CN103473890A - Driver fatigue real-time monitoring system and monitoring method based on multi-information - Google Patents

Abstract

The invention discloses a driver fatigue real-time monitoring system and monitoring method based on multi-information. The monitoring system is characterized by comprising an information acquisition unit, an information processing unit and a warning unit; the information acquisition unit comprises a pressure sensor used for extracting grip strength information of a steering wheel by a driver, an angular transducer used for extracting turning angle information of the steering wheel, and a range sensor used for extracting position information of the head of the driver; the information processing unit is used for processing the information obtained by the information acquisition unit, thus characteristic parameters representing the fatigue degree of the state of the driver are obtained, and the level of the state of fatigue of the driver is judged according to the characteristic parameters; the warning unit is used for giving an alarm when the information processing unit judges that the driver is in a non-waking state. The monitoring system has the advantages of high reliability, low cost, high timeliness and a good monitoring effect, and a road accident caused by fatigue driving can be avoided to a large extent.

Description

Driver fatigue real-time monitoring system and monitoring method based on many information

Technical field

The invention belongs to intelligent transportation and security fields, be specifically related to driver fatigue real-time monitoring system and monitoring method based on many information.

Background technology

In recent years, road traffic accident sharply increased, and one of its major reason is exactly fatigue driving, for this reason, scientificlly and effectively monitor driver's driving condition, and give driver's prompting and report to the police, become the study hotspot in driver's active safety monitoring field.

In current research, a lot of countermeasures are used to carry out the monitoring of fatigue driving state: utilize the most invasives of method of biosensor monitoring driver physiological indexes strong, can cause interference to the driver, poor practicability; And use the method reliability of single fatigue characteristic low, practicality is not strong; And utilize the method for machine vision and image processing techniques, and changed by light and driver's individual factors affects greatly, cause the high or fatigue characteristic of these monitoring device costs to extract difficult, be difficult to obtain widespread use; Although also have many research to merge a plurality of fatigue characteristics, seldom consider driving fatigue state significantly behavioural characteristic relatively more directly perceived.The most monitoring method judges driving condition by the method for setting the evaluation index threshold value, but is subject to the factor impacts such as individual difference and driving habits, causes the evaluation index threshold value difficult definite, and the monitoring real-time is not high, and effect is undesirable.

Summary of the invention

The present invention seeks to overcome the deficiency that prior art exists, provide that a kind of reliability is high, cost is low, real-time, monitoring effect is desirable, general applicability is high, driver fatigue real-time monitoring system and monitoring method based on many information.

The present invention is that the technical solution problem adopts following technical scheme:

The characteristics that the present invention is based on the driver fatigue real-time monitoring system of many information are: described monitoring system is by information acquisition unit, information process unit and warning cell formation;

Described information acquisition unit is provided with: pressure transducer is arranged on the bearing circle top layer, for extracting the grip information of driver to bearing circle; Angular transducer, be arranged on the bearing circle rotary column, for extracting the corner information of bearing circle; Range sensor, be arranged on the pilot set headrest upper, for extracting driver's head position information; Described grip information, corner information and driver's the head position information of take is driving condition information;

Described information process unit, for described driving condition information is processed, obtains the fatigue strength characteristic parameter that characterizes driving condition, with described fatigue strength characteristic parameter judgement driver fatigue state grade;

Described warning unit for carrying out alarm when information process unit judgement driver is in non-waking state.

The characteristics that the present invention is based on the driver fatigue method of real-time of many information are to comprise the steps:

Step 1, gather driver's driving condition information, described driving condition information refer to by pressure transducer detects the driver that obtains to the grip information of bearing circle, by angular transducer, detects acquisition bearing circle corner information and detect the driver's of acquisition head position information by range sensor;

Step 2, the described driving condition information of step 1 is processed, obtained the fatigue strength characteristic parameter that characterizes driving condition;

Step 3, by step 1 and the described method of step 2, obtain the sample set that a sample number is N, described sample refers to that calculating by step 2 the fatigue strength characteristic parameter obtained forms with driver fatigue state grade determined by subjective assessment and that described fatigue strength characteristic parameter is corresponding; Build three layers of BP network, utilize described sample set to carry out off-line training to described three layers of BP network, obtain characterizing the mathematical model of Nonlinear Mapping relation between fatigue strength characteristic parameter and driver fatigue state grade;

Step 4, in each judgement constantly, obtain in real time current judgement fatigue strength characteristic parameter constantly by step 1 and the described method of step 2, the input signal that the described current judgement fatigue strength characteristic parameter constantly of take is step 3 gained mathematical model, utilize the current judgement of described mathematical model judgement driver fatigue state grade constantly.

The characteristics that the present invention is based on the driver fatigue method of real-time of many information also are:

In observation process, system be take the driving condition information of 0.2s as cycle Real-time Collection driver; The collection initial period of 30s of take is the default stage, and in the described default stage, the driver is in abnormal driving state, the 30s of take in the default stage after the image data of the 20s initial acquisition data that are each driving condition information; When finishing, the default stage of described 30s enters the monitoring stage, in the described monitoring stage, system is carried out the judgement of a driver fatigue state grade every 2s, the fatigue strength characteristic parameter that each judgement constantly adopts is that different fatigue strength characteristic parameters have the sampling time window of identical or different duration with before this judgement constantly and be that to take this judgement be that image data of the sampling time window of the finish time is calculated and obtained constantly.

Described fatigue strength characteristic parameter is grip amplitude fatigue strength FGA, corner standard deviation fatigue strength FASD, corner frequency fatigue strength FAF and biased from fatigue strength FPD;

Described grip amplitude fatigue strength FGA obtains as follows:

After the default stage of utilization, the initial acquisition data of 20s calculate driver's grip initial value F ₀for:

in formula with be respectively the driver's left hand of 20s after the default stage and the grip average of the right hand; N judgement grip value F constantly within the monitoring stage _nfor:

n=1,2,3..., in formula

with

before being respectively n judgement constantly and be that to take the n judgement be driver's left hand of 3s sampling time window of the finish time and the grip average of the right hand constantly; N judges grip amplitude fatigue strength FGA constantly _nfor: if FGA _nbe less than zero, by FGA _nassignment is zero;

Described corner standard deviation fatigue strength FASD obtains as follows:

Utilizing after the default stage initial acquisition data of 20s to calculate corner standard deviation initial value is SD ₀, within the monitoring stage before n judgement constantly and be that to take the n judgement be that the corner standard deviation of the 5s sampling time window of the finish time is SD constantly _n, n=1,2,3..., n judgement corner standard deviation fatigue strength FASD constantly _nfor:

Described corner frequency fatigue strength FAF obtains as follows:

Utilizing after the default stage initial acquisition data of 20s to calculate corner correction frequency initial value is AN ₀, within the monitoring stage before n judgement constantly and be that to take the n judgement be that the corner correction frequency of the 20s sampling time window of the finish time is AN constantly _n, n=1,2,3..., n judgement corner frequency fatigue strength FAF constantly _nfor: if FAF _nbe less than zero, by FAF _nassignment is zero;

Described biased obtains as follows from fatigue strength FPD:

Described range sensor is two ultrasonic sensors, described two ultrasonic sensors lay respectively at left side and the right side of pilot set headrest, and after the default stage of utilization, the initial acquisition data of 20s calculate being respectively apart from average of driver's head periphery and described two ultrasonic sensors with

center, driver's head square section and two ultrasonic sensors apart from initial value x ₀and y ₀be respectively:

with

r is the cross section mean radius of driver's head; The distance value x of n judgement moment center, driver's head square section and two ultrasonic sensors within the monitoring stage _nand y _nbe respectively:

with

n=1,2,3..., in formula

with

before being respectively n judgement constantly and be take the n judgement be constantly driver's head of 3s sampling time window of the finish time peripheral with two ultrasonic sensors apart from average;

Order: the center of two ultrasonic sensors is respectively an A and some B, and center, default stage driver head square section is some C; N judgement center, driver's head square section constantly is some D; Have: α is angle between straight line AC and straight line AB; β is angle between straight line AD and straight line AB; θ is the poor of β and α; L is some A and the spacing of putting B.According to the cosine law and trigonometric function formula, have:

$\cos \mathrm{\&alpha;}=\frac{{x_0}^2+L^2-{{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_0}^2}{{2x}_{0}L},\sin \mathrm{\&alpha;}=\sqrt{1-{\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2\mathrm{\&alpha;}}$

$\cos \mathrm{\&beta;}=\frac{{x_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">n</figure-callout>}}}^2+L^2-{{\mathrm{<figure-callout\; id="2"\; label="y\; n"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_n}^2}{{2x}_{n}L},\sin \mathrm{\&beta;}=\sqrt{1-{\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2\mathrm{\&beta;}}$

cosθ=cos(β-α)=cosαcosβ+sinαsinβ

N judgement constantly driver's head position depart from default stage driver head position apart from S _nfor: $S_n=\sqrt{{x_0}^2+{x_n}^2-{2x}_0{x}_n\cos \mathrm{\&theta;}};$ Constantly biased from fatigue strength FPD of n judgement _nfor: $\mathrm{FP}{D}_n=\frac{S_n}{L}.$

Three layers of BP network that described step 3 builds are: ground floor is input layer, 4 input nodes, consists of, and described 4 input nodes are distinguished 4 corresponding input component x ₁, x ₂, x ₃and x ₄correspond to successively grip amplitude fatigue strength FGA, corner standard deviation fatigue strength FASD, corner frequency fatigue strength FAF and biased from fatigue strength FPD, 4 input components form 1 input vector X and are: X=[x ₁, x ₂, x ₃, x ₄] ^t; The second layer is hidden layer; The 3rd layer is output layer, 3 output nodes, consists of, and described 3 output nodes are distinguished 3 corresponding output component y ₁, y ₂and y ₃mean successively waking state, fatigue state and degree of depth fatigue state, 3 output components form 1 output vector Y and are: Y=[y ₁, y ₂, y ₃] ^t, with Y=[1,0,0] ^tbe characterized by waking state, with Y=[0,1,0] ^tbe characterized by fatigue state, with Y=[0,0,1] ^tbe characterized by degree of depth fatigue state; Described fatigue state and degree of depth fatigue state are non-waking state.

Compared with prior art, beneficial effect of the present invention is embodied in:

1, monitoring system of the present invention is in the 30s of observation process presets the stage, every driver must be in abnormal driving state, for the auxiliary fatigue strength characteristic parameter calculated monitoring stage every driver, compare the method that fixed value is set, improved the general applicability of system, practicality strengthens;

2, monitoring system of the present invention is in the monitoring stage, system is carried out the judgement of a driver fatigue state grade every 2s, the fatigue strength characteristic parameter that each judgement constantly adopts is with before this judgement constantly and be that to take this judgement be that image data of the sampling time window of the finish time is calculated and obtained constantly, different fatigue strength characteristic parameters has the sampling time window of identical or different duration, both improve the utilization factor of data resource, also improved the real-time of system;

3, in the present invention, utilize the driver to carry out many information fusion to the grip information of bearing circle, the corner information of bearing circle and driver's head position information, with other, use the monitoring method of single fatigue characteristic to compare with complex appts, reliability is high, fatigue characteristic extracts easily and the monitoring device cost is low;

4, in the present invention, adopt the BP network to be identified, set up the mathematical model of Nonlinear Mapping relation between fatigue strength characteristic parameter and driver fatigue state grade, adopt compared to existing technology the determination methods of evaluation index threshold value, more effectively accurately.

The accompanying drawing explanation

Fig. 1 is the schematic diagram that monitoring system of the present invention is extracted the fatigue strength characteristic parameter;

Fig. 2 is that in the present invention, ultrasonic sensor detects principle schematic.

Specific embodiments

Driver fatigue real-time monitoring system based on many information in the present embodiment is by information acquisition unit, information process unit and warning cell formation; Wherein information acquisition unit is provided with: pressure transducer is arranged on the bearing circle top layer, for extracting the grip information of driver to bearing circle; Angular transducer, be arranged on the bearing circle rotary column, for extracting the corner information of bearing circle; Range sensor, be arranged on the pilot set headrest upper, for extracting driver's head position information; Grip information, corner information and driver's the head position information of take is driving condition information; Information process unit, for driving condition information is processed, obtains the fatigue strength characteristic parameter that characterizes driving condition, with fatigue strength characteristic parameter judgement driver fatigue state grade; The warning unit for carrying out alarm when information process unit judgement driver is in non-waking state.

In the present embodiment, for the collection of driving condition information, can use existing multiple sensing technology.In concrete enforcement, the driver can obtain by the softness haptic perception pressure transducer the grip information of bearing circle; The corner information of bearing circle can obtain by analog angular transducer; Driver's head position information can obtain by ultrasonic sensor.

In the present embodiment, the softness haptic perception pressure transducer is to using sensor prepared as the flexible pressure-sensitive material by carbon black filled silicon rubber, correlation technique is in " functional material " the 2nd phase in 2010, and " for the conducing composite material research of composite flexible touch sensor " that the people such as Zhao Xing, Huang Ying deliver is existing open.In the present embodiment, 16 softness haptic perception pressure transducers are distributed in to the bearing circle top layer equally spacedly, are in the bearing circle periphery; Warning device in softness haptic perception pressure transducer, analog angular transducer, ultrasonic sensor, warning unit all is connected with the information process unit hardware unit with digitizing LCD automobile Displaying Meter.

Driver fatigue method of real-time based on many information in the present embodiment comprises the steps:

Step 1, gather driver's driving condition information, driving condition information refer to by pressure transducer detects the driver that obtains to the grip information of bearing circle, by angular transducer, detects acquisition bearing circle corner information and detect the driver's of acquisition head position information by range sensor;

Step 2, information process unit are processed driving condition information, obtain for judging the fatigue strength characteristic parameter of driver fatigue state grade, the fatigue strength characteristic parameter comprises: grip amplitude fatigue strength FGA, corner standard deviation fatigue strength FASD, corner frequency fatigue strength FAF and biased from fatigue strength FPD;

Step 3, by step 1 and step 2 method, obtain the sample set that a sample number is N, sample refers to be calculated the fatigue strength characteristic parameter that obtains and consisted of the determined driver fatigue state grade corresponding with the fatigue strength characteristic parameter of subjective assessment by step 2; Build three layers of BP network, utilize sample set to carry out off-line training to three layers of BP network, obtain characterizing the mathematical model of Nonlinear Mapping relation between fatigue strength characteristic parameter and driver fatigue state grade;

In concrete enforcement, build three layers of BP network and be: ground floor is input layer, 4 input nodes, consists of, and 4 input nodes distinguish 4 of correspondence and inputted component x ₁, x ₂, x ₃and x ₄correspond to successively grip amplitude fatigue strength FGA, corner standard deviation fatigue strength FASD, corner frequency fatigue strength FAF and biased from fatigue strength FPD, 4 input components form 1 input vector X and are: X=[x ₁, x ₂, x ₃, x ₄] ^t; The second layer is hidden layer, and the hidden layer node number is determined by training; The 3rd layer is output layer, 3 output nodes, consists of, and 3 output nodes are distinguished 3 corresponding output component y ₁, y ₂and y ₃mean successively waking state, fatigue state and degree of depth fatigue state, 3 output components form 1 output vector Y and are: Y=[y ₁, y ₂, y ₃] ^t, with Y=[1,0,0] ^tbe characterized by waking state, with Y=[0,1,0] ^tbe characterized by fatigue state, with Y=[0,0,1] ^tbe characterized by degree of depth fatigue state, wherein fatigue state and degree of depth fatigue state are non-waking state.

For obtain sample set for example: to 14 male sex and 6 women totally 20 drivers tested, age is 24-55 year, average 6.1 years driving ages, test period is afternoon 11 up to 13 o'clock, night 24 up to 2 o'clock mornings and 4 up to 6 o'clock these three periods and process of the test is recorded a video, video be take to 30s is marked to tested personnel's driving condition by 5 trained testing crews are independent as interval, get the subjective assessment of its mean value as tested personnel's driving condition, the classification grade of subjective assessment is identical with the driver fatigue state grade, select wherein 300 groups of samples that comprise various driver fatigue state grades to form sample set.

Step 4, in each judgement constantly, obtain in real time current judgement fatigue strength characteristic parameter constantly by step 1 and step 2 method, the input signal that the current judgement fatigue strength characteristic parameter constantly of take is step 3 gained mathematical model, utilize the current judgement of mathematical model judgement driver fatigue state grade constantly.

In concrete enforcement:

In observation process, system be take the driving condition information of 0.2s as cycle Real-time Collection driver; The collection initial period of 30s of take is the default stage, and in the default stage, the driver is in abnormal driving state, the 30s of take in the default stage after the image data of the 20s initial acquisition data that be each driving condition information, front 10s is used for the driver and adjusts driving condition; When finishing, the default stage of 30s enters the monitoring stage, in the monitoring stage, system is carried out the judgement of a driver fatigue state grade every 2s, the fatigue strength characteristic parameter that each judgement constantly adopts is that different fatigue strength characteristic parameters have the sampling time window of identical or different duration with before this judgement constantly and be that to take this judgement be that image data of the sampling time window of the finish time is calculated and obtained constantly.Fig. 1 is the schematic diagram that monitoring system of the present invention is extracted the fatigue strength characteristic parameter.

Grip amplitude fatigue strength FGA obtains as follows:

The driver holds bearing circle and touches two softness haptic perception pressure transducers, and after the default stage of utilization, the initial acquisition data of 20s calculate driver's grip initial value F ₀for:

in formula

with

be respectively the driver's left hand of 20s after the default stage and the grip average of the right hand; N judgement grip value F constantly within the monitoring stage _nfor:

n=1,2,3..., in formula

with

before being respectively n judgement constantly and be that to take the n judgement be driver's left hand of 3s sampling time window of the finish time and the grip average of the right hand constantly; The driver, when fatigue state, increases and reduces gradually with degree of fatigue the grip of bearing circle, n judgement grip amplitude fatigue strength FGA constantly _nfor:

if FGA _nbe less than zero, by FGA _nassignment is zero;

Corner standard deviation fatigue strength FASD obtains as follows:

Utilizing after the default stage initial acquisition data of 20s to calculate corner standard deviation initial value is SD ₀, within the monitoring stage before n judgement constantly and be that to take the n judgement be that the corner standard deviation of the 5s sampling time window of the finish time is SD constantly _n, n=1,2,3..., the driver is when fatigue state, and with the characteristics of significantly revising bearing circle, when degree of fatigue is deepened, the steering wheel angle amplitude also shows in a long time the characteristics without significant change, n judgement corner standard deviation fatigue strength FASD constantly _nfor: ${\mathrm{FASD}}_n=|1-\frac{{\mathrm{<figure-callout\; id="0"\; label="SD\; n\; SD"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">SD</figure-callout>}}_n}{{\mathrm{SD}}_{}}|;$

Corner frequency fatigue strength FAF obtains as follows:

Utilizing after the default stage initial acquisition data of 20s to calculate corner correction frequency initial value is AN ₀, within the monitoring stage before n judgement constantly and be that to take the n judgement be that the corner correction frequency of the 20s sampling time window of the finish time is AN constantly _n, n=1,2,3..., the driver is when fatigue state, and the driver obviously reduces the correction frequency of bearing circle, n judgement corner frequency fatigue strength FAF constantly _nfor:

if FAF _nbe less than zero, by FAF _nassignment is zero;

Biased from fatigue strength FPD, obtain as follows:

Two ultrasonic sensors are set, lay respectively at left side and the right side of pilot set headrest, after the default stage of utilization, the initial acquisition data of 20s calculate being respectively apart from average of driver's head periphery and two ultrasonic sensors

with

center, driver's head square section and two ultrasonic sensors apart from initial value x ₀and y ₀be respectively: with

r is the cross section mean radius of driver's head; The distance value x of n judgement moment center, driver's head square section and two ultrasonic sensors within the monitoring stage _nand y _nbe respectively:

with n=1,2,3..., in formula

with

before being respectively n judgement constantly and be take the n judgement be constantly driver's head of 3s sampling time window of the finish time peripheral with two ultrasonic sensors apart from average;

Order: the center of two ultrasonic sensors is respectively an A and some B, and center, default stage driver head square section is some C; N judgement center, driver's head square section constantly is some D; Have: α is angle between straight line AC and straight line AB; β is angle between straight line AD and straight line AB; θ is the poor of β and α; L is some A and the spacing of putting B.Fig. 2 is that in the present invention, ultrasonic sensor detects principle schematic.According to the cosine law and trigonometric function formula, have:

$\cos \mathrm{\&alpha;}=\frac{{x_0}^2+L^2-{{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_0}^2}{{2x}_{0}L},\sin \mathrm{\&alpha;}=\sqrt{1-{\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2\mathrm{\&alpha;}}$

$\cos \mathrm{\&beta;}=\frac{{{\mathrm{<figure-callout\; id="2"\; label="x\; n"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">x</figure-callout>}}_n}^2+L^2-{{\mathrm{<figure-callout\; id="2"\; label="y\; n"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">y</figure-callout>}}_n}^2}{{2x}_{n}L},\sin \mathrm{\&beta;}=\sqrt{1-{\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="HDA0000381014040000011.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2\mathrm{\&beta;}}$

cosθ=cos(β-α)=cosαcosβ+sinαsinβ

N judgement constantly driver's head position depart from default stage driver head position apart from S _nfor:

the driver compares the head position when when fatigue state, the head position is with waking state, and the larger degree of fatigue of departure degree is darker, constantly biased from fatigue strength FPD of n judgement _nfor:

In the present embodiment, the selected sampling time window of each fatigue strength characteristic parameter is preferred value, by the lot of experiments analysis, chooses decision.

In the present embodiment, the mathematical model of Nonlinear Mapping relation between the sign fatigue strength characteristic parameter that obtains and driver fatigue state grade is embedded to information process unit; Information process unit is processed the current judgement of driving condition information acquisition fatigue strength characteristic parameter constantly, the driver fatigue state grade of judgement current time in real time; Within the default stage, grip amplitude fatigue strength FGA, corner standard deviation fatigue strength FASD, corner frequency fatigue strength FAF and a biased initial value from fatigue strength FPD are 0.

In the present embodiment, information process unit is hardware unit, be arranged on by digitizing LCD automobile Displaying Meter, the selection window that the driver can provide by the operation push-button on the information process unit hardware unit and digitizing LCD automobile Displaying Meter, according to traffic conditions grade, trackside environmental rating and the road quality classification of current road, regularly early warning indicating mode is set voluntarily.

In the present embodiment, the warning unit is used digitizing LCD automobile Displaying Meter, the monitoring information of real-time display system: driver fatigue state grade, grip, corner, head position, system operation time.When monitoring the driver in fatigue state, carry out voice reminder and light flash and report to the police; When degree of depth fatigue state, carry out voice reminder, light flash, injection irritative gas and vibrating device and report to the police, realize the multifunction type of alarm of the sense of hearing, vision, sense of smell, sense of touch; All there is distribution the warning device of this cover monitoring system a plurality of positions in car, can improve other occupants' safety and remind consciousness, realize multi-facetedization type of alarm; In addition, as the driver, during in degree of depth fatigue state, carry out front vehicles is blown a whistle, front vehicle is opened the external alert mode of two flashing lights, make real-time monitoring system more effectively, more in real time, more extensive.

Claims (5)

1. the driver fatigue real-time monitoring system based on many information, it is characterized in that: described monitoring system is by information acquisition unit, information process unit and warning cell formation;

Described information acquisition unit is provided with: pressure transducer is arranged on the bearing circle top layer, for extracting the grip information of driver to bearing circle; Angular transducer, be arranged on the bearing circle rotary column, for extracting the corner information of bearing circle; Range sensor, be arranged on the pilot set headrest upper, for extracting driver's head position information; Described grip information, corner information and driver's the head position information of take is driving condition information;

Described information process unit, for described driving condition information is processed, obtains the fatigue strength characteristic parameter that characterizes driving condition, with described fatigue strength characteristic parameter judgement driver fatigue state grade;

Described warning unit for carrying out alarm when information process unit judgement driver is in non-waking state.

2. the driver fatigue method of real-time based on many information, is characterized in that, comprises the steps:

Step 1, gather driver's driving condition information, described driving condition information refer to by pressure transducer detects the driver that obtains to the grip information of bearing circle, by angular transducer, detects acquisition bearing circle corner information and detect the driver's of acquisition head position information by range sensor;

Step 2, the described driving condition information of step 1 is processed, obtained the fatigue strength characteristic parameter that characterizes driving condition;

Step 3, by step 1 and the described method of step 2, obtain the sample set that a sample number is N, described sample refers to that calculating by step 2 the fatigue strength characteristic parameter obtained forms with driver fatigue state grade determined by subjective assessment and that described fatigue strength characteristic parameter is corresponding; Build three layers of BP network, utilize described sample set to carry out off-line training to described three layers of BP network, obtain characterizing the mathematical model of Nonlinear Mapping relation between fatigue strength characteristic parameter and driver fatigue state grade;

Step 4, in each judgement constantly, obtain in real time current judgement fatigue strength characteristic parameter constantly by step 1 and the described method of step 2, the input signal that the described current judgement fatigue strength characteristic parameter constantly of take is step 3 gained mathematical model, utilize the current judgement of described mathematical model judgement driver fatigue state grade constantly.

3. the driver fatigue method of real-time based on many information according to claim 2 is characterized in that: in observation process, system be take the driving condition information of 0.2s as cycle Real-time Collection driver; The collection initial period of 30s of take is the default stage, and in the described default stage, the driver is in abnormal driving state, the 30s of take in the default stage after the image data of the 20s initial acquisition data that are each driving condition information; When finishing, the default stage of described 30s enters the monitoring stage, in the described monitoring stage, system is carried out the judgement of a driver fatigue state grade every 2s, the fatigue strength characteristic parameter that each judgement constantly adopts is that different fatigue strength characteristic parameters have the sampling time window of identical or different duration with before this judgement constantly and be that to take this judgement be that image data of the sampling time window of the finish time is calculated and obtained constantly.

4. the driver fatigue method of real-time based on many information according to claim 2 is characterized in that: described fatigue strength characteristic parameter is grip amplitude fatigue strength FGA, corner standard deviation fatigue strength FASD, corner frequency fatigue strength FAF and biased from fatigue strength FPD;

Described grip amplitude fatigue strength FGA obtains as follows:

After the default stage of utilization, the initial acquisition data of 20s calculate driver's grip initial value F ₀for:

in formula

with

be respectively the driver's left hand of 20s after the default stage and the grip average of the right hand; N judgement grip value F constantly within the monitoring stage _nfor:

n=1,2,3..., in formula with

before being respectively n judgement constantly and be that to take the n judgement be driver's left hand of 3s sampling time window of the finish time and the grip average of the right hand constantly; N judges grip amplitude fatigue strength FGA constantly _nfor:

if FGA _nbe less than zero, by FGA _nassignment is zero;

Described corner standard deviation fatigue strength FASD obtains as follows:

Utilizing after the default stage initial acquisition data of 20s to calculate corner standard deviation initial value is SD ₀, within the monitoring stage before n judgement constantly and be that to take the n judgement be that the corner standard deviation of the 5s sampling time window of the finish time is SD constantly _n, n=1,2,3..., n judgement corner standard deviation fatigue strength FASD constantly _nfor:

Described corner frequency fatigue strength FAF obtains as follows:

Utilizing after the default stage initial acquisition data of 20s to calculate corner correction frequency initial value is AN ₀, within the monitoring stage before n judgement constantly and be that to take the n judgement be that the corner correction frequency of the 20s sampling time window of the finish time is AN constantly _n, n=1,2,3..., n judgement corner frequency fatigue strength FAF constantly _nfor:

if FAF _nbe less than zero, by FAF _nassignment is zero;

Described biased obtains as follows from fatigue strength FPD:

Described range sensor is two ultrasonic sensors, described two ultrasonic sensors lay respectively at left side and the right side of pilot set headrest, and after the default stage of utilization, the initial acquisition data of 20s calculate being respectively apart from average of driver's head periphery and described two ultrasonic sensors

with

center, driver's head square section and two ultrasonic sensors apart from initial value x ₀and y ₀be respectively:

with

r is the cross section mean radius of driver's head; The distance value x of n judgement moment center, driver's head square section and two ultrasonic sensors within the monitoring stage _nand y _nbe respectively:

with n=1,2,3..., in formula

with

before being respectively n judgement constantly and be take the n judgement be constantly driver's head of 3s sampling time window of the finish time peripheral with two ultrasonic sensors apart from average;

Order: the center of two ultrasonic sensors is respectively an A and some B, and center, default stage driver head square section is some C; N judgement center, driver's head square section constantly is some D; Have: α is angle between straight line AC and straight line AB; β is angle between straight line AD and straight line AB; θ is the poor of β and α; L is some A and the spacing of putting B; According to the cosine law and trigonometric function formula, have:

$\cos \mathrm{\&alpha;}=\frac{{x_0}^2+L^2-{y_0}^2}{{2x}_{0}L},\sin \mathrm{\&alpha;}=\sqrt{1-{\cos }^2\mathrm{\&alpha;}}$

$\cos \mathrm{\&beta;}=\frac{{x_n}^2+L^2-{y_n}^2}{{2x}_{n}L},\sin \mathrm{\&beta;}=\sqrt{1-{\cos }^2\mathrm{\&beta;}}$

cosθ=cos(β-α)=cosαcosβ+sinαsinβ

N judgement constantly driver's head position depart from default stage driver head position apart from S _nfor: $S_n=\sqrt{{x_0}^2+{x_n}^2-{2x}_0{x}_n\cos \mathrm{\&theta;}};$ Constantly biased from fatigue strength FPD of n judgement _nfor: $\mathrm{FP}{D}_n=\frac{S_n}{L}.$

5. the driver fatigue method of real-time based on many information according to claim 2, it is characterized in that, three layers of BP network that described step 3 builds are: ground floor is input layer, 4 input nodes, consists of, and described 4 input nodes are distinguished 4 corresponding input component x ₁, x ₂, x ₃and x ₄correspond to successively grip amplitude fatigue strength FGA, corner standard deviation fatigue strength FASD, corner frequency fatigue strength FAF and biased from fatigue strength FPD, 4 input components form 1 input vector X and are: X=[x ₁, x ₂, x ₃, x ₄] ^t; The second layer is hidden layer; The 3rd layer is output layer, 3 output nodes, consists of, and described 3 output nodes are distinguished 3 corresponding output component y ₁, y ₂and y ₃mean successively waking state, fatigue state and degree of depth fatigue state, 3 output components form 1 output vector Y and are: Y=[y ₁, y ₂, y ₃] ^t, with Y=[1,0,0] ^tbe characterized by waking state, with Y=[0,1,0] ^tbe characterized by fatigue state, with Y=[0,0,1] ^tbe characterized by degree of depth fatigue state; Described fatigue state and degree of depth fatigue state are non-waking state.

Images