JP2015132671A - unconscious learning method using neurofeedback - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an unconscious learning method using neurofeedback, presenting a task as a game using brain waves (MMN-mismatch negativity) to a subject without making the subject consciously pay attention to an object to learn in the learning method of using an electroencephalograph.SOLUTION: An unconscious learning method using neurofeedback includes the steps of: (1) repeatedly presenting a task to a subject by an oddball method using task presentation means without any explanation of the task, for a learning task selected, with it regarded to have an MMN response, and, while presenting an amplitude of an MMN signal to the subject, prompting the subject to enhance the intensity of the MMN signal; and (2) presenting a discrimination object in the learning task to the subject individually to check a learning result. This method, by repetition of the steps (1) and (2), causes the enhancement of the MMN signal of the subject and consequently improves the task discrimination capability. Further, the amplitude of the MMN signal is input in a game to assist the MMN signal enhancement.

Description

  The present invention relates to an unconscious learning method using neurofeedback that improves the information processing ability of a subject by using neurofeedback that returns a mismatch negativity (MMN) signal to the subject.

In general, it is said that Japanese people have difficulty distinguishing between English L and R sounds. However, although the difference between L and R is not known at the consciousness level, it is known that when an electroencephalogram is measured, the difference appears in an electroencephalogram signal called an MMN signal (hereinafter referred to as MMN). The amplitude of MMN becomes larger when the difference between the two sounds can be recognized well.
Therefore, in the research by the inventors of the present application, when the MMN amplitude is made to correspond to the radius of the circle and the effort is made to increase this radius, it is clear that the discrimination rate between the two sounds that were heard is improved. Became. The important point in this study is that the subject can increase the size of the circle and improve the recognition rate even though he does not know what the radius of the circle corresponds to. .

  In the first place, it is known that MMN occurs even if the subject does not pay attention to the target sound. Therefore, if the amplitude of the MMN is made to correspond to, for example, the maximum speed of the car in the racing game, the subject simply plays the racing game and automatically tries to increase the speed. The difference between two sounds, for example, the difference between English L and R can be learned unconsciously.

  A method of efficiently performing various learnings by feedback (FB) of the brain activity state to the subject is already well known and is called a neurofeedback (NFB) learning method. In general, in the NFB learning method, the subject is not FB whether the answer to the issued question is correct or incorrect, but the result of analyzing the brain activity pattern is NFB (neurofeedback) to the subject. For example, in the disclosure of Patent Document 1, in learning to increase the ability to discriminate sounds with two different pitches, in a general learning method, these two different sounds are heard and whether they are the same or different. The subject is asked to answer and the subject is FB whether or not the answer is correct. In the NFB learning method, the brain activity pattern is read to determine whether the subject perceived the same sound or different sound, and the subject is FBed. Do that. The greatest advantage of this NFB learning is that the subject's FB can be given in a continuous amount rather than a discrete amount. In other words, in general learning, the FB to the subject is either a correct answer or an incorrect answer, whereas in NFB learning, it is possible to analyze how close the brain activity pattern is to the correct answer. Can be provided as a continuous amount. As a result, even if the test subject's answer is wrong, it is considered that the learning is progressing as long as the brain activity pattern is close to the correct answer, so that the learning efficiency is expected. However, until he gets tired, the subject needs to know the degree of learning each time with respect to the learning result, so the subject needs to pay attention to the learning target.

For example, Patent Document 1 is disclosed as being similar to the present invention. This similarity is as follows.
In general, when learning the difference in pitch, according to the style of oddball stimulation, a lot of category 1 sounds (for example, slightly lower sounds) are put in the sound sequence, and sometimes category 2 sounds (for example, slightly higher sounds) are put. To the subject to hear the acoustic stimulus (FIG. 2). At this time, if the difference between category 1 and category 2 sounds can be perceived, the brain wave called MMN (mismatch negativity) when listening to category 2 sound (that is, when you hear a relatively small amount of sound) The waveform is observed. It is known that the amplitude of the MMN increases as the difference between categories can be perceived more clearly. Therefore, by subjecting the subject to NFB the amplitude value of this MMN, the subject is informed how firmly the sound difference between the two categories can be perceived. The subject can learn how to perceive this sound based on this NFB. However, in this disclosure, the NFB has only changed from a correct answer or an incorrect answer to a continuous amount, and learning is naturally performed while doing other things such as a game without paying attention to the subject to be learned. The state of progress has not been reached.

In addition, Non-Patent Document 2 reports a study that brings about the knowledge that the brain can be learned even if the learning target is not clear. In this study, the degree of similarity of the brain activity pattern of a subject observed by functional magnetic resonance imaging (fMRI) to a target brain activity pattern measured in advance by the subject was measured in terms of the size of the disk. Converted and subjects were visually NFB online. The subject was able to enlarge the disc by repeating the training in the above learning without knowing what the brain activity pattern corresponding to the size of the disc is. In addition, after this became possible, we verified the perception corresponding to this target brain activity pattern (perception for lines that are difficult to see in this paper), and found that the discrimination ability in perception improved. It has been reported.
The fMRI is a large apparatus and is difficult to use in a normal educational field. In addition, since this method cannot be used unless the target brain activity pattern is known, it can be used to strengthen its processing capacity from a state where it can already be discriminated to some extent. There is also a problem that cannot be done.

International Publication No. 2010/117264

K. Shibata, T. Watanabe, Y. Sasaki and M. Kawato, “Perceptual learning incepted by decoded fMRI neurofeedback without stimulus presentation”, Science, 2011, 334 (6061): 1413-1415

  It is a learning method using an electroencephalograph that is relatively inexpensive and easy to measure, and it is not necessary for the subject to turn his attention to the subject to be learned, and the subject can be presented to the subject as a game using brain waves, We propose an unconscious learning method using neurofeedback.

The unconscious learning method using the neurofeedback of the present invention includes a task presenting means for presenting a subject with a task for discrimination between two discrimination targets to a subject, an electroencephalogram signal detecting means for detecting the brain wave signal of the subject, A learning device using signal processing means for generating a feedback signal including intensity information of the detected electroencephalogram signal and returning to the subject, and feedback signal presenting means for presenting the generated feedback signal to the subject By using the information indicating the brain wave state of the subject to be returned to the subject, the information can be classified into learning methods that improve the information processing ability of the subject.
In the case of the present invention, the problem presentation means is presented by the oddball method, the electroencephalogram signal is an MMN (mismatch negativeity) signal, the electroencephalogram signal detection means is an MMN signal detection means,
(1) Using the learning device, the first subject who has received an explanation about the above problem learns the problem signal before selection, and when the MMN signal exceeds a predetermined threshold for the problem by the learning, By leaving the assignment, selecting the learning assignment,
(2) About the learning task including the selected learning task or the stimulus that is the basis of the MMN signal, the second subject different from the first subject uses the task presenting means without explaining the learning task. Prompting the second subject to increase the strength of the MMN signal detected using the MMN signal detecting means, and repeatedly presenting the oddball method;
(3) individually presenting the discrimination target in the learning task including a learning task or a stimulus that is a basis of the MMN signal to the second subject and confirming the learning result;
Is included.
When a plurality of subjects tackle the same learning task, for example, step (1) above can be used in common or can be omitted if known, and step (2) above can be performed for each subject. Step (3) above can be performed after experiencing step (2) multiple times. It is possible to return to the step (2) after the step (3).
The present invention makes it possible to discriminate a discrimination task that was originally impossible for the subject to discriminate despite the unconscious state that the subject himself does not know that he / she is performing the discrimination task.

  In the step (2), when the learning task or the learning task including the stimulus that is the basis of the MMN signal is repeatedly presented by the oddball method, the discrimination target may be replaced and presented. . In addition to the change in the order of presentation in the above presentation, the change in the presentation ratio in the oddball method can be said.

  The feedback signal presented to the feedback signal presenting means may be a feedback signal using an accumulated value of the MMN signal strength. Here, the feedback signal is preferably an online feedback, but in addition to the same processing as the processing of the MMN signal with a linear amplifier or the like, a time-series cumulative value of the unsigned MMN signal is used. But you can. Further, this accumulation may be performed by a sum provided with temporal weighting that places importance on the latest time.

  Presenting the video game to the second subject and using the feedback signal as an input signal for controlling the progress of the video game presented to the second subject, the second subject is made to increase the intensity of the MMN signal. Increasing the MMN signal, such as prompting, may be an incentive to the subject. For example, the intensity of the beam gun fired in the game is increased by the intensity of the MMN signal, or the character is advanced.

  The task signal is an acoustic signal. Background sound may be given. In addition to the sound of a single frequency component, there may be a sound consisting of a complex spectrum.

  The task signal is a video signal. In addition to the video signal related to the shape, there may be a video signal related to the color.

In this way, if the subject does not focus on sound and does not understand what the size of the circle means, learning proceeds when neurofeedback learning using brain waves is performed. Whether or not it was unknown, but after actually experimenting, it became possible to distinguish the difference after training compared to before training.
As a result, it is possible to obtain an effect that the learning proceeds while the sound is simply being heard and the neurofeedback does not know what the meaning is.

It is a figure which shows the example of an apparatus for applying the unconscious learning method using the neurofeedback of this invention. (A) shows the procedure of a learning stage, (b) shows four combinations of the acoustic stimulation in a BAD test, (c) is a figure which shows the procedure of the trial of a BAD test stage. (A) accuracy of each BAD test for individual subjects under NFB conditions, (b) accuracy of each BAD test for individual subjects under subject experimental conditions, (c) under the NFB conditions and subject experimental conditions. It is a figure which shows the average of the improvement of discrimination correctness, and (d) the correctness group average under NFB conditions and object experiment conditions. (A) Group average MMN amplitude under NFB condition, (b) Group average peak latency under NFB condition.

  Embodiments of the present invention will be described below in detail with reference to the drawings. In the following description, devices having the same function or similar functions are denoted by the same reference numerals unless there is a special reason.

  FIG. 1 shows an example of an apparatus for applying the unconscious learning method using the neurofeedback of the present invention. The stimulus presentation device 2 is for presenting the stimulus of the subject to the subject 1. In the case of FIG. 1, an example of the auditory stimulus is shown through the earphone 3. In the case of oddball stimulation, it is a device that presents two stimuli to be distinguished. The stimulus may be any sense information such as vision, in addition to hearing. The electroencephalograph 4 is for acquiring an electroencephalogram through an electrode 5 attached to the subject 1. Here, an example using an electroencephalograph is shown, but not only so-called electroencephalograms, but anything that can measure MMN such as magnetoencephalogram and cortical electroencephalogram may be used. The signal processing device 6 is a device that receives the output from the electroencephalograph 4 and evaluates the MMN. It is desirable that latency information can be output in addition to the MMN amplitude. The feedback device 7 receives the output from the signal processing device 6 and FBs the size of the MMN to the subject. The FB method is FB as visual information in FIG. 1, but any sense information such as hearing and taste may be used. Here, the latency information may be used as the FB information.

Also in the case of the present invention, for example, oddball stimulation is used as the stimulation problem. At this time, if the difference between the two sounds can be identified, MMN appears. Since this MMN has a characteristic that it occurs even if it is not conscious if the difference can be identified without paying attention to the acoustic stimulus, the present invention uses the characteristic of this MMN.
In the disclosure of Patent Document 1, the subject receives neurofeedback after paying attention to the acoustic stimulus and trying to distinguish the difference properly.
However, according to the present invention, the subject is instructed that it is not necessary to pay attention to the sound, and the effort is made to increase the size of the circle corresponding to the size of the MMN. Incidentally, the subject does not know what the size of the circle corresponds to. Unlike Non-Patent Document 2, the use of MMN eliminates the need to examine the subject's brain activity pattern in advance.

  The most effective use case is considered to be language learning. For example, the subject is listening to a word that has a difference between R and L, such as right and light (in this case, Category 1 is right and category 2 is light). By having a racing game with a speed corresponding to this MMN, the subject can naturally distinguish between R and L simply by trying to drive the car faster. Examples are possible.

  The MMN can be triggered by any identifiable change in a sequence of auditory stimuli represented by an auditory oddball task where the deviant stimulus is approximately 20% relative to the remaining standard stimulus. The amplitude and latency well indicate how much the change in sound is discriminated. The easier it is to discriminate, the greater the amplitude and the shorter the latency. Interestingly, MMN is observed even when the object is not consciously identified. It is also known that MMN can be triggered without the subject's attention to acoustic stimulation. In this example, the MMN intensity related to identification is fed back to the subject with a visual display, and the subject is instructed to increase the size of the visual mark on the corner. The subject does not know the correspondence between the MMN response and the size of the visual display. It just tries to increase the size of the visual display. In other words, the subject never realizes that he is performing perceptual skill training and does not need to pay attention to the sound. After the above training, perceptual skills improved.

This example will be described in more detail below.
Under the above NFB, subjects were trained to discriminate between 1000 Hz and 1008 Hz unconsciously. In this training, 80% were stimulated at 1000 Hz and the rest were at 1008 Hz to enhance the evoked MMN response. Acoustic stimulation was given randomly within 0.5 seconds. FIG. 2A shows a time series of experiments. The MMN signal strength is calculated for the latest 20 sound data. The contents are (standard stimulus: 16 trials, eccentric stimulus: 4 trials). The calculated data is visually fed back with a green disk. The diameter of the disc corresponds to the MMN signal strength. The instruction to the subject is “ignore the sound from the headphones and make the green disk on the display as large as possible”.

  One frame consists of 300 sounds and takes 150 seconds. The daily training was 12 frames of training per day. The subject performed the above daily routine for 5 days, but provided at least 24 hours between each daily routine. It is not 5 consecutive days, but it is not longer than 10 days. The diameter of the green disk is fixed for the first 20 sounds of one frame. This is because 20 sounds are required to calculate the MMN signal. The above diameter was updated every 0.5 seconds when the MMN was updated. The size of the disk is a value determined by linear mapping of the amplitude of the MMN signal. It is necessary to note that the subject is not informed about what value is displayed. It is concealed that the experimental purpose is perceptual training.

The training flow is as follows. However, it has been found that the results of this subject are higher than before the training in the inventor's research.
(1) First, in the case of the existing method, it is as follows.
1-1: An electroencephalograph is attached to the subject.
1-2: The subject listens to words such as “play” and “pray” and learns so that either can be distinguished.
1-3: Give feedback to the subject whether the brain has started to judge correctly, even if it is wrong, by feeding back the MMN when doing this.
1-3-1: Stimulus may be bought not only according to this feedback but also according to the size of the MMN (for example, changing the speed of the word or changing the size of the sound so that it is easy to hear when the MMN is small) etc).
1-3-2: The subject knows that the fed back corresponds to the MMN.
1-4: Since it is known that the subject is learning the language, the training and the evaluation are performed at the same time (the evaluation is immediately fed back after the training)
(2) On the other hand, in the present invention, it is as follows.
2-1: Put an electroencephalograph on the subject.
2-2: The subject enlarges the circle reflecting the size of the presented MMN while listening to words such as play and pray.
2-2-1: There is no need to learn about the word you are listening to.
2-2-2: There is no need to tell the subject that the size of the circle corresponds to the MMN.
2-2-3: What is presented need not be a circle, and may be associated with, for example, a game parameter.
2-3: By training several times, the circle reflecting MMN can be intentionally enlarged.
2-4: After training, listen and evaluate the difference between L and R.

  A BAD (behavioral auditory discrimination) test was conducted prior to the first day of training in order to evaluate the improved sound discrimination ability. That is a pre-training test. In addition, it was conducted on each training day after training. In that test, the subject's ability to discriminate between the sounds was measured, but the subject answered the difference between the two single sounds. FIG. 3 (a) shows the results of the BAD test of 8 subjects. A one-way analysis of variance with repeated measurements shows that training is effective. The result is F (5,42) = 8.40, p <0.0001.

  A posteriori t-test between the accuracy at any close stage revealed the following. That is, the discriminating ability of two pure tones has been improved except for the third and fourth days. Overall, accuracy and the average accuracy on each training day increased asymptotically, and this trend gradually improved. Eventually, through NFB training, the accuracy of the last day was improved by 25.45 ± 3.09% (mean of subject ± standard error) compared to pre-test.

  In day-to-day training, subjects are instructed to ignore acoustic stimuli, but training may have improved their acoustic discrimination ability by getting used to listening to acoustic stimuli repeatedly. To confirm this possibility, a control experiment was performed on 8 new subjects. In the control experiment, subjects were given the same instructions and electrodes were attached to the head. However, the measured electroencephalogram was ignored and the size of the green disk of visual feedback was presented in the same procedure as recorded under other subjects' NFB conditions. That is, the subjects in the control experiment showed the same visual stimulus as in the NFB condition. It should be noted that the subject does not know whether he participated under NFB conditions or under control experimental conditions.

  BAD testing of individual subjects under control experimental conditions was performed. The results are shown in FIG. FIG. 3 (c) shows that no significant performance improvement was seen in the BAD test under control experimental conditions. Group average comparisons were also made under NFB and control experimental conditions. The results show that NFB conditions are significantly improved compared to control experimental conditions. It can be seen that the ability improvement under NFB conditions does not learn discriminatory power from repeated acoustic signals. It can also be seen from FIG. 3C that learning does not occur due to repeated sounds.

  In addition to improving behavioral abilities, it was determined whether NFB training changed neural activity. For this, the electroencephalogram for calculating the MMN for each training day was examined in detail. The average MMN amplitude and latency of the group were used as indicators of neural activity. FIG. 4A shows a change in the group average MMN amplitude. This result shows that the average MMN amplitude of the last day group is significantly higher than the first day (t (7) = 2.45, p <0.05). In addition, the daily MMN maximum latency during the subject's NFB training was compared. FIG. 4 (b) shows that the value of the last day of the group average daily maximum MMN latency is considerably shorter than the first day (t (7) = 2.71, p <0.05).

  With regard to perceptual learning, recent studies show that specific brain activity corresponding to direction detection in the visual cortex was trained with recorded fMRI without presenting subjective perceptions of stimuli or target behavior. Our research results support this point of view, causing behavioral performance improvement even when brain activity changes unconsciously. The way to use the recorded fMRI is to determine the brain activity pattern of the subject in advance in order to know the brain region position related to the perceptual learning and how to activate it in the specific perceptual learning. It is necessary to measure. Moreover, since it is necessary to measure brain activity patterns in advance in this way, it is only possible to reinforce problems that can be distinguished to some extent even before learning.

  On the other hand, in the learning method of the present invention, by using the MMN response induced as a result of the auditory process, the perceptual change can be trained even when the specific brain activity pattern and the characteristics of individual sounds are not known. Therefore, this learning method can be used even for tasks that cannot be distinguished at all before learning. This MMN signal can be easily captured. Therefore, the present invention provides useful knowledge for sensory training. For example, in light and right, the brain activity pattern corresponding to the sound of / l / or / r / is not clear, but the brain activity induced by the distinction of / l / or / r /, for example, MMN is / l / And / r /. Compared to the expensive fMRI apparatus, the electroencephalogram apparatus is easy to obtain and use. Furthermore, various advanced functions such as a stable active electrode, a fast-wearing cap, or a compact portable device can be added to the electroencephalogram apparatus.

  Previous studies have shown that: In other words, the MMN is determined by the accuracy of identification and the persistence of sensory memory of sound. Since improvement was confirmed in our behavioral test, it makes sense to think that our NFB method improved discrimination. Several studies have shown that MMN has become stronger through behavioral training. These studies show that the size of the MMN is getting stronger by learning the discrimination task.

  However, following behavioral identification training, MMN was found in subjects who learned behaviorally to identify changes in stimuli. These results show that behavioral acoustic discrimination ability is improved only by enhancing brain activity, and there is no need for behavioral discrimination training. That is, the inventor of the present invention has shown for the first time that the discrimination problem causing the MMN is unconsciously learned by increasing the MMN.

  In addition, behavioral discriminative training can be difficult because behavioral discriminative training style feedback is limited to binary (correct / false) information for behavioral responses. Regardless of how close the brain process is to a novel sound, if the behavior is wrong, it is an “incorrect answer”. Such binary information cannot properly evaluate the outcome of the learning process and makes it difficult for the subject to know what to learn. Therefore, the unconscious learning method using the NFB of the present invention is stronger and more useful than behavior training.

Did the subject notice the purpose of the learning phase?
After the BAD test on the last training day, the subjects were asked why the disk size had changed, and their response was not related to acoustic stimulation in the experiment. This indicates that the subject did not know the purpose of our experiment. In the unconscious learning method using the NFB method of the present invention, the learner does not need to pay attention to the acoustic stimulus, and the learner does not need to know the fact of learning and can improve the discrimination power unconsciously.

  Now we can have an interesting view on acoustic perception training using video games. For example, an incentive for the subject, such as achieving some goal in a video game, is associated with the strength of the MMN. At this time, the player only enjoys the game with the learning sound.

In the present invention, the present learning method can be utilized in that discrimination such as foreign language learning, tone learning, color learning, sommelier training, incense training, etc. is an important skill.
In the above embodiment, an oddball task with two simple sounds that are close in frequency is an example, but when performing identification training on two words that are difficult for a subject to identify, they are different but difficult to identify. The position from the start of the pronunciation of the word is found in advance, and the detection of MMN is performed in the vicinity of the time point, whereby a feedback signal can be obtained and the present invention can be applied.
In addition, when performing the present invention on the sense of smell, the oddball problem can be realized by causing the subject to spout a gas containing an odor component onto the pulse in an odorless wind environment. Can be applied.

  Further, at the place confirmed by the inventor of the present invention, this learning effect extends not only to the stimulus used for learning but also to a similar stimulus. That is, for example, even if the speaker is changed, the learning effect is exhibited, and as a result of learning using the pray / play emitted by the first speaker, the discrimination ability of the pray / play emitted by the second speaker is improved. Furthermore, it was confirmed that the learning effect was demonstrated even with different words containing the same elements, and that the ability to distinguish words using the same r / l, such as right / light, was also improving. Therefore, it can be said that the learning effect according to the present invention has an extension effect.

DESCRIPTION OF SYMBOLS 1 Test subject 2 Stimulus presentation apparatus 3 Earphone 4 Electroencephalograph 5 Electrode 6 Signal processing apparatus 7 Feedback apparatus

Claims (6)

Including a task presentation means for presenting a task signal for discriminating between two discrimination targets to a subject, an electroencephalogram signal detection means for detecting the subject's electroencephalogram signal, and intensity information of the detected electroencephalogram signal, Information indicating the state of the brain wave of the subject is obtained using a learning device using a signal processing means for generating a feedback signal to return and a feedback signal presenting means for presenting the generated feedback signal to the subject. A learning method for improving the information processing ability of the subject by returning to the subject,
The problem presenting means is presented by the oddball method, the electroencephalogram signal is an MMN (mismatch negativeity) signal, the electroencephalogram signal detecting means is an MMN signal detecting means,
(1) The first subject is made to learn about the task signal before the selection using the learning device, and the task is left when the MMN signal exceeds a predetermined threshold for the task by the learning. And the steps to select the learning tasks,
(2) About the learning task including the selected learning task or the stimulus based on the MMN signal, the second subject different from the first subject learns including the learning task or the stimulus serving as the MMN signal. Urging the second subject to increase the intensity of the MMN signal detected using the MMN signal detecting means, while repeatedly presenting the oddball method using the problem presenting means without explaining the problem;
(3) individually presenting the discrimination target in the learning task including a learning task or a stimulus that is a basis of the MMN signal to the second subject and confirming the learning result;
An unconscious learning method using neurofeedback characterized by including   In the step (2) in claim 1, when the learning task including the learning task or the stimulus that is the basis of the MMN signal is repeatedly presented by the oddball method, the discrimination target is exchanged and presented. An unconscious learning method using neurofeedback characterized by inclusion.   3. The unconscious learning method using neuro-feedback according to claim 1, wherein the feedback signal presented to the feedback signal presenting means is a feedback signal using an accumulated value of MMN signal intensity. .   Presenting the video game to the second subject and using the feedback signal as an input signal for controlling the progress of the video game presented to the second subject, the second subject is made to increase the intensity of the MMN signal. The unconscious learning method using the neurofeedback according to any one of claims 1 to 3, wherein the method is an urgent one.   The unconscious learning method using neurofeedback according to any one of claims 1 to 3, wherein the task signal is an acoustic signal.   The unconscious learning method using neurofeedback according to any one of claims 1 to 3, wherein the task signal is a video signal.

Images