JPH06107030A - Dozing driving detection device - Google Patents

Abstract

PURPOSE:To minimize a detection delay of dozing by providing a judging means for judging a dozing state of a driver based on a frequency characteristic obtained by a frequency analysis. CONSTITUTION:A prepared analysis data is inputted to an FFT signal processing part 14, therein subjected to a fast Fourier transform and a signal in a time region is transformed to a signal in a frequency region. An obtained signal is fed to a specific frequency domain integration part 16. Increase of a low frequency component due to dozing driving can be detected regardless of a traveling road shape. Whether or not a difference value in a predetermined frequency band is larger than a predetermined value is examined by a judging part 22 so as to detect dozing. A judgement result based on the analysis data is highly correlated with a brain wave (alpha wave), and in accordance with this value the detection of dozing can be conducted effectively and further without a time delay.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a dozing driving detection device,
In particular, the present invention relates to a device for detecting a drowsy driving from a change in vehicle operation amount such as a steering angle.

[0002]

2. Description of the Related Art Conventionally, various devices have been proposed in order to prevent a driver from falling asleep while traveling on a highway or the like. In order to prevent drowsiness efficiently, it is important to detect the drowsiness state of the driver, and various drowsiness detection devices have been developed.

For example, Japanese Laid-Open Patent Publication No. 60-76424 (Prior Art 1) discloses that the drowsiness detection is performed by turning the steering wheel by a certain angle. Prior art 2) shows that the drowsiness detection is performed because the high frequency component of the steering angle change within a predetermined time is equal to or greater than a predetermined value.

However, in Conventional Example 1, there is a case where a curve running is detected as dozing.
In the above, there is a problem that the high frequency component does not always appear when the person falls asleep, and the high frequency component may appear in the normal state, so that the drowsiness detection cannot be sufficiently performed.

Related Art Therefore, the applicant of the present invention has proposed an improved doze driving detection apparatus in Japanese Patent Application No. 3-222967.
In this device, a sample of a vehicle state quantity such as a steering angle for a predetermined time is subjected to frequency analysis (fast Fourier transform), a predetermined low frequency component (a component of a predetermined frequency band) is extracted, and the extracted low frequency component is extracted. The component is compared with the low-frequency component amount during normal operation, and when this component increases by a predetermined amount or more, dozing is detected.

When the driver's consciousness begins to deteriorate, the operations of situation recognition, judgment, and steering become slow, and the low-frequency component of the change in the vehicle state quantity increases. Therefore, by comparing the low frequency component of the vehicle state quantity during normal operation and the low frequency component of the sample during driving, it is possible to detect the drowsy driving relatively accurately.

[0007]

The frequency resolution of the fast Fourier transform depends on the sampling time of the data to be frequency-transformed. That is, the frequency resolution is 1 / data length (sec), and it takes about 9 minutes to analyze the frequency of the target frequency component (0.05 to 0.6 Hz). . Therefore,
Frequency analysis will be performed based on the data for the past 9 minutes. Therefore, in the previously proposed device, even if the frequency analysis is performed while updating the data for 9 minutes every 10 seconds, the obtained result will be delayed by 4.5 minutes, and the detection of drowsiness will be detected. There was a problem that was delayed.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to provide a drowsy driving detection apparatus capable of minimizing the detection delay without lowering the resolution.

[0009]

DISCLOSURE OF THE INVENTION The present invention relates to a drowsiness driving detection device for detecting a drowsiness of a driver from a running state of a vehicle, and an operation amount detecting means for detecting an operation state of the vehicle,
The analysis data is extracted from the operation state data detected by the operation amount detection means for a first predetermined time, and dummy data determined in advance is added to the operation state data for a second predetermined time. Analysis data creating means to be created, frequency analysis means for performing frequency analysis on the obtained analysis data by fast Fourier transform, and determination means for determining the dozing state of the driver based on the frequency characteristics obtained by this frequency analysis And having.

[0010]

As described above, in the present invention, the analysis data creating means adds the dummy data to the actually detected operation state data. Therefore, for example, 9 minutes of analysis data can be created by adding 8 minutes of dummy data to 1 minute of operation state data. The resolution by the fast Fourier transform depends on the data length (time) of the data to be analyzed, but the analysis data of the present invention has a sufficient length by adding the dummy data. Therefore, frequency analysis can be performed with sufficient resolution. Further, since the dummy data is added to create the analysis data, the actual operation state data may have a short data length. Therefore, the time delay of frequency analysis can be minimized.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows a configuration block diagram of the present embodiment, and a steering angle sensor 10 is attached to a steering shaft of a vehicle. As the steering angle sensor 10, for example, a slit plate press-fitted into the steering shaft and two sets of photointerrupters attached to the column tube can be used, and the photointerrupter is shielded from light by the slit plate as the steering shaft rotates. The turning direction and the turning angle of the steering wheel can be detected by turning on and off by turning the steering wheel off.

The detection signal from the steering angle sensor 10 is input to the microcomputer via the input port. In this microcomputer, a CPU and RA
M is built-in, and signal processing is performed by arithmetic processing performed by each arithmetic unit described later, which is configured by a CPU, to detect a dozing driving.
That is, first, the detection signal is R in the microcomputer.
It is input to the steering angle data memory 12 such as AM, and the steering angle data for the past 2 minutes, for example, is stored while being sequentially updated every 10 seconds. And this steering angle data memory 1
The steering angle data stored in 2 is output for one minute each time it is updated.

Data for one minute, which is output from the steering angle data memory 12 every 10 seconds, is input to the dummy data adding section 13 where dummy data is added. That is, as shown in FIG. 2, dummy data of 4 minutes each is added before and after the actual data of the detected steering angle for 1 minute to create analysis data of 9 minutes.
This dummy data is added for the purpose of effectively performing the frequency analysis based on the actual data for 1 minute, and it is preferable that the noise is added as little as possible. Therefore, in this example, data having a value of 0 is adopted as the dummy data. Note that it is also preferable to repeat the actual data for 1 minute 9 times to create the analysis data because the addition of noise is small. Further, actual data and 0 data may be combined at an appropriate ratio.

Further, the length of the dummy data is not limited to 8 minutes in this embodiment, but if it is too long, the load on the hardware for executing the calculation becomes large, so that it depends on the calculation speed of the hardware. It seems that about 8 minutes is appropriate. If the data length is shortened, a problem of resolution will occur.

Furthermore, the length of the actual data is not limited to one minute. However, since the steering cycle is often about 4 to 5 seconds, at least about 1 minute is required, and if the actual data length is increased, the time delay in frequency analysis becomes large. Therefore, about 1 minute in the embodiment is preferable. The positional relationship between the dummy data and the actual data is F
It depends on the window relationship when performing FT analysis. Therefore, the actual data does not need to be in the center, and may be appropriately determined on a case-by-case basis, such as by placing it at the end.

The analysis data thus created is input to the FFT signal processing unit 14 where it is subjected to fast Fourier transform, and the time domain signal is transformed into the frequency domain signal. Then, the obtained signal is supplied to the specific frequency section integrator 16.

The specific frequency section integrator 16 extracts a low frequency component (0.6 Hz or less) from the transmitted signal, and outputs 0.1 Hz to 0.2 Hz and 0.2 Hz to 0.3 H.
z, 0.3 Hz to 0.4 Hz, 0.4 Hz to 0.5 H
z, integration of 5 sections of 0.5 Hz to 0.6 Hz is performed (S
103). In addition, the method of accumulation is 0.05-
0.15, ... 0.45-0.55 Hz or the like may be used. Then, if, for example, 10 minutes have not passed after the start of operation, the data during normal operation can be obtained. Therefore, an appropriate number of frequency analysis data during this period are averaged and stored in the average value memory 18 for each of the 6 frequency sections. For example, it is advisable to calculate and store an average of 100 seconds (10 times of data because analysis data is obtained every 10 seconds) around the time of 5 minutes. On the other hand, if 10 minutes have elapsed after the start of operation, the integrated value of each frequency section calculated by the specific frequency section integration unit 16 is supplied to the comparison unit 20 every 10 seconds.

Further, the comparison unit 20 is supplied with the average of the steering angle data stored in the above-mentioned average value memory 18 before the lapse of 10 minutes from the start of operation (during normal operation). Therefore, the current integrated value and the average value during normal operation before 10 minutes from the start of operation are supplied to the comparing section 20, and the comparing section 20 calculates the difference between them and compares the two values. This is performed for each frequency section. And
The comparison result, that is, the difference value is supplied to the determination unit 22, and the difference value and a predetermined value are compared.

Here, FIG. 3 shows the frequency spectrum of the steering angle during normal driving and during drowsiness driving. Compared with the normal driving time of FIG. 3A, during drowsiness driving of FIG. 3B. It can be seen that the low frequency component (region a in the figure) increases in the case of. The change in the extremely low frequency region (region b in the figure) is due to the change in the steering angle on the curved road, and has no relation to whether or not the driver is dozing. Therefore, the difference value in this area is excluded. Therefore, it is possible to detect an increase in the low frequency component due to the dozing driving regardless of the shape of the running path. Therefore, the determination unit 22 checks whether or not the difference value in the predetermined frequency band is equal to or larger than the predetermined value, and detects dozing.

FIG. 4 shows the temporal change of the difference value obtained in this way in this embodiment together with the result of the measurement of the brain wave (α wave) corresponding to the case of only one minute of data and drowsiness. There is. From this, it can be seen that the determination result based on the analysis data according to the present embodiment has a very high correlation with the α-wave, and depending on this value, the drowsiness detection can be effectively performed.
Further, in the case of this embodiment, there is no time delay. Therefore, suitable snooze detection can be performed.

When the difference value is a predetermined value or more, for example, twice the value of the sample during normal operation or more, the determining unit 22 determines that the dozing driving is occurring. In this case, the determination unit 22 supplies a signal to the alarm unit 24 such as a lamp, a buzzer, and a sound generator to give an alarm to wake up the driver. By repeating these steps every 10 seconds, it is possible to constantly monitor the driver's consciousness and prevent the driver from falling asleep.

In this embodiment, the steering angle sensor 10 is used to detect the dozing driving by focusing on the low frequency component in the frequency spectrum of the steering angle.
A lateral displacement sensor may be used instead of, and the dozing driving can be similarly detected by focusing on the low frequency component in the frequency spectrum of the lateral displacement. This is because the lateral displacement amount of the vehicle and the steering angle are in a proportional relationship. The same analysis can be performed using a lateral G sensor or the like.

In the present embodiment, the steering angle data for 10 minutes after the start of operation is compared with the current value as a sample during normal operation. However, a vehicle speed sensor is further provided in the configuration of the present embodiment, and 10 seconds after the start of operation. Further, the steering angle data for one minute is stored in the average value memory 18 as a sample in a normal operation for each vehicle speed (for example, an average value of 0 to 30 km / h, 30 to 40 km).
/ H, an average value of 40 to 50 km / h), a current steering angle data and an average value corresponding to the current vehicle speed may be compared for each vehicle speed section to determine the dozing driving.

This is because the change in the steering angle generally depends largely on the vehicle speed (as the vehicle speed increases, the driver's perception
Judgment-Since the operation is performed quickly, the steering frequency,
The lateral displacement frequency becomes high. ), Therefore, the steering angle spectrum also changes according to the vehicle speed.Therefore, comparing the current value with the sample at different vehicle speeds, the low-frequency component differs from the sample despite normal operation, and in some cases it is erroneously determined to be dozing driving. This is because it will end up.

Further, in the present embodiment, when a drowsy driving is detected, an alarm is given by a lamp, a buzzer or the like. However, not only the alarm is simply given, but the accelerator is automatically returned or the brake is applied. It can also be configured to automatically apply and decelerate. That is,
The alarm unit 2 when the determination unit 22 determines that the driver is dozing
4 as well as the control signal to the throttle actuator and the brake actuator.

[0026]

As described above, according to the doze driving detection apparatus of the present invention, since dummy data is added and frequency analysis is performed, frequency analysis with high resolution and little time delay is performed. Therefore, it is possible to reliably detect the dozing driving.

[Brief description of drawings]

FIG. 1 is a configuration block diagram of an embodiment of the present invention.

FIG. 2 is a characteristic diagram showing a structure of analysis data.

FIG. 3 is a characteristic diagram showing a frequency spectrum of steering during normal times and during dozing.

FIG. 4 is a characteristic diagram illustrating a relationship between a difference value and α waves.

[Explanation of symbols]

 10 Steering Angle Sensor 13 Dummy Data Addition Section 14 FFT Signal Processing Section 18 Average Value Memory 20 Comparison Section 22 Judgment Section 24 Warning Section

Claims (1)

[Claims] 1. A drowsy driving detection device for detecting a driver's dozing from the running state of a vehicle, comprising: an operation amount detecting means for detecting an operation state of the vehicle; and an operation state data detected by the operation amount detecting means. 1 for a predetermined period of time, and at the same time, add predetermined dummy data to this operating state data for a second predetermined period of time to create analysis data, and an analysis data creating means, A drowsiness driving detection device comprising: a frequency analysis means for frequency-analyzing data by fast Fourier transform; and a judgment means for judging a driver's dozing state based on the frequency characteristics obtained by this frequency analysis.