. In the first measurement step, in each of the first and second states, together with the subject's brain waves, one or more types of physiological indices other than brain waves are measured as auxiliary indices for intention determination,
In the setting step, the state in which the subject selects one of the selection intentions of the first physiological information including the first brain wave and the first physiological index measured in the first state and the first state. And the second physiological information including the second brain wave and the second physiological index measured in the second state and the second physiological index, the target person selects the other of the selection intentions. Set it to be the state and physiological state to select,
In the second measurement step, the subject's brain wave and the physiological index that are measured when the first state and the second state are made and struck are obtained,
In the intention determination step, the physiological information including the electroencephalogram and the physiological index measured in the second measurement step are compared with the first physiological information and the second physiological information, and the question of the subject is answered. The intention estimation method using an electroencephalogram according to claim 1, wherein the intention of selection is determined.   A computer-readable program for estimating an intention which causes a computer to execute the intention estimating method according to any one of claims 1 to 32. A first measurement unit, a setting unit, and a second measurement unit that execute the first measurement step, the setting step, the second measurement step, and the intention determination step by executing the intention estimation program according to claim 33. And a computer functioning as an intention determination unit;
An electroencephalograph connected to the computer and measuring an electroencephalogram of a subject whose intention is estimated;
A display device connected to the computer and displaying operation states of the first measurement unit, the setting unit, the second measurement unit, and the intention determination unit;
An input device connected to the computer for giving an input instruction to the computer;
The intention estimation apparatus characterized by having.   33. A computer-readable setting program for an intention estimation brain wave for causing a computer to execute the first measurement step and the setting step in the intention estimation method according to any one of claims 1 to 32. A computer functioning as a first measurement unit and a setting unit for executing the first measurement step and the setting step by executing the program for setting an electroencephalogram for intention estimation according to claim 35;
An electroencephalograph connected to the computer and measuring an electroencephalogram of a subject whose intention is estimated;
A display device connected to the computer and displaying operation states of the first measurement unit and the setting unit;
An input device connected to the computer for giving an input instruction to the computer;
A device for setting an electroencephalogram for intention estimation, comprising: