JP6750229B2 - Drowsiness detection program, drowsiness detection method, and drowsiness detection device - Google Patents

Description

The present invention relates to a drowsiness detection program, a drowsiness detection method, and a drowsiness detection device.

Although the total number of traffic accidents has been decreasing year by year, the number of accidents due to human error has not decreased so much. One of the causes of accidents due to human error is drowsiness during driving. For this reason, there is a demand for a technique for issuing a warning to the driver based on the awakening level during driving to prevent an accident.

For example, as a conventional technique for determining the arousal level, the driver's arousal level is determined from the change of the maximum points in the frequency range of LF (Low Frequency) and HF (High Frequency) of the spectral density calculated by frequency analysis from the driver's heartbeat signal. There is a technology to judge. In such a conventional technique, when the awakening degree is low, it is determined that the driver is in a sleepy state.

JP, 2010-63682, A JP, 2004-267409, A JP-A-2004-646 Japanese Patent Publication No. 2004-511286 JP, 2014-12042, A JP, 2010-131061, A

However, the state where the awakening degree is low includes states other than the sleepy state of the driver. Therefore, in the above-described conventional technique, the driver may not be able to accurately determine the sleepy state.

In one aspect, it is an object to provide a drowsiness detection program, a drowsiness detection method, and a drowsiness detection device that enable a subject to accurately estimate a sleepy state.

In the first proposal, the drowsiness detection program causes a computer to execute a process of performing frequency analysis on heartbeat data sequentially obtained from the subject. The drowsiness detection program causes a computer to execute a process of estimating whether the subject is drowsy or relaxed according to the degree of variation in the frequency distribution obtained as a result of the frequency analysis.

According to one embodiment of the present invention, a subject can accurately estimate a sleepy state.

FIG. 1 is a functional block diagram showing the configuration of the drowsiness detection device according to the present embodiment. FIG. 2 is a diagram showing an example of heartbeat signal data. FIG. 3 is a diagram showing an example of heartbeat interval variation data. FIG. 4 is a diagram showing the relationship between frequency and spectral density. FIG. 5 is a diagram for explaining the process of determining the arousal level. FIG. 6 is a diagram for explaining the characteristics of the frequency band. FIG. 7 is a diagram showing an example of the frequency distribution. FIG. 8A is a diagram showing an example of the frequency distribution when the physical condition of the subject is normal. FIG. 8B is a diagram showing an example of the frequency distribution when the physical condition of the subject is poor. FIG. 9 is a flowchart showing an example of the procedure of drowsiness detection processing. FIG. 10 is a diagram illustrating an example of a computer that executes a drowsiness detection program.

Hereinafter, embodiments of the drowsiness detection program, the drowsiness detection method, and the drowsiness detection device disclosed in the present application will be described in detail with reference to the drawings. The present invention is not limited to the embodiments.

FIG. 1 is a functional block diagram showing the configuration of the drowsiness detection device according to the present embodiment. As shown in FIG. 1, the drowsiness detection device 100 includes a sensor 110, a heartbeat interval calculation unit 120, an analysis unit 130, a determination unit 140, an estimation unit 150, a setting unit 160, and a warning unit 170. Have.

The sensor 110 is a sensor that detects a heartbeat signal of the subject. The test subject corresponds to, for example, a driver of the vehicle. For example, the sensor 110 is installed on the steering wheel of the vehicle or the like. In the present embodiment, as an example, a case of detecting a heartbeat signal will be described, but the sensor 110 may detect a pulse signal of the subject. The heartbeat signal and the pulse signal are examples of biological signals. The sensor 110 outputs the data of the heartbeat signal to the heartbeat interval calculation unit 120. In the following description, heartbeat signal data will be referred to as heartbeat signal data. In this embodiment, the heartbeat signal data corresponds to the heartbeat data.

FIG. 2 is a diagram showing an example of heartbeat signal data. As shown in FIG. 2, the heartbeat signal data has waveforms called P, Q, R, S, T, and U waves. For example, the horizontal axis of FIG. 2 represents time and the vertical axis represents the amplitude of the heartbeat signal.

The heartbeat interval calculation unit 120 is a processing unit that detects the amplitude peak of the heartbeat signal based on the heartbeat signal data and detects the interval of the detected timing. The interval between the timings at which the amplitude peak of the heartbeat signal is detected is referred to as the heartbeat interval. The processing of the heartbeat interval calculation unit 120 will be described with reference to FIG. As shown in FIG. 2, the heartbeat interval calculation unit 120 detects a point R at which the amplitude of the heartbeat signal is equal to or more than a threshold value, that is, an amplitude peak, and detects an interval between the detected points R as an amplitude interval. The heartbeat interval calculation unit 120 outputs the detected heartbeat interval data to the analysis unit 130. In the following description, heartbeat interval data will be referred to as heartbeat interval data.

The analysis unit 130 is a processing unit that performs frequency analysis based on heartbeat interval data. The analysis unit 130 stores the heartbeat interval data sequentially obtained from the subject and sequentially detected by the heartbeat interval calculation unit 120 for at least a certain period. Then, the analysis unit 130 calculates the spectral density corresponding to the heartbeat interval by performing the frequency analysis of the heartbeat interval data obtained during the immediately preceding constant period at predetermined time intervals.

Hereinafter, an example of the process in which the analysis unit 130 calculates the spectral density corresponding to the heartbeat interval will be described. The analysis unit 130 generates, based on the heartbeat interval data, data of heartbeat intervals that change with time. In the following description, heartbeat interval data that changes with time is referred to as heartbeat interval fluctuation data.

FIG. 3 is a diagram showing an example of heartbeat interval variation data. In FIG. 3, the vertical axis is the axis indicating the heartbeat interval, and the horizontal axis is the axis indicating time. As shown in FIG. 3, the heartbeat interval fluctuates with time.

The analysis unit 130 performs frequency analysis based on the heartbeat interval variation data to calculate the relationship between frequency and spectral density. FIG. 4 is a diagram showing the relationship between frequency and spectral density. In FIG. 4, the vertical axis is the axis showing the spectral density, and the horizontal axis is the axis showing the frequency. In the example shown in FIG. 4, the spectral density has a maximum at points 10a, 10b, 10c, and 10d. In the following description, data indicating the relationship between the spectral density and the frequency will be referred to as spectral density data.

Here, the analyzing unit 130 may calculate the spectral density by using an AR (Autoregressive) model, although any method may be used when calculating the relationship between the spectral density and the frequency. The AR model is disclosed, for example, in non-patent documents (Shunsuke Sato, Akira Yoshikawa, Tohru Kiryu, "Basics of Biosignal Processing", Corona Publishing Co., Ltd.) and the like. The AR model is a model that represents a state at a certain time point by a linear sum of past time series data, and is characterized in that a clear maximum point can be obtained even with a small number of data points as compared with the Fourier transform. The analysis unit 130 may calculate the relationship between the spectral density and the frequency by Fourier transform.

The p-th order AR model of the time series x(s) can be expressed by the equation (1) using the AR coefficient a(m) and the error term e(s), which are weights for past values. In equation (1), e(s) is ideally white noise.

をｋ次のＡＲ係数とすると、スペクトル密度Ｐ_AR（ｆ）は、式（３）によって表される。解析部１３０は、式（３）および心拍間隔変動データを基にして、スペクトル密度データを算出する。 Let p be the identification order, f _s be the sampling frequency, and ε _p be the identification error,

Is a kth-order AR coefficient, the spectral density P _AR (f) is expressed by equation (3). The analysis unit 130 calculates the spectral density data based on the equation (3) and the heartbeat interval variation data.

The determination unit 140 is a processing unit that determines the awakening degree of the subject at each time based on the analysis result by the analysis unit 130. For example, the determination unit 140 determines the arousal level based on the maximum value of the spectral density corresponding to the heartbeat interval obtained by the frequency analysis by the analysis unit 130 and the frequency corresponding to the maximum value of the spectral density.

Hereinafter, an example of the process in which the determination unit 140 determines the awakening degree will be described. First, the process in which the determination unit 140 determines the arousal level based on the maximum value of the spectral density and the frequency corresponding to the maximum value of the spectral density will be described. In the following description, the maximum value of the spectral density will be referred to as the maximum spectral density. Further, the frequency corresponding to the maximum spectral density will be referred to as the maximum frequency.

The determination unit 140 calculates the frequency f that satisfies the relationship of Expression (4) as the maximum frequency. The determination unit 140 obtains the maximum spectral density by substituting the maximum frequency into the equation (3).

The determination unit 140 selects one of the maximum spectral densities based on the spectral density data. For example, the determination unit 140 selects any of the maximum spectral densities 10a to 10d in FIG. 4, and pays attention to the selected maximum spectral density and the change with time of the maximum frequency corresponding to the selected maximum spectral density.

For example, the determination unit 140 plots the relationship between the maximum spectral density of interest and the maximum frequency corresponding to the maximum spectral density in a graph. The points on the graph determined by the maximum spectral density and the maximum frequency are referred to as feature points. The determination unit 140 determines the waking degree of the subject based on the positions of the feature points on the graph.

FIG. 5 is a diagram for explaining the process of determining the arousal level. The vertical axis of the graph 20 shown in FIG. 5 is the axis corresponding to the maximum spectral density. In the graph 20, the maximum spectral density decreases from the lower side to the upper side. The horizontal axis of the graph 20 is an axis corresponding to the maximum frequency. In the graph 20, the maximum frequency increases from the left side to the right side. When the maximum frequency decreases and the maximum spectrum increases, the arousal level of the subject decreases. On the other hand, when the maximum frequency is high and the maximum spectrum is low, the arousal level of the subject is high. That is, it can be said that when the feature point moves from the lower left to the upper right, the awakening degree of the subject moves in the awakening direction.

For example, when the position of the feature point is included in the area 20a, the determination unit 140 determines that the subject's arousal level is “arousal level 1”. The determination unit 140 determines that the arousal level of the subject is “arousal level 2” when the position of the feature point is included in the area 20b. When the position of the feature point is included in the area 20c, the determination unit 140 determines that the subject's arousal level is “arousal level 3”. When the position of the feature point is included in the region 20d, the determination unit 140 determines that the subject's arousal level is “awakening level 4”. The determination unit 140 determines that the arousal level of the subject is “arousal level 5” when the position of the feature point is included in the area 20e.

In the graph shown in FIG. 5, as an example, the area of the graph 20 is divided into 20a to 20e, and the arousal level of the subject is classified into any of the arousal levels 1 to 5, but the present invention is not limited to this. Not a thing. For example, the area of the graph 20 may be further divided to further finely classify the arousal level of the subject.

The determination unit 140 compares the awakening degree with a predetermined threshold TH1 and determines whether the awakening degree is less than the threshold TH1. When the arousal level is less than the threshold value TH1, the determination section 140 outputs information indicating that the arousal level of the subject is low to the estimation section 150.

Next, the process in which the determination unit 140 determines the driver's arousal level from the characteristic amount of each frequency range of the spectral density will be described. The heartbeat interval changes with respiration, that is, with the adjustment of autonomic nerves. Factors of fluctuation include, for example, HRSA called Mayer Wave Sinus Arrhythmia (MWSA) and respiratory sinus arrhythmia called RSA (Respiratory Sinus Arrhythmia). The cycle of the respiratory fluctuation in the heartbeat interval data includes a low frequency component (LF) near 0.05 to 0.15 Hz corresponding to MWSA and a high frequency side (LF) near 0.15 to 0.4 Hz corresponding to RSA ( HF) component. Therefore, the spectral density has the following characteristics. FIG. 6 is a diagram for explaining the characteristics of the frequency band. The horizontal axis of FIG. 6 represents frequency and the vertical axis represents spectral density. For example, 0.05 to 0.15 Hz is the frequency range of LF (Low Frequency). 0.15-0.4 Hz is the frequency range of HF (High Frequency). The sympathetic nerve activity state is likely to appear in the spectral density component of the LF frequency range. The parasympathetic nerve activity state is likely to appear in the spectral density component of the HF frequency range. The example of FIG. 6 shows the waveform of the distribution of the spectral density for each frequency in the high awakening state and the low awakening state. As shown in FIG. 6, the peak with the maximum spectral density in the HF frequency range is higher in the high awakening state than in the low awakening state. In addition, the high awakening state has a higher spectral density in the HF frequency range than the low awakening state.

Therefore, the determination unit 140 determines the awakening degree of the driver from the feature amount of the frequency distribution for each frequency range of LF and HF. This feature amount may be the spectral density of the maximum point that is a peak, or may be a value obtained by integrating the spectral density of the frequency range. The determination unit 140 obtains the arousal level based on the feature amount for each frequency range of LF and HF. For example, the determination unit 140 determines that the arousal level is higher as the feature amount in the HF frequency range is larger than the feature amount in the LF frequency range, and determines the arousal level. The determination unit 140 compares the arousal level with the threshold TH1 and determines whether the arousal level is equal to or higher than the threshold TH1. When the arousal level is not equal to or higher than the threshold value TH1, the determination section 140 outputs information indicating that the arousal level of the subject is low to the estimation section 150. For example, the determination unit 140 calculates an LF component that integrates the spectral density in the LF frequency range and an HF component that integrates the spectral density in the HF frequency range. The determination unit 140 determines that the driver is in an awake state when the ratio of the LF component and the HF component is 7:3, and when the ratio of the LF component and the HF component is 3:7. , And outputs information indicating that the awakening degree of the subject is low to the estimation unit 150.

When the estimation unit 150 receives the information indicating that the subject is in a low awakening degree from the determination unit 140, the estimation unit 150 estimates what state the subject is in in a low awakening state.

Here, the low awakening state may be a state other than the sleepy state of the driver. For example, the low awakening state includes a sleepy state and a relaxed state in which tension is released. The sleepy state is a state in which drowsiness is felt and attention is reduced. The relaxed state is a state in which the person is not drowsiness and is alert. When the driver is in a relaxed state, he/she is kept alert and quickly reacts to changes in the situation, so that the variation in the frequency distribution becomes large. On the other hand, when the driver is in a sleepy state, his alertness decreases and his reaction to changes in the situation decreases, so that the frequency distribution is biased and the variation in the frequency distribution is reduced.

Therefore, the estimation unit 150 estimates whether the subject is drowsy or relaxed according to the degree of variation in the frequency distribution obtained as a result of frequency analysis of the heartbeat interval data by the analysis unit 130.

Hereinafter, an example of a process in which the estimation unit 150 estimates whether the subject is drowsy or relaxed according to the degree of variation in the frequency distribution will be described. First, a process in which the estimation unit 150 estimates whether the subject is drowsy or relaxed by using entropy as the variation degree of the frequency distribution will be described.

FIG. 7 is a diagram showing an example of the frequency distribution. In the example of FIG. 7, patterns are given to the frequency distribution in the LF frequency range (0.05 to 0.15 Hz) and the frequency distribution in the HF frequency range (0.15 to 0.4 Hz). The estimation unit 150 obtains an integral value of the spectral density by integrating the spectral density of the frequency width for each predetermined frequency width of the frequency distribution. The frequency width may be any value as long as an integrated value roughly indicated by the frequency distribution in the target frequency range can be obtained, and is set to, for example, 0.01 Hz. The estimation unit 150 obtains, for each frequency width, the ratio of the integrated value of the spectral density of the frequency width to the integrated value of the spectral density of all frequencies. The ratio of the integral value of the spectral density for each frequency width is the probability of appearance in each frequency width. The appearance probability may be obtained as the ratio of the integral value of the spectral density of the frequency width to the integral value of the spectral density of the LF and HF frequency ranges.

The appearance probability Pi of the frequency width i can be represented by Expression (5-1) using a PSD (Power Spectral Density Function) representing the spectral density of the frequency f. Entropy H is represented by equation (5-2). Here, Pi has a probability distribution, i=1, 2,... N, and Δf=0.05 Hz. Δf may have a slightly smaller value other than 0.05 Hz, or may have a large value. However, it is desirable that the lower limit of the integration is 0.05 Hz, which is the lower limit of the LF of the heartbeat fluctuation frequency, and the upper limit is 0.4 Hz, which is the upper limit of the HF. The estimation unit 150 calculates the entropy H from the appearance probability Pi of the frequency width i using the equations (5-1) and (5-2).

The estimation unit 150 estimates that the subject is in a relaxed state when the calculated entropy H is greater than or equal to a predetermined threshold TH2, and estimates that the subject is drowsy when the entropy H is less than the threshold. The threshold value TH2 may be a fixed value or may be dynamically changed. For example, a standard threshold value TH2 may be stored in advance for each sex and age group, and the threshold value TH2 may be set according to the subject's sex and age group. In this embodiment, the threshold value TH2 is set to an initial value and is dynamically changed by the setting unit 160 described later. The initial value may be a single value, or may be set according to the subject's sex or age group. When the estimation unit 150 estimates that the subject is drowsy, the estimation unit 150 outputs information indicating that the subject is estimated to be drowsiness to the warning unit 170.

Next, a process in which the estimation unit 150 estimates whether the subject is drowsy or relaxed by using variance as the variation degree of the frequency distribution will be described.

The estimation unit 150 obtains the variance of the spectral density in a fixed section of the frequency distribution. The fixed section may be the entire frequency range or a specific frequency range such as the LF and HF frequency ranges (0.05 to 0.4 Hz). When n frequencies f ₁ , f ₂ ,..., F _n in a certain section are given, the estimation unit 150 expresses the average value thereof by the formula (6). Further, the dispersion S ² is represented by the equation (7).

The estimation unit 150 estimates that the subject is in a relaxed state when the calculated variance S ² is greater than or equal to a predetermined threshold TH2, and estimates that the subject is drowsy when the variance S ² is less than the threshold TH2. The threshold TH2 may be a fixed value or may be dynamically changed. In this embodiment, the threshold value TH2 is set to an initial value and is dynamically changed by the setting unit 160 described later. When the estimation unit 150 estimates that the subject is drowsy, the estimation unit 150 outputs information indicating that the subject is estimated to be drowsiness to the warning unit 170.

Here, the spectral density of the frequency distribution also changes depending on the physical condition of the subject. For example, in the case where the physical condition of the subject is poor, the spectral density of the frequency distribution is generally lower than in the case where the physical condition is normal (the physical condition is healthy). FIG. 8A is a diagram showing an example of the frequency distribution when the physical condition of the subject is normal. FIG. 8B is a diagram showing an example of the frequency distribution when the physical condition of the subject is poor. FIG. 8B is generally lower than FIG. 8A. When the physical condition of the subject is poor, the frequency distribution is less likely to be biased. Also, when the subject is in a poor physical condition, he or she tends to be less alert.

Therefore, when the spectral density of the frequency distribution is generally low, the estimation unit 150 may estimate that the subject is drowsy regardless of the degree of variation in the frequency distribution. For example, when the integrated value or the peak of the spectral density is equal to or less than the predetermined threshold TH3, the estimation unit 150 estimates that the subject is drowsy regardless of the degree of variation in the frequency distribution. The threshold value TH3 is set to a value at which the physical condition and the spectral density of the frequency distribution are obtained for a large number of subjects and the physical condition is considered to be poor.

The setting unit 160 sets a threshold TH2 used for distinguishing between a relaxed state and a sleepy state. Generally, a driver is low in drowsiness when he/she gets into a vehicle, and tends to be drowsiness when the driving time is long. Therefore, the setting unit 160 sets the threshold value TH2 from the frequency distribution for a predetermined period from the time when the subject gets in the vehicle. This predetermined period may be any period as long as the subject gets in the vehicle and the frequency distribution in a stable state is obtained, and is, for example, 5 minutes. For example, when entropy is used as the variation degree of the frequency distribution, the setting unit 160 calculates the entropy of the frequency distribution for a predetermined period from the time when the subject gets on the vehicle. Then, the setting unit 160 sets the threshold value TH2 to a value smaller than the calculated entropy. For example, the setting unit 160 sets the threshold TH2 to the value of 70% of the calculated entropy. Similarly, for example, when the variance is used as the variation degree of the frequency distribution, the setting unit 160 calculates the variance S ² of the frequency distribution in a predetermined period from the time when the subject gets in the vehicle. Then, the setting unit 160 sets the threshold value to a value smaller than the calculated variance S ² . For example, the setting unit 160 sets the threshold value TH2 to 70% of the value of the calculated variance ^{S 2.} As a result, the setting unit 160 can set the threshold value TH2 according to the driver, and thus can accurately identify the driver state.

The warning unit 170 is a processing unit that receives the estimation result from the estimation unit 150 and issues a warning according to the estimation result. Specifically, when the warning unit 170 receives information indicating that the subject is in a drowsiness state, the warning unit 170 warns the subject. The warning unit 170 may give a warning by sound, or may give a warning by video using a display installed in the vehicle.

Next, a flow in which the drowsiness detection device 100 according to the present embodiment executes drowsiness detection processing for detecting drowsiness will be described. FIG. 9 is a flowchart showing an example of the procedure of drowsiness detection processing. The drowsiness detection process is executed at a predetermined timing, for example, when the vehicle engine is started and a processing start instruction is received from the vehicle control unit.

As shown in FIG. 9, the sensor 110 of the drowsiness detection device 100 detects the heartbeat signal of the subject and acquires heartbeat signal data (S10). The heartbeat interval calculation unit 120 detects the amplitude peak of the heartbeat signal based on the heartbeat signal data, and calculates the interval of the detected timing (S11).

The analysis unit 130 calculates the spectral density corresponding to the heartbeat interval by performing frequency analysis on the heartbeat interval data obtained during the immediately preceding constant period (S12). The determination unit 140 derives the arousal level based on the maximum value of the spectral density corresponding to the heartbeat interval obtained by the frequency analysis and the frequency corresponding to the maximum value of the spectral density (S13).

The determination unit 140 determines whether the derived arousal level is equal to or higher than the threshold TH1 (S14). When the awakening level is equal to or higher than the threshold TH1 (Yes in S14), the process proceeds to S21 described later.

On the other hand, when the arousal level is not equal to or higher than the threshold TH1 (No in S14), the estimation unit 150 determines whether the spectral density of the frequency distribution is low overall. For example, the estimation unit 150 determines whether the peak of the spectral density is the threshold TH3 or less (S15). When the peak of the spectral density is less than or equal to the threshold TH3 (Yes in S15), the estimation unit 150 estimates that the subject is drowsy because the spectral density of the frequency distribution is low overall (S16). The warning unit 170 gives a warning to the subject (S17), and shifts to S21 described later.

On the other hand, when the peak of the spectral density is not equal to or lower than the threshold TH3 (No in S15), the estimation unit 150 calculates the entropy H of the spectral density (S18). The estimation unit 150 determines whether the entropy H is greater than or equal to the threshold TH2 (S19). If the entropy H is not equal to or greater than the threshold TH2 (No at S19), the process proceeds to S16. As a result, the subject is estimated to be in a drowsiness state, and the subject is warned.

On the other hand, when the entropy H is equal to or greater than the threshold value TH2 (Yes in S19), the estimation unit 150 estimates that the subject is in the relaxed state (S20), and proceeds to S21 described below.

The analysis unit 130 determines whether or not the end of processing has been instructed (S21). For example, when the engine of the vehicle is stopped and an instruction to end the process is received from the control unit of the vehicle, the analysis unit 130 determines that the instruction to end the process has been given. When the end of processing is instructed (Yes in S21), the processing ends.

On the other hand, if the end of processing has not been instructed (No at S21), the process proceeds to S10 described above.

As described above, the drowsiness detection device 100 according to the present embodiment performs frequency analysis on heartbeat data sequentially obtained from the subject. The drowsiness detection device 100 estimates whether the subject is drowsy or relaxed according to the degree of variation in the frequency distribution obtained as a result of the frequency analysis. Accordingly, the drowsiness detection device 100 can accurately estimate the sleepy state of the subject.

Further, the drowsiness detection device 100 according to the present embodiment calculates entropy or variance from the frequency distribution. The drowsiness detection apparatus 100 estimates that the subject is in a relaxed state when the calculated entropy or variance is equal to or greater than a predetermined threshold (threshold TH2). When the calculated entropy or variance is less than a threshold value (threshold value TH2), the drowsiness detection apparatus 100 estimates that the subject is drowsy. Accordingly, the drowsiness detection device 100 can determine whether the variation in the frequency distribution is large or small, so that the subject can accurately estimate the drowsiness state.

In addition, the drowsiness detection device 100 according to the present embodiment calculates entropy or variance from the frequency distribution of a predetermined period from the time when the subject gets in the vehicle, and sets a threshold value (threshold TH2) to a value smaller than the calculated entropy or variance. Set. Thereby, the drowsiness detection device 100 can set the threshold value TH2 according to the driver, and can accurately identify the driver's state.

Further, the drowsiness detection device 100 according to the present embodiment determines the arousal level of the subject from the spectral density of the heartbeat intervals calculated from the heartbeat data. The drowsiness detection apparatus 100 estimates whether the subject is drowsy or relaxed depending on the degree of variation in the frequency distribution when the waking degree of the subject is equal to or less than a predetermined value (threshold TH1). Accordingly, the drowsiness detection device 100 can estimate whether the state in which the degree of awakening is low is the drowsiness state or the relaxed state of the subject.

In addition, the drowsiness detection device 100 according to the present embodiment estimates that the subject is drowsiness regardless of the degree of variation in the frequency distribution when the integrated value or the peak of the spectral density is equal to or less than a predetermined value. As a result, the drowsiness detection device 100 can suppress the occurrence of an accident by issuing a warning when the awakening degree is low when the physical condition of the subject is poor.

Although the embodiments of the disclosed device have been described above, the disclosed technique may be implemented in various different forms other than the above-described embodiments. Therefore, other embodiments included in the present invention will be described below.

For example, in the above embodiment, the case has been described in which the subject is estimated to be in a drowsy state or a relaxed state based on the overall degree of variation in the frequency distribution. However, the disclosed device is not limited to this. The estimation unit 150 may estimate whether the subject is drowsy or relaxed based on the degree of variation in the frequency distribution in a specific frequency range. For example, the estimation unit 150 calculates entropy or variance for each frequency range of LF and HF from the frequency distribution. Then, the estimation unit 150 may estimate whether the subject is drowsy or relaxed based on the calculated entropy or variance of the frequency ranges of LF and HF. For example, the estimation unit 150 estimates a relaxed state when the entropy or variance of the LF and HF frequency ranges is greater than or equal to a predetermined threshold value, and when one or both of them does not exceed the threshold value, a drowsiness state is determined. It may be estimated. The estimation unit 150 may also estimate whether the subject is drowsy or relaxed based on the degree of variation in the frequency range of LF or HF.

Further, each component of each device shown in the drawings is functionally conceptual and does not necessarily have to be physically configured as shown. That is, the specific state of distribution/integration of each device is not limited to that shown in the figure, and all or a part of them may be functionally or physically distributed in arbitrary units according to various loads or usage conditions. It can be integrated and configured. For example, the processing units of the heartbeat interval calculation unit 120, the analysis unit 130, the determination unit 140, the estimation unit 150, and the setting unit 160 may be integrated as appropriate. Further, each processing unit may be divided into a plurality of processing units depending on the function. Further, each processing function performed by each processing unit may be realized in whole or in an arbitrary part by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by a wired logic. ..

Next, an example of a computer that executes a drowsiness detection program that realizes the same function as the drowsiness detection device 100 described in the above embodiment will be described. FIG. 10 is a diagram illustrating an example of a computer that executes a drowsiness detection program.

As shown in FIG. 10, the computer 200 includes a CPU 201 that executes various arithmetic processes, an input device 202 that receives data input from a user, and a display 203. The computer 200 also includes a reading device 204 that reads a program or the like from a storage medium, and an interface device 205 that exchanges data with another computer via a network. The computer 200 also has a RAM 206 for temporarily storing various information and a hard disk device 207. Then, each of the devices 201 to 207 is connected to the bus 208.

The hard disk device 207 has, for example, a drowsiness detection program 207a. The CPU 201 reads the drowsiness detection program 207a and loads it on the RAM 206. The drowsiness detection program 207a functions as a drowsiness detection process 206a. The drowsiness detection process 206a corresponds to, for example, the heartbeat interval calculation unit 120, the analysis unit 130, the determination unit 140, the estimation unit 150, and the setting unit 160.

The drowsiness detection program 207a does not have to be stored in the hard disk device 207 from the beginning. For example, each program is stored in a “portable physical medium” such as a flexible disk (FD), a CD-ROM, a DVD disk, a magneto-optical disk, an IC card, which is inserted into the computer 200. Then, the computer 200 may read the drowsiness detection program 207a from these and execute it.

100 drowsiness detection device 110 sensor 120 heartbeat interval calculation unit 130 analysis unit 140 determination unit 150 estimation unit 160 setting unit 170 warning unit

Claims (6)

Frequency analysis is performed on the heartbeat data sequentially obtained from the subject,
Determine the arousal level of the subject from the spectral density of the heartbeat interval calculated from the heartbeat data,
When the arousal level of the subject is less than or equal to a predetermined value, entropy or variance is calculated from the frequency distribution obtained as a result of the frequency analysis, and when the calculated entropy or variance is greater than or equal to a predetermined threshold value, the subject is in a relaxed state. If the entropy or variance is less than the threshold value , the computer causes the computer to perform a process of estimating that the subject is drowsy . A method for calculating entropy or variance from a frequency distribution for a predetermined period from the time when a subject gets into a vehicle, and causing a computer to further execute a process of setting the threshold value to a value smaller than the calculated entropy or variance. drowsiness detection program according to 1. The process of estimating calculates entropy or variance for each frequency range of LF (Low Frequency) and HF (High Frequency) from the frequency distribution, and based on the calculated entropy or variance of the frequency range of LF and HF. The drowsiness detection program according to claim 1 or 2 , wherein the subject is estimated to be in a drowsiness state or a relaxed state. In the estimating process, when the integrated value or the peak of the spectral density is equal to or less than a predetermined value, the subject is estimated to be in a drowsiness state regardless of the variation degree of the frequency distribution. The drowsiness detection program according to any one of 1 to 3 . Frequency analysis is performed on the heartbeat data sequentially obtained from the subject,
Determine the arousal level of the subject from the spectral density of the heartbeat interval calculated from the heartbeat data,
When the arousal level of the subject is less than or equal to a predetermined value, entropy or variance is calculated from the frequency distribution obtained as a result of the frequency analysis, and when the calculated entropy or variance is greater than or equal to a predetermined threshold value, the subject is in a relaxed state. And the entropy or variance is less than the threshold value , the computer executes a process of estimating that the subject is drowsy . An analysis unit that performs frequency analysis on heartbeat data sequentially obtained from the subject,
A determination unit that determines the arousal level of the subject from the spectral density of heartbeat intervals calculated from the heartbeat data,
As a result of the determination by the determination unit, if the arousal level of the subject is less than or equal to a predetermined value, entropy or variance is calculated from the frequency distribution obtained as a result of the frequency analysis by the analysis unit , and the calculated entropy or variance is a predetermined threshold value. In the above case, the subject is estimated to be in a relaxed state, if the entropy or variance is less than the threshold value, the estimation unit to estimate that the subject is drowsy ,
A drowsiness detection device comprising:

Images