JP2015009047A - Sensitivity state determination device, sensitivity state determination method, and computer program for sensitivity state determination - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sensitivity state determination device which demonstrates higher determination accuracy with using an electrocardiographic signal than in the past.SOLUTION: A beat interval average value calculation section 4 calculates an average value of beat intervals from electrocardiographic signals containing a plurality of R-waves acquired from a person to be measured. A delay time calculation section 5 calculates a delay time as a parameter having a time dimension required for reconfiguration of attractor in order to calculate Liapunov index from the electrocardiographic signal. A Liapunov dimension calculation section 6 calculates a Liapunov dimension of the electrocardiographic signal using the delay time. A sensitivity state determination section 7 determines the sensitivity state of the person to be measured by making the average value of beat intervals, delay time, and Liapunov dimension as feature quantity.

Description

  The present invention relates to a sensory state determination device, a sensory state determination method, and a computer program for sensory state determination that determine a sensory state of a measurement subject using an electrocardiogram signal.

  Japanese Patent Laid-Open No. 2006-150065 (Patent Document 1) describes a subject's autonomic nerve by determining a heart rate displacement degree of the subject using the RR interval of the electrocardiogram signal of the subject as a feature amount. An apparatus for inspecting is disclosed.

  Japanese Patent Laying-Open No. 2012-63885 (Patent Document 2) obtains a Lyapunov index from biological information such as pulsating flow information of a measured person, and determines whether or not the measured person is concentrated from the Lyapunov index. A biological information management device is disclosed.

 Furthermore, in Japanese Patent Laid-Open No. 2003-233601 (Patent Document 3), when the maximum Lyapunov exponent estimated value of the measured person's pulse group is monitored every 10 s, according to the measured person's mental, emotional, and behavioral state, It is described that the value varies in the range of 5-12.

JP 2006-150065 A JP 2012-063885 A JP 2003-233601A [0097] Paragraph

 In the technique described in Patent Literature 1, the autonomic nerve (a mind-body state or a sensory state) of the measurement subject is determined based on the degree of heartbeat displacement. However, the heartbeat displacement degree is unstable as a feature quantity, and there is a limit to increasing the determination accuracy.

 In addition, there is a limit to increasing the determination accuracy even when determining the mental / emotional / behavioral state of the person to be measured using the Lyapunov exponent shown in Patent Document 2 or the maximum Lyapunov exponent estimated value shown in Patent Document 3 as a feature quantity. is there.

 An object of the present invention is to provide an emotional state determination device, an emotional state determination method, and a computer program for emotional state determination with higher determination accuracy than in the past when an electrocardiogram signal is used.

  The sensitivity state determination device according to the first invention of the present application includes a pulsation interval average value calculation unit, a delay time calculation unit, and a sensitivity state determination unit in order to determine the sensitivity state of the person to be measured using an electrocardiogram signal. Prepare. The pulsation interval average value calculation unit obtains an average value of pulsation intervals from an electrocardiogram signal including a plurality of R waves obtained from the measurement subject. In this specification, the R wave is a signal waveform of one cycle that repeatedly appears in an electrocardiogram signal. The delay time calculation unit calculates a delay time as a parameter having a time dimension required when performing attractor reconstruction in order to obtain a Lyapunov exponent from an electrocardiogram signal. The sensitivity state determination unit determines the sensitivity state of the measurement subject using the average value of the pulsation interval and the delay time as feature amounts.

  The sensitivity state determination device according to the second aspect of the present invention further includes a Lyapunov dimension calculation unit that obtains a Lyapunov dimension of an electrocardiogram signal using a delay time, and the sensitivity state determination unit includes an average value of a pulsation interval, a delay time, and a Lyapunov. Using the dimension as a feature amount, the sensitivity state of the measurement subject is determined.

  The inventor studied to determine the sensitivity state of the measurement subject based on the feature quantity based on the heartbeat and the feature quantity based on the chaos analysis of the electrocardiogram signal as a biological signal. As a result, an average value of the beat interval calculated from the electrocardiogram signal including a plurality of R waves obtained from the measurement subject, and a delay time required when performing the attractor reconstruction to obtain the Lyapunov exponent from the electrocardiogram signal And the Lyapunov dimension was found to be a feature quantity for determining the sensitivity state. In particular, it is important to find that the delay time required when attractor reconstruction is performed in order to obtain the Lyapunov exponent used for chaos analysis is one of the characteristic quantities for determining the sensitivity state. When the delay time is included in the feature amount, the determination accuracy can be significantly improved as compared with the case where only the average value of the beat interval is used as the feature amount. Furthermore, when the Lyapunov dimension, not the Lyapunov exponent, is included as the feature amount, the determination accuracy can be further improved as compared with the case where the beat interval and the delay time are used as the feature amount. The present invention and its effects are all found and confirmed based on research results.

  The pulsation interval average value calculation unit calculates an average value of intervals (RR intervals) between maximum amplitudes (R waves) of a plurality of electrocardiogram signals as an average value of pulsation intervals. As the electrocardiographic signal, an electrocardiographic signal within a predetermined sampling time (for example, 1 minute) is used. The delay time calculation unit calculates the time τ when the mutual information amount calculated based on the joint probability density of the electrocardiogram signal v (t) at time t and the electrocardiogram signal v (t + τ) delayed by time τ from time t. Is calculated as a delay time. This delay time is a time required for the electrocardiogram signal v (t) and the electrocardiogram signal v (t + τ) to be statistically independent, and is an amount that characterizes the electrocardiogram signal in terms of information theory.

  The Lyapunov dimension calculation unit calculates a Lyapunov exponent from a reconstructed attractor obtained by reconstructing an electrocardiogram signal into an n-dimensional delay coordinate system using a delay time, and calculates the Lyapunov dimension from the calculated n Lyapunov exponents. calculate. The Lyapunov dimension is considered to represent the degree of freedom of ECG signal dynamics estimated from the instability of ECG signals. By the way, in the case of an electrocardiographic signal, it has been confirmed by experiments that “n-dimensional” is preferably “four-dimensional”.

  In addition, the emotional state determination unit specifies and stores in advance a classification plane that can separate one or more emotional states from data obtained by measuring feature amounts when a plurality of subjects feel a plurality of types of emotional states. In addition, the sensitivity state of the measurement subject is determined based on the feature amount obtained from the measurement subject and the classification plane. If such a classification plane is determined in advance, the sensitivity state can be determined easily and with high accuracy by pattern classification.

  In the first sensibility state determination method of the present invention, the processor of the computer has a pulsation interval average value calculation function for obtaining an average value of pulsation intervals from an electrocardiogram signal including a plurality of R waves obtained from the measurement subject. A delay time calculation function for calculating a delay time as a parameter having a time dimension required for performing an attractor reconstruction to obtain a Lyapunov exponent from an electrocardiogram signal, an average value of the pulsation interval and the delay The sensitivity state of the measurement subject is determined by realizing a sensitivity state determination function that determines the sensitivity status of the measurement subject using time as a feature amount.

 In the second sensitivity state determination method of the present invention, the processor of the computer has a beat interval average value calculating function for obtaining an average value of the beat interval from an electrocardiogram signal including a plurality of R waves obtained from the measurement subject. A delay time calculation function for calculating a delay time as a parameter having a time dimension required for performing an attractor reconstruction to obtain a Lyapunov exponent from the electrocardiogram signal, and the electrocardiogram signal using the delay time By implementing a Lyapunov dimension calculation function for determining the Lyapunov dimension of the subject, and a sensitivity state determination function for determining the sensitivity state of the person being measured using the average value of the beat interval, the delay time, and the Lyapunov dimension as feature quantities, The sensitivity state is determined.

 A first computer program for determining a sensibility state according to the present invention includes a pulsation interval average value calculating function for obtaining an average value of pulsation intervals from an electrocardiogram signal including a plurality of R waves obtained from a measurement subject, and an electrocardiogram signal. A delay time calculation function for calculating a delay time as a parameter having a time dimension necessary for performing attractor reconstruction to obtain the Lyapunov exponent from the average value of the beat interval and the delay time as feature quantities As described above, the sensitivity state determination function for determining the sensitivity state of the measurement subject is realized by the processor in the computer.

 A second computer program for determining a sensibility state according to the present invention includes a pulsation interval average value calculating function for obtaining an average value of pulsation intervals from an electrocardiogram signal including a plurality of R waves obtained from a measurement subject, and an electrocardiogram signal. A delay time calculation function for calculating a delay time as a parameter having a time dimension required for attractor reconstruction in order to obtain a Lyapunov exponent, and a Lyapunov dimension of the electrocardiogram signal using the delay time. The processor in the computer realizes the Lyapunov dimension calculation function to be obtained and the sensitivity state determination function for determining the sensitivity state of the measurement subject using the average value of the beat interval, the delay time, and the Lyapunov dimension as the feature quantities.

 The sensitivity state determination computer program is preferably stored in a computer-readable storage means.

It is a block which shows roughly the structure of an example of embodiment of the sensitivity state determination apparatus of this invention. Is a diagram used to explain the computing the beat interval T ₁ through T _M (generally referred to as R-R interval). It is a diagram showing the relationship between the average value Tmean pulsation interval T ₁ through T _M and pulsation interval. It is a figure used in order to explain the intuitive meaning of mutual information amount. It is a figure used in order to explain the calculation method of delay time. It is a figure used in order to explain the calculation method of delay time. It is a figure which shows the calculation procedure of a Lyapunov dimension. It is a figure which shows the structure of the sensitivity state determination part. It is a figure used in order to explain the classification plane used by pattern classification. It is a flowchart which shows the algorithm of the computer program installed in a computer, when using the average value Tmean of a pulsation interval, and delay time (tau) as a feature-value, and determining a sensitivity state using a computer. Mean value Tmean pulsation intervals, as the characteristic amount of the delay time τ and Lyapunov dimension D _L, when determining the emotional state using a computer, is a flowchart showing an algorithm of a computer program installed in the computer. It is a figure which shows the implementation protocol of experiment. It is a figure which shows experimental data. It is a figure which shows the test result at the time of determining the isolation | separation of a resting state only about the average value Tmean of a pulsation interval about the test subject 4 as a feature vector. It is a figure which shows the test result when the isolation | separation of a resting state is determined about the test subject 3 by using the average value Tmean and delay time (tau) of a pulsation interval as a feature vector. It is a figure which shows the test result at the time of determining the separation of a resting state about the test subject 5 by making the average value Tmean of pulsation interval and delay time (tau) into a feature vector. It is a figure which shows the test result at the time of determining the separation of a resting state about the test subject 7 by making the average value Tmean and delay time (tau) of a pulsation interval into a feature vector. The average value Tmean pulsation intervals for subjects 3 is a diagram showing the test results when determining the separation of the rest state as a feature vector of the delay time τ and Lyapunov dimension D _L. The average value Tmean pulsation intervals for subjects 5 is a diagram showing the test results when determining the separation of the rest state as a feature vector of the delay time τ and Lyapunov dimension D _L. The average value Tmean pulsation intervals for subjects 7 is a diagram showing the test results when determining the separation of the rest state as a feature vector of the delay time τ and Lyapunov dimension D _L.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram schematically showing a configuration of an example of an embodiment of a sensitivity state determination apparatus according to the present invention. In FIG. 1, a block denoted by reference numeral 1 is a person to be measured, and a block denoted by reference numeral 2 is a biological amplifier including an electrocardiograph. This biological amplifier 2 is attached to the left chest of the measurement subject, and an electrocardiographic signal v (t) including a plurality of R waves based on the heartbeat of the measurement subject for a predetermined sampling time T (for example, 1 minute). Is output. The R wave is a signal waveform of one cycle that repeatedly appears in the electrocardiogram signal v (t) shown in FIG. An electrocardiogram signal v (t) including a plurality of R waves is input to the heartbeat detection unit 3. As shown in FIG. 2, the heartbeat detection unit 3 has a time interval between portions having the maximum amplitude of two adjacent R waves, that is, pulsation fluctuations or pulsation intervals T _{1 to} T _M (generally R−R). (Referred to as interval). The pulsation interval average value calculation unit 4 calculates an average value Tmean of pulsation intervals based on the following formula. FIG. 3 illustrates the relationship between the beat intervals T _{1 to} T _M and the average value Tmean of the beat intervals.

In the above equation, T _k is the time interval (RR interval) between the k th R wave and the k + 1 th R wave of the electrocardiogram signal. M is the number of heartbeats during time T.

 The delay time calculation unit 5 in FIG. 1 calculates a delay time τ as a parameter having a time dimension required when performing attractor reconstruction in order to obtain a Lyapunov exponent from an electrocardiogram signal. The delay time τ is a time required for the electrocardiogram signal to become statistically independent from itself, and is an amount that characterizes the electrocardiogram signal in terms of information theory. The inventor has found through experiments that the delay time τ required when performing attractor reconstruction in order to obtain the Lyapunov exponent used for chaos analysis is one of the characteristic quantities for determining the sensitivity state. It was confirmed by experiments that the inclusion of the delay time τ in the feature value can significantly improve the determination accuracy compared to the case where only the average value Tmean of the pulsation interval is used as the feature value. The delay time τ is a parameter calculated so that the mutual information amount calculated based on the joint probability density of the electrocardiogram signals v (t) and v (t + τ) is minimized. To explain conceptually, as shown in FIG. 4, the mutual information amount I of the average information amount (entropy) possessed by the electrocardiogram signal v (t) and the average information amount (entropy) possessed by v (t + τ). The time when (information overlap) is minimized is the delay time τ.

Regarding the calculation of the delay time τ, Andrew M. Fraser and Harry L. Swinney have published a paper entitled “Independent coordinates for strange attractors from mutual information” (Phys. Rev. A Vol. 33, pp. 1134-1140 (1986). )). Therefore, the calculation of the delay time τ will be briefly described below as a concept. In order to obtain the delay time τ for the electrocardiogram signal v (t) shown on the left side of FIG. 5, the electrocardiogram signal v (t) is separated from the signal v (t + s) and the signal v (t) separated by time s. Plot time as a parameter. In this plot, the range v _range of the electrocardiogram signal v (t) is calculated by the following formula, and the electrocardiogram signal v (t) and the signal v (t + s) separated by time s coincide with each other (○ Plot). The diagram on the right side of FIG. 5 shows a plurality of data points in a simulated manner.

The front term in the above formula means the maximum value of the signal v (t), and the rear term means the minimum value. Next, as shown in FIG. 6, a rectangular region having an area v _range × v _range in which data points are plotted is divided into L ² cells (L × L) by L cells in the column direction and L cells in the row direction. Then, the probability p _{S, L} (l, m) that a data point enters each cell (l, m) is calculated. In the diagram on the right side of FIG. 6, the shade of the cell color indicates the probability. The lighter the color, the lower the probability. The delay time τ is calculated from the following equation.

However, the mutual information I (s, L) is obtained by the following equation.

Σ _i p _{S, L} (l, i) is a marginal probability density obtained by fixing the column and summing the probabilities of the cells on the row side, and Σ _j p _{S, L} (j, m) is a fixed row. The marginal probability density is the total probability of the cells on the column side. The equation for calculating the delay time is an equation which means that the median value for the delay time in the space division number L (2, 3,..., 12) is determined as the actually used delay time τ. The reason for performing such processing is that the calculated delay time value depends on the number L of space divisions. By increasing L and making the division finer, it becomes possible to estimate the joint probability density with higher accuracy. However, since the number of data is actually finite, if L is too large, the number of data points entering each cell becomes extremely small, and the estimation of the joint probability density becomes inappropriate. On the other hand, when L is small and the division is coarse, sufficient data is secured in each cell, but the spatial resolution of the joint probability density becomes coarse, and a mutual information amount different from the original is estimated. Occurs. In this embodiment, the delay time is calculated for each of 2, 3,..., 12, and the median value is used to avoid an extreme estimation result, and stable delay time is estimated.

The Lyapunov dimension calculation unit 6 in FIG. 1 uses the Lyapunov exponents λ ₁ , λ ₂ , λ ₃ ... Λ _n from the reconstructed attractor obtained by reconstructing the electrocardiogram signal into the n-dimensional delay coordinate system using the delay time τ. Lyapunov dimension D _L is calculated from the calculated _n Lyapunov exponents λ ₁ , λ ₂ , λ ₃ . For calculating the Lyapunov index, the “Sano-Sawada method” is well known. Regarding the “Sano-Sawada method”, M. Sano and Y. Sawada wrote a paper entitled “Measurement of the Lyapunov Spectrum from a Chaotic Time Series” [Phys. Rev. Lett. 55, 1082-1085 (1985)]. Are explained in detail. The procedure for obtaining the Lyapunov dimension by the “Sano-Sawada method” is as shown in FIG. As shown in FIG. 7, the electrocardiogram signal is reconstructed into an n-dimensional delay coordinate system using the delay time τ to obtain reconstructed attractors f (t), f (t + τ), and f (t + 2τ). However, FIG. 7 is a schematic diagram in the case of n = 3. Next, a microsphere containing other data points is set for a certain data point, and the instability of the orbit is determined by calculating how the hypersphere containing these data expands and contracts with respect to time evolution. index λ _1, λ _2, quantified as λ ₃ ·· λ _n. The Lyapunov exponent is the logarithm of the expansion / contraction magnification. Then, Lyapunov dimension D _L is calculated for the obtained Lyapunov exponent according to the following equation.

In the above formula, k is the maximum value of k at which the numerator of the subsequent fractional expression does not become negative. In the example of FIG. 7, when the Lyapunov exponent is added up to λ ₃ , the numerator becomes a negative value, so k = 2. The Lyapunov dimension is considered to represent the degree of freedom of ECG signal dynamics estimated from the instability of ECG signals. By the way, in the case of an electrocardiographic signal, it has been confirmed by experiments that “n-dimensional” is preferably “four-dimensional”.

FIG. 8 shows an example of the configuration of the sensitivity state determination unit 7 shown in FIG. The sensitivity state determination unit 7 used in the present embodiment includes a pattern classification unit 71, a classification plane storage unit 72, and a state display unit 73. In the classification surface storage unit 72, a three-dimensional feature vector (average value Tmean of pulsation interval, delay time τ, and Lyapunov dimension D _L ) that is a feature amount when a plurality of subjects feel a plurality of types of emotional states in advance. ) Is identified and stored in advance as a classification plane from which one or more sensitivity states (for example, rest, comfort, discomfort, etc.) can be separated from the data obtained as a feature quantity. FIG. 9 is a diagram conceptually illustrating an example of a classification surface in a resting state. In the example shown in FIG. 9, a separation plane in a resting state is defined as a virtual plane that can separate a feature vector group at rest from a feature vector group at comfort and a feature vector group at comfort, and a classification plane memory is stored. Store in the unit 72. Similarly, a separation surface in a pleasant state and a separation surface in an unpleasant state are determined in advance and stored in the classification surface storage unit 72. The pattern classification unit 71 determines the sensitivity state of the measurement subject based on the three-dimensional feature vector (average value Tmean of pulsation interval, delay time τ and Lyapunov dimension D _L ) obtained from the measurement subject and the classification plane. . If such a classification plane is determined in advance, the sensitivity state can be determined easily and with high accuracy by pattern classification. Then, the status display unit 73 displays the determination result on the monitor screen.

As in the present embodiment, when the sensitivity state is determined using the three-dimensional feature vector (average value Tmean of pulsation interval, delay time τ, and Lyapunov dimension D _L ) as the feature amount, the sensitivity state of the measurement subject is fairly high. Can be determined. The sensitivity state determination unit 7 can also be configured to determine the sensitivity state of the person to be measured using the average value Tmean of the pulsation interval and the delay time τ (two-dimensional feature vector) as feature amounts.

The heartbeat detection unit 3, the beat interval average value calculation unit 4, the delay time calculation unit 5, the Lyapunov dimension calculation unit 6, and the emotional state determination unit 7 according to the present embodiment can be realized using a computer. FIG. 10 shows an algorithm of a computer program installed in the computer when the emotional state is determined using the computer using the average value Tmean of the pulsation interval and the delay time τ (two-dimensional feature vector) as the feature amount. It is a flowchart. This algorithm includes steps ST1 to ST3, and these steps cause the computer to realize a beat interval average value calculation function, a delay time calculation function, and a sensitivity state determination function. Further, FIG. 11 is installed in the computer when the emotional state is determined using the computer using the average value Tmean of the pulsation interval, the delay time τ, and the Lyapunov dimension D _L (three-dimensional feature vector) as the feature amount. It is a flowchart which shows the algorithm of a computer program. This algorithm includes steps ST11 to ST13, and these steps cause the computer to realize a beat interval average value calculation function, a delay time calculation function, and a sensitivity state determination function. In practice, these computer programs are stored in a computer-readable storage medium.

[Experiment]
In the present invention, an experiment for confirming that it is effective to use the two-dimensional feature vector and the three-dimensional feature vector as the feature amount of the emotional state will be described below.

 In the experiment, three sets of sensitivity recall tasks were performed for 11 subjects, each reminding them of rest, pleasantness, and discomfort. FIG. 12 shows the protocol for carrying out the experiment. Sensitivity recall teaches the subject in advance that the subject performs each sensitivity recall in accordance with the instructions of the measurer, and the measurer asks the subject to “rest,” “imagine pleasure,” “discomfort” It was carried out by instructing "Please image". As for the recall contents, each subject was taught to recall the contents that they felt "rest", "pleasant", and "unpleasant" in the past experience. Since human sensibility includes subjectivity, it is possible to control the sensibility state during the experiment based on each subject's criteria by reminding subjects that they feel pleasant and uncomfortable based on past experiences. It becomes possible. This experiment is a sufficiently valid experiment system for the purpose of estimating an individual's emotional state as a problem of the present embodiment.

FIG. 13 shows a case where a sensitivity state (especially separation of a resting state) is determined using only the average value Tmean of the beat interval as a feature vector, and a case where a sensitivity state (particularly separation of a resting state) is determined using only the delay time τ as a feature vector. When the sensitivity state (especially separation of the resting state) is determined using only the Lyapunov dimension D _L as the feature vector, and the sensitivity state (particularly separation of the resting state) is determined using the delay time τ and the Lyapunov dimension D _L as the feature vector. and if it is determined, and when determining the emotional state of the average value Tmean and Lyapunov dimension D _L of the heartbeat interval (RR interval) as a feature vector (in particular the resting state separation), the average value of the pulsation interval (RR interval) When Tmean and delay time τ are used as a two-dimensional feature vector to determine a sensibility state (particularly separation of a resting state), an average value Tmean, delay time τ and rear of a pulsation interval (RR interval) It shows a classification rate results when determining the emotional state (especially the resting state separation) the Nof dimension D _L as a three-dimensional feature vector. Here, the “classification rate” is obtained by multiplying the value obtained by dividing the number of classification success data by the number of classification target data by 100. A classification rate of 100% means that the sensitivity state has been correctly determined 100%. From FIG. 13, the determination result (correct answer) with a classification rate of 100% is obtained at 80% (correct answer rate) or higher because the average value Tmean and delay time τ of the pulsation interval (RR interval) are two-dimensional features. and when determining the emotional state (especially the resting state separation) as a vector, beat interval (RR interval) mean value Tmean, the emotional state (especially resting state delay time τ and the Lyapunov dimension D _L as a three-dimensional feature vector It can be seen that this is the case where separation is determined. In other cases, it can be seen that the accuracy rate is quite bad.

 FIG. 14 shows the test results when the subject 4 determines the sensitivity state (particularly separation of the resting state) using only the average value Tmean of the pulsation interval as a feature vector. The threshold settable line shown in FIG. 14 corresponds to the classification plane.

 FIG. 15 shows a test result in the case where the subject 3 determines the sensitivity state (particularly separation of the resting state) using the average value Tmean of the pulsation interval and the delay time τ as the feature vector. A straight line shown in FIG. 15 corresponds to a classification plane. FIG. 16 shows the test results when the subject 5 determines the sensitivity state (particularly the separation of the resting state) using the average value Tmean of the pulsation interval and the delay time τ as the feature vector. A straight line shown in FIG. 16 corresponds to a classification plane. FIG. 17 shows a test result when the sensibility state (particularly separation of the resting state) is determined for the subject 7 using the average value Tmean of the pulsation interval and the delay time τ as the feature vectors. A straight line shown in FIG. 17 corresponds to a classification plane. 15 to 17, it can be seen that there are a plurality of classification plane patterns.

Figure 18 shows the test results when determining the average value Tmean pulsation intervals for subjects 3, emotional state of the delay time τ and Lyapunov dimension D _L as a feature vector (in particular the resting state separation). The grid shown by the solid line in FIG. 17 corresponds to the classification plane. Figure 19 shows the average value Tmean pulsation intervals for subjects 5, the test results obtained by determining the emotional state (especially the resting state separation) the delay time τ and Lyapunov dimension D _L as a feature vector. The grid shown by the solid line in FIG. 19 corresponds to the classification plane. Figure 20 shows the average value Tmean pulsation intervals for subjects 7, the test results obtained by determining the emotional state (especially the resting state separation) the delay time τ and Lyapunov dimension D _L as a feature vector. The grid shown by the solid line in FIG. 20 corresponds to the classification plane. 18 to 20, it can be seen that there are a plurality of classification plane patterns.

 In the above experiment, the resting state is determined, but according to the present invention, it is naturally possible to determine the pleasant state or the unpleasant state. Of course, other sensitivity states such as “surprise”, “anger”, “tension”, “fear”, and “anxiety” may be determined as the sensitivity state determined in the present invention.

 In the above embodiment, the sensitivity state is determined using the pattern recognition technology that classifies data based on a plurality of classification planes specified in advance by the sensitivity state determination unit, but other recognition technologies may be used. Of course.

 According to the present invention, when an electrocardiogram signal is used, the average value and delay time of the pulsation interval are used as feature quantities, or the average value, delay time and Lyapunov dimension of the pulsation intervals are used as feature quantities. By determining the sensitivity state, it is possible to provide a sensitivity state determination device, a sensitivity state determination method, and a computer program for sensitivity state determination that have higher determination accuracy than conventional methods.

DESCRIPTION OF SYMBOLS 1 Person to be measured 2 Biological amplifier 3 Heartbeat detection part 4 Beat interval average value calculation part 5 Delay time calculation part 6 Lyapunov dimension calculation part 7 Kansei state determination part 71 Pattern classification part 72 Classification surface memory | storage part 73 State display part

Claims (10)

A sensitivity state determination apparatus for determining a sensitivity state of a measurement subject using an electrocardiogram signal,
A pulsation interval average value calculating unit for obtaining an average value of pulsation intervals from an electrocardiogram signal including a plurality of R waves obtained from the measurement subject;
A delay time calculation unit that calculates a delay time as a parameter having a dimension of time required when performing attractor reconstruction in order to obtain a Lyapunov exponent from the electrocardiogram signal;
An emotional state determination device including an emotional state determination unit that determines an emotional state of the measurement subject using the average value of the beat interval and the delay time as feature amounts. A sensitivity state determination apparatus for determining a sensitivity state of a measurement subject using an electrocardiogram signal,
A pulsation interval average value calculating unit for obtaining an average value of pulsation intervals from an electrocardiogram signal including a plurality of R waves obtained from the measurement subject;
A delay time calculation unit that calculates a delay time as a parameter having a dimension of time required when performing attractor reconstruction in order to obtain a Lyapunov exponent from the electrocardiogram signal;
A Lyapunov dimension calculation unit for obtaining a Lyapunov dimension of the electrocardiogram signal using the delay time;
An emotional state determination apparatus comprising: an emotional state determination unit that determines the emotional state of the measurement subject using the average value of the beat interval, the delay time, and the Lyapunov dimension as feature amounts.   The sensation state determination device according to claim 1 or 2, wherein the pulsation interval average value calculation unit calculates an average value of intervals between maximum amplitudes of the plurality of electrocardiogram signals as an average value of the pulsation intervals.   The delay time calculation unit is a time when the mutual information amount calculated based on the joint probability density of the electrocardiogram signal v (t) at time t and the electrocardiogram signal v (t + τ) delayed by time τ from time t is minimized. The sensitivity state determination apparatus according to claim 1, wherein τ is calculated as the delay time.   The Lyapunov dimension calculation unit calculates a Lyapunov exponent from a reconstructed attractor obtained by reconfiguring the electrocardiogram signal into the n-dimensional delay coordinate system using the delay time, and calculates the Lyapunov exponent from the calculated n Lyapunov exponents. The sensitivity state determination apparatus according to claim 2, wherein a dimension is calculated.   The sensitivity state determination unit includes a pattern classification classification plane that can separate one or more emotional states from data obtained by measuring the feature amounts when a plurality of subjects feel a plurality of types of emotional states in advance. The sensitivity state determination apparatus according to claim 1 or 2, wherein the sensitivity state determination device according to claim 1 or 2, wherein the sensitivity state of the measurement subject is determined based on the feature amount obtained from the measurement subject and the classification plane. A sensitivity state determination method for determining a sensory state of a measured person using a computer,
A pulsation interval average value calculating function for obtaining an average value of pulsation intervals from an electrocardiographic signal including a plurality of R waves obtained from the measurement subject;
A delay time calculation function for calculating a delay time as a parameter having a dimension of time required when performing an attractor reconstruction in order to obtain a Lyapunov exponent from the electrocardiogram signal;
Using the average value of the pulsation interval and the delay time as feature quantities, the sensitivity state determination function for determining the sensitivity state of the measurement subject is realized, thereby determining the sensitivity state of the measurement subject. Sensitivity state determination method. A sensitivity state determination method for determining a sensory state of a measured person using a computer,
A pulsation interval average value calculating function for obtaining an average value of pulsation intervals from an electrocardiographic signal including a plurality of R waves obtained from the measurement subject;
A delay time calculation function for calculating a delay time as a parameter having a dimension of time required when performing an attractor reconstruction in order to obtain a Lyapunov exponent from the electrocardiogram signal;
A Lyapunov dimension calculation function for obtaining a Lyapunov dimension of the electrocardiogram signal using the delay time;
Using the average value of the pulsation interval, the delay time, and the Lyapunov dimension as a feature amount, a sensitivity state determination function for determining the sensitivity state of the measurement subject is realized, thereby determining the sensitivity state of the measurement subject. A method for determining a sensitivity state. A computer program for determining a sensitivity state installed in the computer used in the sensitivity state determination method for determining a sensitivity state of a measurement subject using a computer,
A pulsation interval average value calculating function for obtaining an average value of pulsation intervals from an electrocardiogram signal including a plurality of R waves obtained from the measurement subject;
A delay time calculation function for calculating a delay time as a parameter having a dimension of time required when performing an attractor reconstruction in order to obtain a Lyapunov exponent from the electrocardiogram signal;
A computer program for determining a sensibility state, which realizes a sensibility state determination function for determining a sensibility state of the measurement subject by a processor in the computer, using the average value of the beat interval and the delay time as feature amounts. A computer program for determining a sensitivity state installed in the computer used in the sensitivity state determination method for determining a sensitivity state of a measurement subject using a computer,
A pulsation interval average value calculating function for obtaining an average value of pulsation intervals from an electrocardiogram signal including a plurality of R waves obtained from the measurement subject;
A delay time calculation function for calculating a delay time as a parameter having a dimension of time required when performing an attractor reconstruction in order to obtain a Lyapunov exponent from the electrocardiogram signal;
A Lyapunov dimension calculation function for obtaining a Lyapunov dimension of the electrocardiogram signal using the delay time;
A computer program for determining a sensory state, wherein a processor in the computer realizes a sensory state determination function for determining the sensory state of the measurement subject using the average value of the beat interval, the delay time, and the Lyapunov dimension as feature amounts.