JP2011182843A - Method and apparatus for determining human sensation - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and an apparatus for enhancing accuracy of determining human sensations based on electroencephalogram signals. <P>SOLUTION: The apparatus 2 for determining human sensations is equipped with a biological measurement device 4 and an operation part 14. The biological measurement device 4 is configured to measure the electroencephalogram signals and mioelectric signals of a human. The operation part 14 is configured to calculate an electroencephalogram restoration signal, to determine a characteristic value of the electroencephalogram restoration signal by analyzing a frequency of the electroencephalogram restoration signal, and to determine an objective evaluation corresponding to the characteristic value. The electroencephalogram restoration signal is configured to be calculated from the electroencephalogram signal based on the electroencephalogram signal and the mioelectric signal by an independent component analysis (ICA). In the method of determining human sensations, the electroencephalogram restoration signal is calculated by applying the independent component analysis (ICA) on the measured electroencephalogram signal. The human sensations are determined by the electroencephalogram restoration signal. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a method for determining a human sense based on an electroencephalogram and a determination apparatus using the method.

Human sensory evaluation is performed at the product development stage. As an objective evaluation method, various methods for evaluating human senses based on human biological information such as heartbeat, respiration, myoelectric potential, eye movement, blinking, and electroencephalogram have been proposed. Japanese Patent Application Laid-Open No. 2004-138814 discloses a method for evaluating human sensitivity based on an electrocardiogram signal. Japanese Patent Application Laid-Open No. 2002-577 discloses a method for evaluating a human mental state by analyzing an electroencephalogram.
JP 2004-138814 A JP 2002-577 A

  In such an evaluation method, it is difficult to appropriately select biometric information, and thus complicated analysis may be required. Furthermore, the information obtained may not lead to a sufficiently objective evaluation. It is not easy to apply evaluation results obtained by conventional methods to product development.

  The inventors have proposed a method in which human senses can be objectively determined in Japanese Patent Application Nos. 2008-334006 and 2009-196810. In these determination methods, human senses are determined based on the measured electroencephalogram signals. The measured electroencephalogram signal includes an artifact component derived from blinking or body movement muscle activity. By removing the artifact component from the measured electroencephalogram signal, the determination accuracy of human senses can be improved.

  An object of the present invention is to provide a method and apparatus for improving the accuracy of human sense determination based on an electroencephalogram signal.

The human sensory determination method according to the present invention includes:
(1) a step of measuring a human electroencephalogram signal and a muscle action potential signal (myoelectric signal);
(2) A step of obtaining an electroencephalogram restoration signal by independent component analysis (ICA) based on the electroencephalogram signal and the myoelectric signal;
(3) a step in which the arithmetic unit performs frequency analysis of the electroencephalogram restoration signal;
(4) The calculation unit includes a step of determining a feature value of the electroencephalogram restoration signal based on the result of the frequency analysis. (5) The calculation unit includes a step of determining an objective evaluation corresponding to the feature value.

  Preferably, a plurality of restoration signals are obtained in the step of obtaining an electroencephalogram restoration signal by ICA based on the electroencephalogram signal and myoelectric signal. The electroencephalogram restoration signal is determined from the plurality of restoration signals based on the characteristics of the electroencephalogram signal.

  Preferably, in the step of obtaining an electroencephalogram restoration signal by ICA based on the above electroencephalogram signal and myoelectric signal, a restoration signal in which the amplitude spectrum value in a band of 1 Hz or less is larger than the amplitude spectrum values in other frequency bands from the plurality of restoration signals To extract. The extracted electroencephalogram signal is determined as an electroencephalogram restoration signal.

Preferably, in this sense determination method, the determination of the feature value (4) is as follows:
(4-1) A step of extracting a frequency from a plurality of analyzed frequencies,
(4-2) a step of calculating a feature value based on the extracted frequency;
(4-3) evaluating the recognition rate of the calculated feature value;
(4-4) a step in which a frequency is determined based on the recognition rate evaluation;
And (4-5) including a step of determining a feature value based on the determined frequency.

  Preferably, in this sensory determination method, a database is prepared in advance, and this database stores feature value data and subjective evaluation data obtained from frequency analysis. The recognition rate is evaluated (4-3) based on the feature value data and the subjective evaluation data.

  Preferably, in this sensory determination method, frequency extraction (4-1) and frequency determination (4-4) are performed using particle swarm optimization (PSO).

  Preferably, in this sensory determination method, the determination of the feature value (4) is the maximum value, the minimum value, the average value, the variation or the change rate of the values calculated from the power spectrum obtained by the frequency analysis, or of these It is made based on a combination of two or more.

  Preferably, in this sensory determination method, a feature value obtained by frequency analysis and used for human sensory determination and a subjective evaluation heard from the human are added to the characteristic value data and the subjective evaluation data, respectively. And further including the step of updating the database.

    Preferably, the sensory determination method further includes a step of predicting a human reaction or behavior obtained from the stimulus based on the feature value.

  The human sense determination apparatus according to the present invention includes a biometric meter and a calculation unit. This biometer is configured to measure human brain wave signals and myoelectric signals. The arithmetic unit is configured to calculate an electroencephalogram restoration signal, analyze the frequency of the electroencephalogram restoration signal, determine a feature value of the electroencephalogram restoration signal, and determine an objective evaluation corresponding to the feature value. The electroencephalogram restoration signal is configured to be calculated by the ICA based on the electroencephalogram signal and the myoelectric signal from the electroencephalogram signal.

The product evaluation method according to the present invention includes:
Measuring the electroencephalogram and myoelectric signals of the person using the product,
A step of calculating an electroencephalogram restoration signal, a step of performing frequency analysis of the electroencephalogram restoration signal,
Determining a feature value of the electroencephalogram restoration signal based on a result of the frequency analysis, and determining an objective evaluation corresponding to the feature value.
This electroencephalogram restoration signal is calculated by the ICA based on the electroencephalogram signal and the myoelectric signal from the electroencephalogram signal.

  In the sensory determination method according to the present invention, an electroencephalogram restoration signal is obtained from an electroencephalogram signal. A feature value is obtained from the electroencephalogram restoration signal. Based on this feature value, the calculation unit determines the evaluation. In this determination method, it is possible to search for an electroencephalogram restoration signal optimal for determination. This determination method can obtain objective evaluation with higher accuracy.

FIG. 1 is a block diagram illustrating a sensory determination device according to an embodiment of the present invention. FIG. 2 is a flowchart showing a sensory determination method using the apparatus of FIG. FIG. 3 is a schematic diagram showing an example of the structure of a database of feature value data and subjective evaluation data used in the method of FIG. FIG. 4 is an explanatory diagram illustrating the conversion state of an electroencephalogram signal by the apparatus of FIG. FIG. 5 is a graph showing an example of a power spectrum obtained by frequency analysis of the method of FIG. FIG. 6 is a schematic diagram in which feature values of the method of FIG. 2 are represented in space. FIG. 7 is a flowchart illustrating a feature value determination method of the method of FIG. FIG. 8 is a conceptual diagram showing the particle group of FIG.

  Hereinafter, the present invention will be described in detail based on preferred embodiments with appropriate reference to the drawings.

  As shown in FIG. 1, the apparatus 2 includes a biometer 4, a filter 6, an ICA analyzer 8, a converter 10, a storage unit 12, a calculation unit 14, a monitor 16, a storage unit 18, and an input unit 20. ing.

  FIG. 2 shows a sensory determination method using the device 2 of FIG. In this method, an electroencephalogram signal is first measured by the biometer 4. An action potential signal (hereinafter, referred to as a myoelectric signal) of muscle strength at a site of N (N is a natural number of 1 or more) is measured by the biometer 4 (STEP 1). N portions when N is a plurality are different portions. By this measurement, an electroencephalogram signal and N myoelectric signals are obtained. The electroencephalogram signal and myoelectric signal are passed through the filter 6. Noise is removed by the filter 6 (STEP 2).

  The electroencephalogram signal and N myoelectric signals from which noise has been removed are sent to the ICA analyzer 8. The ICA analyzer 8 converts this electroencephalogram signal into a restoration signal by independent component analysis (ICA) (STEP 3). Here, the electroencephalogram signal and the myoelectric signal are converted into (N + 1) restoration signals based on the electroencephalogram signal and the N myoelectric signals. This restoration signal is sent to the converter 10. This restoration signal is converted into a digital signal (STEP 4). This digital restoration signal is stored in the storage unit 12 (STEP 5). The conversion to the restoration signal may be performed after the conversion to the digital signal. From measurement (STEP 1) to storage (STEP 5) may be repeated a plurality of times.

  The calculation unit 14 extracts the digital restoration signal from the storage unit 12. The calculation unit 14 performs frequency analysis of the digital restoration signal (STEP 6). A digital restoration signal representing an electroencephalogram signal is determined from (N + 1) digital restoration signals. This determination is made based on the characteristics of the measured electroencephalogram signal. This determined signal is used as an electroencephalogram restoration signal.

  The calculation unit 14 determines the feature value of the electroencephalogram restoration signal (STEP 7). Further, the calculation unit 14 determines an objective evaluation corresponding to the feature value (STEP 8). This determination is made based on a database stored in the storage unit 18 (typically a hard disk). By these steps, human senses are determined.

  Hereinafter, the determination procedure by the apparatus 2 is illustrated using the putter of a golf club. In this example, the sense of the subject (human) who strokes is determined by the device 2 of FIG. More specifically, the sense of “easy to pull” or “difficult to pull” when taking back the putter is determined. This determination procedure will be described in detail. The putter is also evaluated by this sensory determination method. The subject's stroke of the putter corresponds to “use of product” in the present invention.

  In this determination method, a database stored in the storage unit 18 is used. The structure of this database is schematically shown in FIG. This database was obtained by letting 20 humans make 10 strokes with 2 types of putters. This database includes feature value data (X (n), Y (n)) and subjective evaluation data. This database contains 40 records. One record includes feature value data (X (n), Y (n)) and subjective evaluation data. A method for calculating the feature value data (X (n), Y (n)) will be described in detail later.

  This subjective evaluation data is heard from a person who strokes the putter. This interview is easy to pull. By calculating feature values and listening for subjective evaluations 40 times, a database with 40 records can be obtained. The subject strokes a plurality of times with one putter, and the calculation of one feature value data and subjective evaluation are obtained. Here, the subject strokes 10 times.

  This database preferably has a large number of records, but may be less than 40. For example, a database may be constructed by calculating feature values and listening for subjective evaluations by six people with two types of putters. In creating this database, three or more types of putters may be used.

  In FIG. 3, two feature value data (X (n), Y (n)) are obtained. A two-dimensional position vector is obtained from the feature value data (X (n), Y (n)). The position vectors of these 40 records are positioned on a two-dimensional space. In this space, an identification boundary is calculated. The 40 position vectors are divided into class A and class B by the identification boundary. Feature value data of subjective evaluation data “easy to pull” belongs to one of class A and class B, and feature value data of subjective evaluation data “difficult to pull” mainly belongs to the other. The frequency selection method, the feature value data (X (n), Y (n)), and the identification boundary calculation method include particle group optimization (hereinafter referred to as PSO) and a support vector machine (hereinafter referred to as the following). SVM) is used. Although not shown in FIG. 3, this identification boundary is also recorded in the recording unit 18.

  In this determination method, a device known as the biometer 4 is used. As a preferable apparatus, a trade name “Polymate II” of TEAC and a trade name “Brain Builder Unit” of Ability Development Laboratory are exemplified.

  Here, an example will be described in which the biometer 4 measures an electroencephalogram signal and four myoelectric signals. The first electrode of the biometer 4 is attached to the subject's scalp. One reference electrode is attached to the earlobe. Preferably, an electrode is attached to Fp1 or Fp2 defined in the International 10-20 Act. Here, an electrode is attached to Fp1 defined in the International 10-20 Act.

  The second electrode of the biometer 4 is attached to the subject's right arm. The second electrode is attached to the epidermis where the right upper biceps is located. The third electrode is attached to the subject's right shoulder. The third electrode is attached to the epidermis where the right deltoid muscle is located. The fourth electrode is attached to the subject's right breast. The fourth electrode is attached to the epidermis where the right greater pectoral muscle is located. The fifth electrode is attached to the back of the subject's right neck. The fifth electrode is attached to the epidermis where the right trapezius muscle is located. In this way, the second to fifth electrodes are attached to different parts.

  The subject to whom the electrode is attached has a putter. Take a stance at the position where the subject hits the ball. Subject strokes putter. The subject takes back the putter during this stroke. Subject hits the ball with a putter. The subject strokes 10 times. After that, the subject feels comfortable using the putter. This comfort is a subjective evaluation. Specifically, this putter is asked whether it is “easy to pull” or “not easy to pull” during takeback.

  An electroencephalogram signal and myoelectric signal are measured from the time when the subject starts taking putter to 1023 msec. FIG. 4A shows an example of the measured electroencephalogram signal. The electroencephalogram signal and myoelectric signal are passed through the filter 6. This filter 6 has a low-pass function. The filter 6 removes a high frequency band from these signals. As for the electroencephalogram signal and the myoelectric signal, signals in a frequency band exceeding 40 Hz are removed. FIG. 4B illustrates a signal obtained by passing the electroencephalogram signal of FIG. 4A through the filter 6. The electroencephalogram signal and myoelectric signal that have passed through the filter 6 are sent to the ICA analyzer 8.

  The ICA analyzer 8 converts the electroencephalogram signal and the myoelectric signal from restoration signals 1 to 5. The electroencephalogram signal and the myoelectric signal are converted into five restoration signals. FIG. 4C shows an example of the restoration signal obtained in this way. The ICA analyzer 8 performs independent component analysis on the electroencephalogram signal and the myoelectric signal. Independent component analysis is a method for obtaining an original signal by separating mixed signals from a plurality of signal sources into statistically independent signals.

  It is assumed that the measured electroencephalogram signal is mixed with the action potential component of muscle strength in addition to the electroencephalogram component. The purpose of this independent component analysis is to remove myoelectric artifacts from the measured electroencephalogram signal. The electroencephalogram signal is restored by applying ICA to the measured electroencephalogram signal and myoelectric signal.

  The converter 10 converts the restored signals 1 to 5 from analog signals to digital signals. The digital restoration signals 1 to 5 are stored in the storage unit 12.

  The calculation unit 14 in FIG. 1 performs frequency analysis of the digital restoration signals 1 to 5. In frequency analysis, a power spectrum is obtained from an electroencephalogram signal by FFT (Fast Fourier Transform). Here, the frequency analysis of the converted signal in the frequency band from 0 to 40 Hz is performed.

  The measured electroencephalogram signal has an amplitude spectrum peak in a frequency band of 1 Hz or less. On the other hand, the measured myoelectric signal does not have an amplitude spectrum peak in a frequency band of 1 Hz or less. One of the features of the electroencephalogram signal is that an amplitude spectrum peak exists in a frequency band of 1 Hz or less. Reconstructed signals 1 to 5 having this characteristic are extracted. For example, a restoration signal having the largest amplitude spectrum peak in a frequency band of 1 Hz or less is extracted. This extracted restoration signal is determined to represent an electroencephalogram signal. This restoration signal is used as an electroencephalogram restoration signal.

  FIG. 5 is a graph illustrating the power spectrum of the electroencephalogram restoration signal obtained by frequency analysis. As shown in FIG. 5, by this frequency analysis, a power value is used for each frequency of 1 Hz from 7 Hz to 30 Hz.

  The electroencephalogram restoration signal is divided into a front signal and a rear signal. The previous signal is a signal from the start of takeback to 511 msec. The rear signal is a signal from 512 msec to 1023 msec. The power spectrum of the front signal and the power spectrum of the rear signal are respectively obtained.

A power value at a specific frequency is extracted from the power spectrum of the previous signal. For one putter, 20 power values of a specific frequency are obtained. This power value of 10 is obtained from the previous signal. The other power value 10 is obtained from the post signal. An average value of the power values of the previous signal is calculated. This average value is defined as a feature value X _(n) . Here, n is an integer indicating a specific frequency from 7 to 30. The average value of the power value of the rear signal is calculated. This average value is defined as a feature value Y _(n) . Similarly, the average value of the power value of the previous signal and the average value of the power value of the subsequent signal are calculated for other patterns, and the feature value X _(n) and the feature value Y _(n) are obtained. In this way, two records are obtained by two kinds of putters.

A two-dimensional position vector is obtained from the feature value data (X _(n) , Y _(n) ) obtained in this way. This two-dimensional position vector is positioned on the two-dimensional space of FIG. Whether the position vector belongs to class A or class B is determined by the identification boundary. Thereby, an objective determination of “easy to pull” or “difficult to pull” is made.

  When the feature value data belongs to class A, “favorability A” is output to the monitor 16. When the feature value data belongs to class B or is on the identification boundary, “favorability B” is output to the monitor 16. Thereby, objective determination is made.

  The calculation unit 14 may add a position vector obtained from a new subject to the database. The subjective evaluation heard from the subject is input from the input unit 20 (for example, a keyboard). The computing unit 14 adds this subjective evaluation to the subjective evaluation data in the database. The added position vector and the subjective evaluation data form one record. The database is updated by adding the position vector and the subjective evaluation data. By repeating the update of the database, the number of records in the database increases and the objectivity of the determination is improved.

  In this embodiment, two types of objective evaluations are performed, ie, likability A and B. In this embodiment, two types of subjective evaluation are used: “easy to pull” and “hard to pull”. The number of types of subjective evaluation may be three or more. In this embodiment, the number of data records is 40, but it may be smaller or larger.

In this embodiment, the feature value data (X _(n) , Y _(n) ) is calculated from the average value of the power values, but the standard deviation σx _(n) of the power value of the previous signal and the power of the rear signal Feature value data (X _(n) , Y _(n) ) may be calculated from the standard deviation σy _(n) of the values. A position vector may be obtained from the feature value data (X _(n) , Y _(n) ), and an identification boundary may be calculated. Furthermore, instead of the average of the power values, a maximum value, a minimum value, variation, a change rate, or a combination of two or more of these may be used.

  FIG. 7 shows a method for determining feature values. Along with the determination of the feature value, the frequency is selected and the identification boundary is calculated. With reference to FIG. 7, a method of determining the feature value, selecting a frequency, and calculating an identification boundary will be described. PSO and SVM are used for the frequency selection method and the feature value data and identification boundary calculation method.

  In selecting the frequency, the frequency is selected by PSO. The selection of the frequency includes selecting a combination of frequencies. The calculation unit 14 generates an arbitrary number of particle groups. FIG. 8 schematically shows the generated particle group. Here, a particle group composed of five particles 1 to 5 is generated. The particles 1 to 5 have real values corresponding to respective frequencies from 7 Hz to 30 Hz. The real number is given as a random number from 0 to 1.00. Here, an example in which five particles are generated is illustrated, but fewer particles may be generated, or more particles may be generated.

  The frequency with the largest real value is extracted. In FIG. 6, real values of some frequencies are shown and other real values are omitted, but in the particle 1, the highest frequency is 9 Hz. In the particle 1, 9 Hz is extracted as a frequency.

From the result of the frequency analysis, the average power value of the frequency 9 Hz of the previous signal is calculated. Thereby, the feature value data X ₍₉₎ is obtained. Similarly, the average power value of the rear signal frequency of 9 Hz is calculated. Thereby, feature value data Y ₍₉₎ is obtained. Characteristic value data (X ₍₉₎ , Y ₍₉₎ ) at ₉ Hz is obtained from these data.

A two-dimensional position vector is obtained from the feature value data (X ₍₉₎ , Y ₍₉₎ ) obtained in this way. In this manner, 40 record position vectors are obtained. In this way, position vectors obtained from 40 records are positioned in a two-dimensional space. An identification boundary is calculated by SVM from the subjective evaluation data of the corresponding records with these 40 position vectors. The combination of frequencies, the feature value data (X ₍₉₎ , Y ₍₉₎ ), and the identification boundary are stored in the storage unit 18.

The computing unit 14 calculates a recognition rate from the subjective evaluation data and the identification boundary of each record. The recognition rate is the ratio of the number of data C whose subjective evaluation data matches the determination result to the total number of determinations A. Here, the evaluation function RA shown in the following equation is calculated as the recognition rate.
RA = (C / A)
This evaluation function RA is recorded in the recording unit 18.

  Similarly, for particles 2 through 5, the frequency is extracted. Feature value data is calculated from the extracted frequency. An identification boundary is calculated from the feature value data and the subjective evaluation data. The extracted frequency, feature value data, and identification boundary are stored in the storage unit 18. The calculation unit 14 calculates an evaluation function RA from the feature value data, the subjective evaluation data, and the identification boundary. This evaluation function RA is recorded in the recording unit 18.

  When the evaluation function is calculated for all particles, the particles in the particle group are updated. The calculation unit 14 specifies the best solution among the evaluation functions of the particles 1 to 5. The best solution is the particle with the largest evaluation function. For example, when the evaluation function of the particle 1 is the largest, the particle 1 is stored in the recording unit 18 as the best solution of this particle group.

For particles 1 to 5, the position vector and velocity vector of each particle are updated by the following equations.
V _i (t + 1) = wV _i (t) + c ₁ r ₁ (P _b (t) −X _i (t)) + c ₂ r ₂ (P _g (t) −X _i (t))
X _i (t + 1) = X _i (t) + V _i (t + 1)
_i : particle index X _i (t): position vector of i-th particle in t-th iteration V _i (t): velocity vector of i-th particle in t-th iteration w: weighting factor (positive) value)
c ₁ and c ₂ : weighting factors r ₁ and r ₂ : random numbers from 0 to 1.00 P _b (t): the best solution (position vector) of the particle group in the t-th iteration
P _g (t): Best solution (position vector) from the start of search in the t-th iteration
Here, the weighting factors w, c _1, and c ₂ are determined by performing an iterative simulation and detecting parameter values that can empirically obtain a good classification result. Here, w is 0.10, c ₁ is 0.50, and c ₂ is 0.50.
In this way, based on the updated position vectors X _i (t + 1) of the particles 1 to 5, the particles 1 to 5 shown in FIG. 6 are updated with real values corresponding to respective frequencies from 7 Hz to 30 Hz. And updated to a new particle group.

With respect to the updated particle group, the frequency is extracted for each of the particles 1 to 5 in the same manner as the generated particle group described above. A position vector for each record is calculated. An identification boundary is calculated. An evaluation function is calculated. Then, the updated best solution P _b (2) of the particle group is specified. For example, in this way, in the updated particle group, the particle 3 is specified as the best solution P _b (2).

The evaluation function of particle 1 that is the best solution P _b (1) of the first particle group is compared with the evaluation function of particle 3 that is the best solution P _b (2) of the updated particle group. Of the best solution P _b (1) and the best solution P _b (2), the larger evaluation function is specified as the best solution P _g (2) from the start of the search. For example, if the best solution P _b (2) is larger than the best solution P _b (1), this particle 3 is the best solution P _b (2) of this group and the best solution P _g (2 from the start of the search). It is also.

The position vector X _i (t) and velocity vector V _i (t) of the particle group are updated by the best solution P _b (1) and the best solution P _g (2). With this updated particle group, the steps are similarly repeated from the extraction of the frequency to the update of the next particle group. This repetition is repeated a predetermined m times. For example, when m = 10, when the repetition reaches 10 times, the update of the particle group is completed.

By repeating this 10 times, the best solution P _g (10) is obtained. The frequency of the particle of this best solution P _g (10) is determined as the selected frequency. For example, 9 Hz is determined as the frequency in this way.

Based on the determined frequency, the average value X _(n) of the previous signal and the average value Y _{(n) of the} subsequent signal are calculated. Feature value data (X _(n) , Y _(n) ) is calculated. Here, n is 9. A two-dimensional position vector is calculated from the characteristic value data of the frequency thus obtained. The position vector and subjective evaluation data of each record are stored in the database. The database shown in FIG. 3 schematically shows the data stored for 40 records in this way.

  Here, the number of updates of the particles was set to 10 times. By repeating this 10 times, the frequency having the largest evaluation function is searched. PSO is used for this frequency search. This makes it possible to search for an optimum frequency. This search can be performed at high speed. This search can be implemented in the arithmetic unit 14 with a simple algorithm. Here, the number of repetitions is ten, but the number of repetitions may be determined in advance and may be more or less than ten.

  According to the method of the present invention, the influence of myoelectricity is reduced from the measured electroencephalogram signal. Thereby, a signal close to the original brain wave signal is extracted. Based on the extracted signal, a feature value is determined, and a human sense is determined. Thereby, the determination accuracy can be improved.

  Here, the feature value is determined using the SVM, but the feature value may be determined by another method. For example, CMAC or neural network (NN) may be used instead of SVM. The CMAC or neural network (NN) may be applied to the restoration signal obtained by the ICA analysis to perform human sensory evaluation.

  Here, the myoelectric signal is measured from the epidermis where the right upper biceps, right deltoid, right pectoralis and right trapezius are located, but may be measured from other parts of the body. When the putter is taken back, an electromyogram signal assumed to be mixed in the electroencephalogram signal is measured. When the putter is taken back, the electromyogram of the active muscle is measured. When using the product, it is only necessary to measure the electromyogram of the active muscle.

[Example]
Two types of putters were prepared. These two kinds of putters were stroked by three subjects. Each subject stroked two putters 10 times each. The subjective evaluation at that time was heard. With this as one trial, 10 trials were carried out. Based on the electroencephalogram signal and myoelectric signal obtained in this trial and the subjective evaluation, the sensory determination of the subject was made by the sensory determination method of FIG.

  The feature value was determined by the method shown in FIG. Here, a database consisting of 60 records was obtained. Based on this feature value, an identification boundary was calculated. Based on this identification boundary, the position vector of 60 records was divided into class A and class B. The coincidence rate between the classified result and the subjective evaluation data, that is, the recognition rate RA was calculated.

[Comparative example]
On the other hand, 60 records were obtained in the same manner as in Example 1 except that an electroencephalogram signal without ICA analysis was used. The recognition rate RA was calculated for these 60 records in the same manner as in the first embodiment.

  The greater the value of the recognition rate RA, the higher the correct answer rate. From this example and the comparative example, it was confirmed that the recognition rate RA of the example increases. The discrimination method according to the present invention can objectively determine human subjectivity with higher accuracy.

  By the method according to the present invention, it is possible to determine the feel at impact of a golf ball, the comfort of shoes, the feeling of wearing gloves, the feeling of wearing of wear, and the like.

DESCRIPTION OF SYMBOLS 2 ... Sensory determination apparatus 4 ... Biometer 6 ... Filter 8 ... ICA analyzer 10 ... Converter 12 ... Storage part 14 ... Calculation part 16 ... Monitor 18 ... Storage unit 20 ... Input unit

Claims (11)

(1) a step of measuring a human electroencephalogram signal and a muscle action potential signal (myoelectric signal);
(2) A step of obtaining an electroencephalogram restoration signal by independent component analysis (ICA) based on the electroencephalogram signal and the myoelectric signal;
(3) a step in which the arithmetic unit performs frequency analysis of the electroencephalogram restoration signal;
(4) A step of determining a feature value of the electroencephalogram restoration signal based on a result of frequency analysis, and (5) a human sensory determination method including a step of determining an objective evaluation corresponding to the feature value by the calculation unit . In the step of obtaining an electroencephalogram restoration signal by ICA based on the above electroencephalogram signal and myoelectric signal, a plurality of restoration signals are obtained,
The sensory determination method according to claim 1, wherein an electroencephalogram restoration signal is determined based on characteristics of the electroencephalogram signal from the plurality of restoration signals. In the step of obtaining an electroencephalogram restoration signal by ICA based on the above electroencephalogram signal and myoelectric signal,
The restoration signal of which the amplitude spectrum value in the band of 1 Hz or less is larger than the amplitude spectrum value of the other frequency band from the plurality of restoration signals, and the extracted electroencephalogram signal is determined as the electroencephalogram restoration signal. Sensory judgment method. The determination of the feature value (4)
(4-1) A step of extracting a frequency from a plurality of analyzed frequencies,
(4-2) a step of calculating a feature value based on the extracted frequency;
(4-3) evaluating the recognition rate of the calculated feature value;
(4-4) a step in which a frequency is determined based on the recognition rate evaluation;
And (4-5) The sensory determination method according to any one of claims 1 to 3, including a step of determining a feature value based on the determined frequency. A database is prepared in advance, and this database stores feature value data and subjective evaluation data obtained from frequency analysis,
The sensory determination method according to claim 4, wherein the recognition rate evaluation (4-3) is performed based on feature value data and subjective evaluation data.   The sensory determination method according to claim 4 or 5, wherein the frequency extraction (4-1) and frequency determination (4-4) are performed using particle swarm optimization (PSO).   The determination of the feature value (4) is made based on the maximum value, minimum value, average value, variation or change rate of values calculated from the power spectrum obtained by frequency analysis, or a combination of two or more of these values. The sensory determination method according to claim 1.   The feature value obtained by the frequency analysis and used for the human sense determination and the subjective evaluation heard from the human are added to the feature value data and the subjective evaluation data, respectively, and the database is updated. The sensory determination method according to claim 1, further comprising a step.   The sensory determination method according to claim 1, further comprising a step of predicting a human reaction or behavior obtained from a stimulus based on the feature value. It is equipped with a biometer and a calculation unit,
This biometer is configured to measure human brain wave signals and myoelectric signals,
The calculation unit is configured to calculate an electroencephalogram restoration signal, determine the feature value of the electroencephalogram restoration signal by frequency analysis of the electroencephalogram restoration signal, and determine an objective evaluation corresponding to the feature value,
A human sensory determination device configured such that an electroencephalogram restoration signal is calculated by an ICA based on an electroencephalogram signal and an electromyogram signal from the electroencephalogram signal. Measuring the electroencephalogram and myoelectric signals of the person using the product,
A step of calculating an electroencephalogram restoration signal, a step of performing frequency analysis of the electroencephalogram restoration signal,
Determining a characteristic value of the electroencephalogram restoration signal based on a result of the frequency analysis, and determining an objective evaluation corresponding to the characteristic value,
A product evaluation method in which the electroencephalogram restoration signal is calculated by the ICA based on the electroencephalogram signal and the electromyogram signal from the electroencephalogram signal.

Images