JP2018057449A - Drowsiness estimation device and drowsiness estimation program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable drowsiness to be estimated appropriately by following an activity index that changes with time in consideration of a lapse of time based on human physiology.SOLUTION: A drowsiness estimation device includes RRI acquisition means 21 for acquiring RRI data, which is data on an R-R interval, index value calculation means 23 for subjecting the RRI data to statistical processing, and calculating an index value on an activity index concerning an activity of autonomic nerves, drowsiness estimation means 24 for evaluating the index value calculated by the index value calculation means and estimating drowsiness based on a drowsiness estimation rule configured including an estimation function for evaluation by a threshold for each activity index, and threshold function calculation means 26 that, based on index value time series data in which the index values are arranged in a time-series manner, acquires an approximate function close to a change in the index value time series data, obtains a threshold function by giving a change of a predetermined value to the approximate function, and uses the threshold function as the threshold.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a drowsiness estimation device and a drowsiness estimation program.

  The inventors of the present application filed a sleepiness estimation apparatus including an RRI acquisition means, an index value calculation means, and a sleepiness estimation means. The present invention provides RRI acquisition means for acquiring RRI data, which is R-R interval data, from a signal obtained by an RRI sensor that detects a signal corresponding to an R wave of an electrocardiogram signal, a result of statistical processing of the RRI data, and And / or index value calculation means for calculating index values for a plurality of types of activity indexes related to autonomic nerve activity based on the result of spectral analysis of the RRI data, and an estimation function that is evaluated by a threshold value for each activity index. And drowsiness estimation means for estimating the drowsiness by evaluating the index value calculated by the index value calculation means (see Patent Document 1).

  By the way, it is known that the heart rate tends to decrease with time as compared to the start of work. In addition, activity indices related to parasympathetic nerves such as the activity index used in the above invention, such as SDSD (RMSSD), pNN50, etc. tend to increase with time.

  However, the threshold value used in the sleepiness estimation rule of the above invention is the average value, minimum value, or maximum value of the activity index. That is, the threshold value is a constant value regardless of the passage of time. Therefore, while the threshold value is a constant value, the activity index to be evaluated using this threshold value varies with the passage of time. There was a problem that it was impossible to estimate the sleepiness.

  On the other hand, in Patent Document 2, in the dozing determination apparatus, the threshold for blink detection is changed according to the illuminance, which leads to appropriate determination. However, it has not been able to follow fluctuations over time.

  Further, Patent Document 3 discloses a snooze driving prevention device including an alarm output unit that outputs an alarm when the drowsiness level is higher than a predetermined alarm threshold. The invention of Patent Document 3 includes an alarm threshold correction unit that corrects an alarm threshold, but performs correction based on the number of times the sleepiness level exceeds the threshold, the total time that the sleepiness level exceeds the threshold, and the like. It does not take into account the passage of time based on human physiology.

JP 2006-120061 A JP-A-8-153288 Japanese Patent Laying-Open No. 2015-219771

  The present invention has been made in view of the present state of sleepiness estimation technology as described above, and its purpose is to allow time lapse based on human physiology and to appropriately estimate sleepiness by following a time-varying activity index. An object of the present invention is to provide a sleepiness estimation apparatus and a sleepiness estimation program.

  An apparatus for estimating sleepiness according to the present invention includes an RRI acquisition unit that acquires RRI data that is data of an RR interval from a signal obtained by an RRI sensor that detects a signal corresponding to an R wave of an electrocardiogram signal; Index value calculation means for calculating an index value for an activity index related to autonomic nerve activity based on a result of statistical processing and / or a spectrum analysis result of the RRI data, and an estimation function evaluated by a threshold value for each activity index A drowsiness estimation unit that evaluates the index value calculated by the index value calculation unit and estimates drowsiness based on a drowsiness estimation rule that includes the index value time-series data in which the index values are arranged in time series Then, an approximate function approximating the change in the index value time series data is obtained, and a threshold function that gives a predetermined value change to this approximate function is obtained. Characterized in that the threshold function equipped with the threshold value function calculating means and the threshold value.

  In the sleepiness estimation apparatus according to the present invention, the threshold function calculation means obtains the threshold function for each day for the index value time-series data for a plurality of days obtained in the past, and statistically processes the threshold function. To obtain a final threshold function.

  In the sleepiness estimation apparatus according to the present invention, the threshold function calculation means obtains one threshold function for each time period in which the fluctuation rate of the index value in a day is substantially the same.

  In the sleepiness estimation apparatus according to the present invention, the threshold function calculation means obtains one threshold function for each of two morning and afternoon time zones before and after lunch break.

  In the sleepiness estimation apparatus according to the present invention, the threshold function calculation means obtains one threshold function for each time period until a break.

  In the sleepiness estimation apparatus according to the present invention, the threshold function calculation means obtains a threshold function by giving a change of a predetermined value to the approximate function based on the difference between the average value of the index value time-series data and the previous threshold. It is characterized by that.

In the sleepiness estimation apparatus according to the present invention, as an activity index as a result of statistical processing of the RRI data,
SDNN: (RRI standard deviation)
RMSSD: (the square root of the root mean square of the difference between adjacent RRIs)
SDSD: (standard deviation of the difference between adjacent RRIs)
pNN50: (Ratio in which the difference between adjacent RRIs exceeds 50 (milliseconds))
At least one of the above is used.

In the sleepiness estimation apparatus according to the present invention, as an activity index as a result of spectrum analysis of the RRI data,
LF: (PSD (power spectral density function) 0.04 to 0.15 [Hz] power)
HF: (PSD 0.15-0.40 [Hz] power)
HF / (LF + HF)
pi (i = 0, 1, 2,..., 9): (PSD 0.15 + i × 0.025 to 0.15+ (i + 1) × 0.025 [Hz] power)
At least one of the above is used.

  The drowsiness estimation program according to the present invention is an RRI acquisition unit that acquires RRI data, which is R-R interval data, from a signal obtained by an RRI sensor that detects a signal corresponding to an R wave of an electrocardiogram signal. Index value calculation means for calculating an index value for an activity index related to autonomic nerve activity based on a result of statistical processing of data and / or a result of spectrum analysis of the RRI data, and an estimation function evaluated by a threshold value for each activity index Drowsiness estimation means for evaluating the index value calculated by the index value calculation means and estimating sleepiness based on a drowsiness estimation rule comprising: based on index value time-series data in which the index values are arranged in time series Obtain an approximate function that approximates the change in the index value time series data, and apply a threshold value that gives a predetermined value change to this approximate function. Seeking function, characterized in that to function this threshold function as a threshold function calculating means and the threshold value.

  In the sleepiness estimation program according to the present invention, the threshold function calculation means obtains the threshold function for each day for the index value time-series data for a plurality of days obtained in the past, and statistically processes the threshold function. And functioning to obtain a final threshold function.

  The sleepiness estimation program according to the present invention is characterized in that the threshold function calculation means functions so as to obtain one threshold function for each time period in which the fluctuation rate of the index value in a day is substantially the same.

  The sleepiness estimation program according to the present invention is characterized in that the threshold function calculation means functions so as to obtain one threshold function for each of two morning and afternoon time zones before and after lunch break.

  The sleepiness estimation program according to the present invention is characterized in that the threshold function calculation means functions so as to obtain one threshold function for each time period until a break.

  In the sleepiness estimation program according to the present invention, the threshold function calculation means obtains a threshold function by giving a change of a predetermined value to the approximate function based on the difference between the average value of the index value time-series data and the previous threshold. It is made to function as follows.

In the sleepiness estimation program according to the present invention, as an activity index as a result of statistical processing of the RRI data,
SDNN: (RRI standard deviation)
RMSSD: (the square root of the root mean square of the difference between adjacent RRIs)
SDSD: (standard deviation of the difference between adjacent RRIs)
pNN50: (Ratio in which the difference between adjacent RRIs exceeds 50 (milliseconds))
At least one of the above is used.

In the sleepiness estimation program according to the present invention, as an activity index as a result of spectrum analysis of the RRI data,
LF: (PSD (power spectral density function) 0.04 to 0.15 [Hz] power)
HF: (PSD 0.15-0.40 [Hz] power)
HF / (LF + HF)
pi (i = 0, 1, 2,..., 9): (PSD 0.15 + i × 0.025 to 0.15+ (i + 1) × 0.025 [Hz] power)
At least one of the above is used.

  According to the present invention, based on index value time-series data in which index values are arranged in time series, an approximate function that approximates the change in the index value time-series data is obtained, and a change in a predetermined value is performed with respect to the approximate function. Since a threshold function calculation means that obtains a given threshold function and uses this threshold function as the threshold is provided, it is possible to appropriately estimate sleepiness by following a time-varying activity index in consideration of the passage of time based on human physiology. To do.

The block diagram which shows embodiment of the drowsiness estimation apparatus which concerns on this invention. The flowchart which shows the operation mode of embodiment of the sleepiness estimation apparatus which concerns on this invention. The flowchart which shows operation | movement of the threshold value function production | generation mode of embodiment of the sleepiness estimation apparatus which concerns on this invention. The figure which shows the average value and standard deviation which are calculated in the threshold value function production | generation mode of embodiment of the sleepiness estimation apparatus which concerns on this invention. The figure which shows an example of the approximate function calculated | required in the threshold value function production | generation mode of embodiment of the sleepiness estimation apparatus which concerns on this invention. The figure which shows an example of the threshold value function calculated | required in the threshold value function production | generation mode of embodiment of the sleepiness estimation apparatus which concerns on this invention.

  Embodiments of a drowsiness estimation device and a drowsiness estimation program according to the present invention will be described below with reference to the accompanying drawings. In the drawings, the same components are denoted by the same reference numerals and redundant description is omitted. FIG. 1 shows a block diagram of an embodiment of a sleepiness estimation apparatus according to the present invention. In the present embodiment, a heart rate sensor can be used as the RRI sensor 10 that detects a signal corresponding to the R wave of the electrocardiogram signal.

  In addition to the heart rate sensor, the RRI sensor 10 may use a configuration of a part for taking out an electrocardiogram signal of an electrocardiograph or a pulse wave sensor. The RRI sensor 10 is provided in a living body, detects an electrocardiogram signal wirelessly or by wire, and outputs an RRI (before shaping) 1001.

  As shown in FIG. 1, in the sleepiness estimation apparatus of this embodiment, the terminal 20 includes an RRI acquisition unit 21, a data shaping unit 22, an index value calculation unit 23, a sleepiness estimation unit 24, and an output unit 25. The functions and operations of these means are described in detail in Patent Document 1. The sleepiness estimation apparatus according to the present embodiment further includes threshold function calculation means 26.

  The RRI acquisition unit 21 acquires RRI data transmitted by the RRI sensor 10. The RRI sensor 10 may output an electrocardiogram signal. In this case, the RRI acquisition unit 21 creates RRI data based on the electrocardiogram signal. The data shaping means 22 has a configuration for performing at least one of trend removal, abnormal value removal, data interpolation, and filter processing, which is removal of a predetermined frequency component. Details of these configurations are as described in Patent Document 1.

  The index value calculation means 23 calculates an index value of the autonomic nerve based on the RRI data. For example, the index value calculation means 23 can calculate a plurality of values based on the result of statistical processing of the RRI data and / or the result of spectrum analysis of the RRI data. It can be configured to calculate the index value of the autonomic nerve activity index as a type. The index value calculation means 23 calculates an index value for unit time using the RRI data for each first time, collects the index values for unit time for each type of activity index, and vectorizes the index values. Create an index value vector time series in a series.

  The sleepiness estimation means 24 estimates sleepiness by evaluating the activity index calculated by the index value calculation means 23 on the basis of a sleepiness estimation rule constituted by an estimation function that is evaluated based on a threshold value related to an activity index and / or a fluctuation state. To do.

  The output unit 25 outputs the estimation result obtained by the sleepiness estimation unit 24 and is used for generating an alarm, stopping the operation of the device, warning, or the like. The terminal 20 can be configured by a smartphone, a tablet terminal, a mobile terminal, or the like.

  A cloud storage 30 is connected to the terminal 20. The cloud storage 30 includes sensor characteristic information 1101 and sleepiness estimation rules 1102 in advance. The sensor characteristic information 1101 is used when the data shaping unit 22 executes data shaping. The sleepiness estimation rule 1102 is used when the sleepiness estimation means 24 estimates sleepiness. The cloud storage 30 indicates that the acquired RRI data acquired in the terminal 20 may be stored as the acquired RRI 1012X of history data. This acquired RRI 1012X is used for updating the sleepiness estimation rule 1102 You may use for.

  The terminal 20 performs processing under the control of the CPU, and each means is realized by the CPU executing a program. The threshold function calculation means 26 obtains an approximate function that approximates the change of the index value time-series data based on the index value time-series data in which the index values are arranged in time series, and changes the predetermined value with respect to the approximate function. Is obtained, and this threshold function is used as the threshold value.

  The sleepiness estimation apparatus according to the present embodiment configured as described above operates in the standby operation mode S11, the threshold function generation mode S12, and the practical mode S13 as shown in FIG. The preliminary operation mode S11 is a mode in which the processing shown in Patent Document 1 is performed, and is a premise of the present invention. Since the operation of the standby operation mode S11 is described in detail in Patent Document 1, the description thereof is omitted.

  In the threshold function generation mode S12, as shown in FIG. 3, the index value calculation means 23 calculates an index value for an activity index related to autonomic nerve activity using the acquired RRI data of the acquired RRI 1012X (S21). For example, an index value is calculated for an activity index using acquired RRI data for n days. If this calculation is performed in advance and stored in the cloud storage 30, this process is omitted. Further, when a threshold value is not used for the evaluation of the index value, this process is not performed for the index value.

Next, the average value μ _i and standard deviation σ _i of the i (1 ≦ i ≦ n) -th index value are calculated (S22). Next, since the threshold value for this index value is determined (used in the standby operation mode S11), r _i is calculated assuming that the distance from the average value μ _i to the threshold value K is the distance r _i σ _i ( S23). A diagram showing indices corresponding to the processes of steps S22 and S23 is shown in FIG. Here, the description uses SDSD as an index. Of course, other indicators may be used.

Next, an approximate function f _i (t) is obtained for the i-th index value (S24). An appropriate method such as a least square method can be employed for the process of obtaining the approximate function. FIG. 5 shows an example in which a linear function is adopted as the approximate function.

Next, the value of the approximation function f _i (t) is subtracted from the i-th index value to obtain the standard deviation σA _i of the obtained data (S25). A multiplication value r _i .SIGMA.A _i of .SIGMA.A _i determined in r _i and the step S25 determined in the step S23, in addition to the approximation function f _i (t), determined as the i-th threshold function h _i (t) (S26).

The processing result of step S26 is shown in FIG. Here, if f _i (t) is f _i (t) = a _i t + b _i , the threshold function h _i (t) is
h _i (t) = a _i t + b _i + r _i σA _i
It becomes. In addition, although the example of the linear function was shown here, it is not limited to this, It becomes an appropriate function by approximation.

Steps S22 to S26 are repeated n times. Finally, threshold functions h ₁ (t) to h _n (t) are obtained for the first to n-th indices, and these averages are obtained as a final threshold function H (t ) Is obtained (S27). That is, it is the following formula 1.

Through the above processing, all threshold functions for the necessary types of indexes are obtained, and in the practical mode S13 in FIG. 2, the threshold used in the preliminary operation mode S11 is replaced with the threshold function obtained in the threshold function generation mode. Processing similar to that in mode S11 is performed.

  In this embodiment, the index value calculation means 23 calculates an index value for an activity index related to autonomic nerve activity based on the result of statistical processing of RRI data and the result of spectrum analysis of RRI data. An index value may be calculated for an activity index related to autonomic nerve activity based on a result of statistical processing or a result of spectrum analysis of RRI data.

  Further, the sleepiness estimation rule 1102 is configured by an estimation function that is evaluated based on a threshold value related to each activity index and / or a variation state. It ’s fine.

The activity index used in this embodiment is an activity index as a result of statistical processing of RRI data.
SDNN: (RRI standard deviation)
RMSSD: (the square root of the root mean square of the difference between adjacent RRIs)
SDSD: (standard deviation of the difference between adjacent RRIs)
pNN50: (Ratio in which the difference between adjacent RRIs exceeds 50 (milliseconds))
At least one of The details of this activity index are described in Patent Document 1, so the description thereof is omitted.

The activity index used in the present embodiment is an activity index as a result of spectrum analysis of the RRI data.
LF: (PSD (power spectral density function) 0.04 to 0.15 [Hz] power)
HF: (PSD 0.15-0.40 [Hz] power)
HF / (LF + HF)
pi (i = 0, 1, 2,..., 9): (PSD 0.15 + i × 0.025 to 0.15+ (i + 1) × 0.025 [Hz] power)
At least one of The details of this activity index are described in Patent Document 1, so the description thereof is omitted.

  Furthermore, it has been confirmed that the time change of each index becomes different depending on the break. For this reason, it is preferable to function so as to obtain one threshold value function for each time zone in which the fluctuation rate of the index value in one day is substantially the same. In this case, a time zone break may be obtained by obtaining a variation rate of the index value.

  Further, it can be made to function so as to obtain one threshold function for each of two time zones of morning and afternoon before and after the lunch break. Or it can also be made to function so that one threshold value function may be obtained for every time zone until a break. For example, in the case of a work that continues for several hours from the start, breaks, then performs several hours, breaks, and so on, it is preferable to obtain one threshold function for each time period until the break. is there.

  Furthermore, since the above temporal variation is noticeable in desk work or driving sitting work, and so-called manual labor or exercise hardly appears, this embodiment is in desk work or driving sitting state. It is preferable to apply to the person of the business of.

  1 is merely an example, and the terminal 20 may include what the cloud storage 30 holds. Further, a server may be provided in place of the cloud storage 30, and the cloud storage 30 may be substituted. Furthermore, a server may be provided in place of the cloud storage 30, and the cloud storage 30 may be substituted, and some of the functions of the terminal 20 may be performed in the server. In addition, the system configuration can be changed as appropriate.

DESCRIPTION OF SYMBOLS 10 Sensor 20 Terminal 21 Acquisition means 22 Data shaping means 23 Index value calculation means 24 Sleepiness estimation means 25 Output means 26 Threshold function calculation means 30 Cloud storage

Claims (16)

RRI acquisition means for acquiring RRI data, which is R-R interval data, from a signal obtained by an RRI sensor that detects a signal corresponding to an R wave of an electrocardiogram signal;
Index value calculation means for calculating an index value for an activity index related to autonomic nerve activity based on a result of statistical processing of the RRI data and / or a result of spectrum analysis of the RRI data;
Sleepiness estimation means for evaluating the index value calculated by the index value calculation means and estimating sleepiness based on a sleepiness estimation rule including an estimation function that is evaluated by a threshold value for each activity index;
Based on the index value time-series data in which the index values are arranged in time series, an approximate function that approximates the change in the index value time-series data is obtained, and a threshold function that gives a change in a predetermined value to the approximate function is obtained. A drowsiness estimation device comprising: a threshold value function calculating unit that obtains the threshold value function as the threshold value.   The threshold function calculation means obtains the threshold function for each day of index value time series data for a plurality of days obtained in the past, and obtains a final threshold function by statistically processing the threshold function. The sleepiness estimation apparatus according to claim 1, wherein:   The sleepiness estimation apparatus according to claim 1, wherein the threshold function calculation unit obtains one threshold function for each time period in which the fluctuation rate of the index value in a day is substantially the same.   The sleepiness estimation apparatus according to claim 1 or 2, wherein the threshold function calculation means obtains one threshold function for each of two time zones of morning and afternoon before and after lunch break.   The sleepiness estimation apparatus according to claim 1, wherein the threshold function calculation unit obtains one threshold function for each time period until a break.   The threshold function calculating means obtains a threshold function by changing a predetermined value to an approximate function based on a difference between an average value of index value time-series data and a previous threshold. The drowsiness estimation apparatus according to any one of 5. As an activity index as a result of statistical processing of the RRI data,
SDNN: (RRI standard deviation)
RMSSD: (the square root of the root mean square of the difference between adjacent RRIs)
SDSD: (standard deviation of the difference between adjacent RRIs)
pNN50: (Ratio in which the difference between adjacent RRIs exceeds 50 (milliseconds))
The sleepiness estimation apparatus according to claim 1, wherein at least one of the following is used. As an activity index as a result of spectrum analysis of the RRI data,
LF: (PSD (power spectral density function) 0.04 to 0.15 [Hz] power)
HF: (PSD 0.15-0.40 [Hz] power)
HF / (LF + HF)
pi (i = 0, 1, 2,..., 9): (PSD 0.15 + i × 0.025 to 0.15+ (i + 1) × 0.025 [Hz] power)
The sleepiness estimation apparatus according to any one of claims 1 to 7, wherein at least one of the following is used. Computer
RRI acquisition means for acquiring RRI data, which is R-R interval data, from a signal obtained by an RRI sensor that detects a signal corresponding to an R wave of an electrocardiogram signal;
Index value calculation means for calculating an index value for an activity index related to autonomic nerve activity based on a result of statistical processing of the RRI data and / or a result of spectrum analysis of the RRI data;
Sleepiness estimation means for evaluating the index value calculated by the index value calculation means and estimating sleepiness based on a sleepiness estimation rule configured to include an estimation function evaluated from a threshold value for each activity index;
Based on the index value time-series data in which the index values are arranged in time series, an approximate function that approximates the change in the index value time-series data is obtained, and a threshold function that gives a change in a predetermined value to the approximate function is obtained. A drowsiness estimation program that is obtained and functions as threshold function calculation means using the threshold function as the threshold.   The threshold function calculation means obtains the threshold function for each day of index value time series data for a plurality of days obtained in the past, and statistically processes the threshold function to obtain a final threshold function. The sleepiness estimation program according to claim 9, wherein the sleepiness estimation program is executed.   The sleepiness estimation program according to claim 9 or 10, wherein the threshold function calculation unit is caused to function so as to obtain one threshold function for each time period in which the fluctuation rate of the index value in a day is substantially the same.   The sleepiness estimation program according to claim 9 or 10, wherein the threshold function calculation means functions so as to obtain one threshold function for each of two morning and afternoon time zones before and after lunch break.   The sleepiness estimation program according to claim 9 or 10, wherein the threshold function calculation means functions so as to obtain one threshold function for each time period until a break.   The threshold function calculation means is made to function so as to obtain a threshold function by giving a change of a predetermined value to an approximate function based on a difference between an average value of index value time-series data and a previous threshold value. The sleepiness estimation program according to any one of claims 9 to 13. As an activity index as a result of statistical processing of the RRI data,
SDNN: (RRI standard deviation)
RMSSD: (the square root of the root mean square of the difference between adjacent RRIs)
SDSD: (standard deviation of the difference between adjacent RRIs)
pNN50: (Ratio in which the difference between adjacent RRIs exceeds 50 (milliseconds))
The sleepiness estimation program according to any one of claims 9 to 14, wherein at least one of the following is used. As an activity index as a result of spectrum analysis of the RRI data,
LF: (PSD (power spectral density function) 0.04 to 0.15 [Hz] power)
HF: (PSD 0.15-0.40 [Hz] power)
HF / (LF + HF)
pi (i = 0, 1, 2,..., 9): (PSD 0.15 + i × 0.025 to 0.15+ (i + 1) × 0.025 [Hz] power)
The sleepiness estimation program according to any one of claims 9 to 15, wherein at least one of the following is used.

Images