CN106491156A - A kind of fatigue drive of car detection method based on Multi-source Information Fusion - Google Patents
A kind of fatigue drive of car detection method based on Multi-source Information Fusion Download PDFInfo
- Publication number
- CN106491156A CN106491156A CN201610890731.7A CN201610890731A CN106491156A CN 106491156 A CN106491156 A CN 106491156A CN 201610890731 A CN201610890731 A CN 201610890731A CN 106491156 A CN106491156 A CN 106491156A
- Authority
- CN
- China
- Prior art keywords
- signal
- filter
- fatigue
- frequency
- electric capacity
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/02—Measuring pulse or heart rate
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/44—Constructional features of the ultrasonic, sonic or infrasonic diagnostic device
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/48—Diagnostic techniques
- A61B8/488—Diagnostic techniques involving Doppler signals
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/52—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/06—Alarms for ensuring the safety of persons indicating a condition of sleep, e.g. anti-dozing alarms
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Animal Behavior & Ethology (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Cardiology (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
The invention discloses a kind of fatigue drive of car detection method based on Multi-source Information Fusion, comprises the following steps:By persistently being detected to physiological driver's signal intensity and obtaining in real time and store physiological signal installed in the doppler radar unit for driving interior;Variable quantity of the steering wheel angle information in Preset Time is obtained, and data processing is carried out to the physiological signal acquired in working as in the last period at the continuous moment according to variable quantity startup;Judge whether fatigue driving occur according to physiological signal change.Compared with prior art, present invention firstly provides monitoring the degree of fatigue of driver using Doppler measurement technique, while by detecting steering wheel angle information, improving fatigue driving accuracy of measurement by multi-source information process;General doppler radar sensor is operated in continuous wave mode, and respective design multi-stage filter circuit simultaneously, so as to realize non-contact detecting physiology signal, and then fatigue driving is judged on this basis.
Description
Technical field
A kind of the present invention relates to fatigue driving monitoring field, more particularly to fatigue drive of car based on Multi-source Information Fusion
Detection method.
Background technology
Fatigue driving is to cause one of modal reason of vehicle accident in the world.According to WHO's (World Health Organization (WHO))
Report, the people for having more than 1,300,000 every year dies from vehicle accident, have 2,000 ten thousand to 5 thousand ten thousand people because vehicle accident be subjected to non-lethal
Injury.Wherein, about 20% fatal traffic accident is caused by fatigue driving.If it is possible to research and develop one kind certainly
The system of dynamic detection fatigue driving, it is possible to avoid substantial amounts of death by accident.
In order to prevent the generation of this kind of accident, effectively the system of monitoring fatigue driving be necessary, can be in thing
Therefore driver is reminded before occurring, such that it is able to avoid or reduce the generation of such vehicle accident.Driving fatigue is mental, physical
The technology fatigue for simultaneously participating in.Due to driving human action repeatedly, continuously, and the number of times for repeating is too many so as to physiology, psychology
Upper there is certain change, occurs driving the low phenomenon of function, is mainly shown as distractibility, dozes off, and the visual field narrows, and believes
Breath leakage sees that reaction and judgement is blunt, and driver behavior is slipped up or completely loses driving ability.
In existing research, many research worker all think that following physiological signal can be used to detection fatigue, including:Electroencephalogram
(EEG), electrocardiogram (ECG), electro-oculogram (EOG), breath signal and electrodermal activity (EDA) etc..Test is proved, using physiology
The reliability and degree of accuracy of the degree of fatigue of signal detection driver is very high, because it can really reflect driver
The situation of body interior.Meanwhile, physiological signal fatigue early stage will change, therefore using physiological signal be inspection
Survey the most suitable method of fatigue driving and error rate be low, make in this way can timely alerting drivers, so as to reduce
Many road traffic accidents.However, prior art physiological detection is generally by the way of contact measurement, these technology need
Adhesive electrodes with driver, and this is also unpractical while typically dislike by driver.
In order to overcome above-mentioned technological deficiency, Doppler radar technique is realized that contactless bio-signal acquisition becomes ability
Domain study hotspot.Doppler radar, also known as pulse Doppler radar, are usually operated at pulse-triggered pattern, are that a kind of utilization is more
General Le effect is detecting the position of moving target and the radar of speed of related movement.In prior art, Doppler radar is widely used in
Military field and civil area, such as airborne early warning, navigation, missile guidance, Satellite Tracking, battle reconnaissance, target range measurement, weapon
Etc. military aspect, and human body sensing, gate control system, test the speed range finding etc. civil area.However, the spy due to bio-signal acquisition
Different property, it is difficult to which general for prior art doppler radar module is directly applied to bio-signal acquisition;Because breathing and heart beating are believed
Number extremely faint, it is easy to be submerged in the noise and clutter of radar, using prior art Doppler radar routine application circuit
The contactless detection of the vital signss such as the breathing to human body and heart beating cannot be realized.Therefore, those skilled in the art are generally led to
The accuracy of identification and sensitivity for crossing improvement radar reaches application requirement, this considerably increases and realizes difficulty, while on cost
It is greatly improved.
Therefore, for drawbacks described above present in currently available technology, it is necessary in fact to be studied, to provide a kind of scheme,
Solve defect present in prior art.
Content of the invention
In view of this, it is necessory to provide a kind of fatigue drive of car detection method based on Multi-source Information Fusion, first
Propose to realize contactless fatigue driving detection using Doppler radar, while by detecting that steering wheel angle information improves fatigue
Drive certainty of measurement.
In order to overcome the defect of prior art, technical scheme is as follows:
A kind of fatigue drive of car detection method based on Multi-source Information Fusion, comprises the following steps:
Step S1:By the doppler radar unit installed in driving interior persistently to physiological driver's signal intensity
Detected and obtain in real time and store physiological signal;
Step S2:Variable quantity of the steering wheel angle information in Preset Time is obtained, and is started to working as according to the variable quantity
Acquired physiological signal in the last period at the continuous moment carries out data processing;
Step S3:Judge whether fatigue driving occur according to physiological signal change.
Preferably, in step s3, judge whether fatigue driving occur by the way of neural network machine study, its
In, the data template of various fatigue state grades is prestored in neural network machine identification module, by the physiology letter of collection
Number input neural network machine identification module carries out data analysiss so as to judging fatigue state grade.
Preferably, neural network machine identification module adopts extreme learning machine model.
Preferably, the data template of various fatigue state grades is by the breath signal for being obtained and heartbeat signal to be input into
Neutral net learning is determined based on the expert analysis mode method of facial video.
Preferably, in step sl, further comprising the steps:
Step S11:Continuous wave radar signal is launched to torso model by doppler radar sensor;
Step S12:Echo-signal and transmitting concussion frequency signal are carried out obtaining reaction human body after Frequency mixing processing detection
Breathing and the low frequency signal of heart beating change;
Step S13:Impedance matching is carried out to doppler radar sensor outfan and filters the direct current in low frequency signal point
Amount;
Step S14:Signal after processing through step 3 is carried out signal amplification;
Step S15:Process is filtered to its input signal by the band filter of 0.1Hz-10Hz;
Step S16:Frequency filtering is carried out using digital filtering technique prize through the signal after the process of step S25 to exhale so as to obtain
Inhale signal and heartbeat signal;
In step S3, the change according to the breath signal and heartbeat signal judges whether driver is driven in fatigue
Sail state.
Preferably, in step S15, by quadravalence Butterworth LPF and second order Butterworth high-pass filtering
Band filter realized by device.
Preferably, in step S16, using FIR filter, iir filter or zero phase iir filter in appoint
A kind of separation for realizing breath signal and heartbeat signal.
Preferably, zero phase iir filter realizes that step is as follows:
Step S161:Feature according to breath signal and heartbeat signal separately designs breath signal iir filter and heart beating
Signal iir filter;
Step S162:Input signal is carried out signal sampling and is stored as digital signal sequences;
Step S163:The digital signal sequences are separately input to breath signal iir filter and heartbeat signal IIR filtering
Device carries out first time Filtering Processing;
Step S164:The signal exported through above-mentioned first time Filtering Processing is executed the upset of first time time domain;
Step S165:Step S64 output signal is again inputted into breath signal iir filter and heartbeat signal IIR filters
Ripple device carries out second Filtering Processing;
Step S166:The signal exported through above-mentioned second Filtering Processing is executed second time domain upset, so as to obtain
Filtered breath signal and heartbeat signal;
Step S167:Frequency spectrum is carried out obtaining respectively to filtered breath signal and heartbeat signal after FFT so as to reality
Existing breath signal and the separation of heartbeat signal.
Preferably, in step step S11, the doppler radar sensor adopts working frequency range for 10.525GHz's
Microwave Doppler radar detedtor probe sensor HB100 modules.
Preferably, in step s 13, impedance is carried out to doppler radar sensor outfan using voltage follower
Match somebody with somebody, adopt band connection frequency that the DC component in low frequency signal is filtered for the passive RC filter of 0.1Hz-150Hz.
Compared with prior art, technical scheme has following technique effect:
(1) present invention firstly provides monitoring the degree of fatigue of driver using Doppler measurement technique, while by detection
Steering wheel angle information, improves fatigue driving certainty of measurement by multi-source information process.
(2) general doppler radar sensor is operated in continuous wave mode, and respective design multi-stage filter circuit, so as to
Non-contact detecting physiology signal is realized, and then judges fatigue driving on this basis.Wherein, active filter adopts Bart
Butterworth wave filter, Butterworth filter passable frequency response curve are flat, decline in suppressed frequency band slow, it is to avoid signal loses
Very, the amplification of signal is can achieve while filtering, the signal to noise ratio of signal is improved, and is realized that signal is undistorted and is amplified filtering.
(3) digital filter isolates breathing and heartbeat signal using zero phase iir filter, reduces the same of operand
When, the phase distortion of signal is eliminated, is realized that patient's physiological change and monitoring show synchronous, is improved the real-time of monitoring device.
Description of the drawings
Fig. 1 is radar echo signal detection torso model expansion model.
Fig. 2 is FB(flow block) of the present invention based on the contactless method for detecting fatigue driving of Multi-source Information Fusion.
Fig. 3 is extreme learning machine illustraton of model.
Algorithm flow block diagrams of the Fig. 4 for extreme learning machine model.
Fig. 5 is the oscillogram (level of fatigue 0) of normal driving physiological signal and hand-wheel signal.
Fig. 6 is the oscillogram (level of fatigue 1) for driving physiological signal and hand-wheel signal under alert status.
Fig. 7 is the oscillogram (level of fatigue 2) for driving physiological signal and hand-wheel signal under drowsy state.
Fig. 8 is the oscillogram (level of fatigue 3) for driving physiological signal and hand-wheel signal under notable drowsy state.
Fig. 9 is the oscillogram (level of fatigue 4) for driving physiological signal and hand-wheel signal under extreme drowsy state.
FB(flow block)s of the Figure 10 for Radar Signal Processing.
Flowcharts of the Figure 11 for zero phase iir filter.
Figure 12 is the system block diagram for realizing contactless bio-signal acquisition method of the invention.
Figure 13 is the theory diagram of doppler radar unit in the present invention.
Figure 14 is the circuit theory diagrams of radar power supply in power module.
Figure 15 is the circuit theory diagrams of amplifier power supply in power module.
Figure 16 is the circuit theory diagrams of digital power in power module.
Figure 17 is the circuit theory diagrams of ADC reference power supplies in power module.
Figure 18 is the circuit theory diagrams of signal pre-processing module of the present invention.
Figure 19 is a kind of circuit theory diagrams of embodiment of difference amplifier of the present invention.
Figure 20 is a kind of circuit theory diagrams of embodiment of active band-pass filter of the present invention.
Circuit theory diagrams of the Figure 21 for voltage movement circuit.
Circuit theory diagrams of the Figure 22 for analog-digital converter.
Figure 23 is that FIR and IIR filtering separates the contrast of breath signal time domain.
Figure 24 is that FIR and IIR filtering separates the contrast of breath signal frequency domain.
Figure 25 is zero-phase filtering breath signal time-domain diagram.
Figure 26 is zero-phase filtering heartbeat signal time-domain diagram.
Figure 27 breath signals and heartbeat signal separation frequency domain figure.
Specific examples below will further illustrate the present invention in conjunction with above-mentioned accompanying drawing.
Specific embodiment
The fatigue drive of car detection method based on Multi-source Information Fusion that the present invention is provided is made below with reference to accompanying drawing
Further illustrate.
Research finds, driver's steering wheel rotation generally by a small margin is constantly adjusting the horizontal shifting of vehicle under normal circumstances
Dynamic position travels the center in track all the time to keep vehicle.Micro- correction data when driver is drowsy, on steering wheel
The data that can be less than during normal driving.Do not consider track change, only consider (SWM, the 0.5-5 degree of rotation by a small margin of steering wheel
Between) in the case of, compared with normal driver, the number of times of steering wheel rotation is less for the driver of fatigue.So, driver's
Degree of fatigue can be reacted by the angle of wheel steering SWM (steering wheel movement) to a certain extent
Out.
Research also finds simultaneously, and in fatigue state, which breathes human body and heart rate signal can all decline, and therefore passes through
Breathing and heart rate signal can accurately judge the fatigue state of human body.In order to overcome prior art to detect using touch sensor
The technological deficiency of physiological signal, applicant propose for Doppler radar general for prior art to be applied to contactless fatigue first
Drive detection field.
Doppler radar be widely used in airborne early warning, navigation, missile guidance, Satellite Tracking, battle reconnaissance, target range measurement,
The military fields such as weapon.Its operation principle can be expressed as follows:When the impulse wave of one fixed frequency of radar emission is to empty scanning, such as
Moving target is run into, the frequency of echo is poor with the frequency frequency of occurrences of transmitted wave, referred to as Doppler frequency.According to Doppler frequency
Size, diametrically movement velocity of the target to radar can be measured;According to transmitting pulse and the time difference for receiving, can measure
The distance of target.Therefore, the Doppler radar of military domain is usually operated at pulse mode, detects work by detecting difference on the frequency
Moving-target.In prior art, Doppler radar also has the application in civil area, such as, using Doppler radar (Doppler
Radar) the microwave detector for moving object HB100 microwave modules of principle design, are widely used in automatic door control switch, safety
The places such as crime prevention system, the automatic video recording control system of ATM Automatic Teller Machines, train automatic signal.However, such Doppler
During application of the radar in civil area, frequency is detected after typically directly amplify output signal, then according to frequency size
Obtain and speculate human motion speed.
Doppler radar sensor can eliminate the shadow of particular medium (such as cloth, silk etc.) in specific distance range
Ring, detect the fine motion change of torso model, therefrom get physiological parameter information, realize the detection of contactless physiological signal.
Contactless monitoring system overcomes the shortcoming of traditional physiological monitoring system, simple with noncontact, remote monitoring, operation
The advantages of, increasing concern has been obtained in fields such as clinical medicine, disaster medicine, military medicine, city anti-terrorisms, with wide
General application prospect.However, doppler radar sensor is realized contactless physiology in research by those skilled in the art
During signal detection, it is typically directed to design high accuracy of identification and highly sensitive doppler radar sensor, reality has been significantly greatly increased
Existing difficulty.
On the basis of existing technology, by repeatedly theoretical and experimental study, applicant has found that continuous wave radar is with human body
Thoracic cavity as detection target, through chest cavity movement return radar emission signal can produce phase-modulation, the radar for receiving return
Ripple signal extracts the phase information being associated with chest cavity movement, according to phase information from demodulating information through phase demodulating
The situation of change of the breathing and heart beating of change reflection tester.
Referring to Fig. 1, radar echo signal detection torso model expansion model is shown, it is now assumed that radar emission signal T (t)
For
T (t)=cos [2 π f0t+Φ(t)] (1)
F in formula0It is radar emission signal frequency, Φ (t) is phase noise.
If chest cavity movement amplitude is x (t), radar sensor to human body distance is d0, transmitting radar signal to thoracic wall away from
From for d (t), then round trip delay time isDue to the chest cavity movement cycleThen adjust through radar reflection
Reception signal R (t) after system is:
Reception echo-signal R (t) is multiplied with radar emission signal T (t) after low-pass filtering and demodulates modulated signal, obtains
Taking baseband signal is:
In formulaIt is residual phase noise,It is that radar and human body spacing determine
Intrinsic phase shift.When θ isOdd-multiple when, x (t) < < λ can be obtained:
Wherein ΔΦ (t) is the DC component for fixing target generation, can obtain chest displacement x (t) by formula (4) and export with base band
Amplitude linear.From the foregoing, doppler radar sensor can obtain the signal of torso model vibration, in signal
Comprising breathing, heart beating and human body disturbance, by certain signal processing method, human body respiration can be extracted and heart beating is special
Reference number.Research shows that during normal driving and fatigue driving, the physiological signal such as breathing and heart rate bright driver can occur
Aobvious change.Therefore, fatigue driving just can accurately be judged by detecting physiological signal.But cardiopulmonary information is carried out with radar
Real-time detection and process have the problem of the following aspects:
(1) the mobile interference of human body in driving.During normal driving, driver needs constantly adjustment direction at any time
Disk, now, the limbs such as human arm are also constantly being moved.As human body is in narrow and small driving space, mobile scope compared with
Little, and translational speed is slower, then and the frequency of the signal comprising human body disturbance in the signal of detections of radar is just very low, connects very much
Closely even with breathing heart rate signal frequency in same frequency range, this interference can cause breathing and heart rate signal to be submerged, it is impossible to
Detect breathing heart rate signal.
(2) process signal time span exists and limits.On the one hand, along with the change of physiological statuss, hrv parameter is breathed
Also can change accordingly, need timely to detect this change.When the signal for intercepting the long period is processed,
The shorter change information of physiological parameter is difficult to be detected, for real-time detection and prediction fatigue driving very unfavorable.Separately
On the one hand theoretical according to time frequency analysis, the resolution of the data of short time on frequency domain is inevitable very low, is unfavorable for breathing and heart rate
The separation of signal.To sum up need from the aspect of two to choose suitable data treated length, to improve the real-time of detection and improve letter
The resolution of number frequency.
(3) radar signal is different from conventional electrocardio and pulse wave signal detection method, and what it detected is heart beating and breathing
Compound signal.And respiratory movement is stronger than heart beating in amplitude so that heart rate signal is not readily separated extraction.And breathe and heart beating
The fine motion for causing can be overlapped in body surface space, be also easy to produce frequency domain intermodulation.
In order to solve above-mentioned technical problem, referring to Fig. 2, a kind of automobile based on Multi-source Information Fusion of the present invention is shown tired
Please the FB(flow block) of detection method is sailed, is comprised the following steps:
Step S1:By the doppler radar unit installed in driving interior persistently to physiological driver's signal intensity
Detected and obtain in real time and store physiological signal;
Step S2:Variable quantity of the steering wheel angle information in Preset Time is obtained, and is started to working as according to the variable quantity
Acquired physiological signal in the last period at the continuous moment carries out data processing;
Step S3:Judge whether fatigue driving occur according to physiological signal change.
Using above-mentioned technical proposal, by steering wheel angle information and the use processing of physiological driver's signal, from
And effectively overcome the mobile interference of human body in driving;Meanwhile, by real-time storage physiological signal, can when data analysiss are carried out
Choosing suitable data length in time carries out data processing, so as to improve data processing speed.
Although breath signal of the theoretical proof human body under fatigue state and non-fatigue state and heartbeat signal have significantly
Difference, but the research of fatigue driving decision algorithm is carried out under true driving environment with very big challenge, also only recognize
The fatigue driving decision algorithm of high precision just has actual application value.
In prior art, using the expert analysis mode method based on facial video, the method is driver most practical at present
Fatigue state evaluation methodology.The method is by one group of trained expert according to the facial expression of driver and head pose to which
Fatigue state is scored.Concrete operation step is:It is video segment that human face's video slicing will be driven;Several scoring expert's roots
Rub one's eyes according to driver, scratch face, yawn, closed-eye time, the adjustment fatigue such as posture are characterized, video segment is carried out by random sequences
Marking, the tired score of the average of several expert analysis modes as this section of video.The system is carried out to driver's fatigue degree's grade
Classification, specific requirement are as shown in table 1 below:
Table 1:Driver's fatigue degree's grade classification and its description
In order to realize accurate fatigue driving detection, the present invention is adopting neural network machine identification module, using nerve net
The mode of network machine learning judges whether fatigue driving occur, wherein, prestores each in neural network machine identification module
The data template of fatigue state grade is planted, the physiological signal input neural network machine identification module of collection is carried out data analysiss
So as to judge fatigue state grade.Neural network machine identification module can be arranged in MCU module or vehicle-mounted computer.Make
With front, need first to train neural network machine identification module.In the present invention, first by the expert analysis mode based on facial video
Breath signal and heartbeat signal input neutral net learning that classification is completed, so as in neural network machine identification module
Determine the data model of each level of fatigue.So as to when actual fatigue driving judges, can according to the breath signal of input and
Heartbeat signal judges level of fatigue.
In the present invention, neural network machine identification module show pole using extreme learning machine illustraton of model referring to Fig. 3
Limit learning machine illustraton of model, the data template of various fatigue state grades is by the breath signal for being obtained and heartbeat signal to be input into
Neutral net learning is determined based on the expert analysis mode method of facial video.Referring to Fig. 4, extreme learning machine model is shown
Algorithm flow, the rudimentary algorithm of extreme learning machine model are as described below:
N:Training sample sum
Hiding layer unit number
n、m:The dimension (i.e. the length of input and output vector) of input and output layer
(xj,tj), j=1,2 ..., N:Training sample, wherein,
xj=(xj1,xj2,...,xjn)T∈Rn,tj=(tj1,tj2,...,tjn)T∈Rm
All output vectors are got up by row spelling, overall output matrix is obtained:
oj, j=1,2 ..., N:With mark tjCorresponding reality output vector;
The corresponding vector of i-th row of the weight matrix between input layer and hidden layer, wherein W
wi=(wi1,wi2,...,win)T
Represent the weight vector of i-th unit of connection hidden layer and input block;
Bias vector, biRepresent the threshold value of i-th hidden unit;
The corresponding vectorial β of i-th row of the weight matrix between hidden layer and output layer, wherein βi=
(βi1,βi2,...,βim)T
Represent that the weight vector of i-th unit of connection hidden layer and output unit, matrix β can be written as piecemeal shape by row
Formula:
G (x) is activation primitive.
Mathematically, the model of the SLFNs of standard is
Wherein wi·xjRepresent wiAnd xjInner product.
Make model above zero error approaches above-mentioned N number of sample, then
There is W, β and b so that
Represented using matrix form, above-mentioned expression formula can be reduced to
H β=T
In general SLFNs learning algorithms, the bias vector b for being input into weights W and hidden unit needs constantly to adjust by iteration
Whole refreshing, when learning algorithm starts, the value of any given W and b calculates H with which and makes its holding constant, so, we
Only need to determine that parameter beta just can be with.Therefore, when W and b is fixed, the least square solution for seeking formula equation is equivalent toI.e.
Correlation theorem according to Moore-Penrose generalized inverse matrix and LS solution of the least norm is obtained
β=H+T
Given training sample setActivation primitive g (x), hidden unit number
ELM algorithms realize that process can be summarized as:First, input weights and the threshold value of network are randomly assigned;Secondly, meter
Calculate hidden layer neuron output matrix;Finally, according to β=H+T calculates output weight matrix.According to above reasoning process, we can
Know that therefore ELM iterative steps not during traditional neural computing considerably reduce the training time.
In order to verify the actual techniques effect of the present invention, 8 tested drivers have been recruited in the experiment test stage, the age exists
Between 22-60 year.Healthy, audition is normal, without anerythrochloropsia.Ensure before the experiments abundance sleep, test front 24 little
When interior do not take any zest article.Every driver carries out drive simulation experiment twice daily:All drivers are every
Its same time is tested, and notifies driver's preparing experiment before experiment, it is desirable to which driver before experiment forbids driving in 24 hours
Sail people drink, tea and coffee.Experimental selection at noon 12:00 and evening 20:00 proceeds by, each continue two hours or so,
This time period people is typically easiest to, in state of feeling sleepy, therefore can to observe driver in two of short duration hours
Tired whole process from regaining consciousness to being absorbed in.Require to keep environment quiet in experimentation, the photographic head in driving cabin is used
Have recorded driver's facial expression information, image acquisition rates are 30Hz, and the data collecting system record of testing stand drives
People's breath signal, heart rate signal, human mobile signal, steering wheel small angle tower signal, steering wheel big corner signal, sample frequency are equal
For 50Hz.
Video signal and MCU collection signal synchronizations, are 30 seconds by per section of segmentation length of video signal of collection, using expert
Standards of grading are given a mark to the degree of fatigue of driver, used as the physiological signal judging basis of radar collection.Synchronous physiology
Signal is divided into five class signals according to based on facial video experts standards of grading, and correspondingly 5 class degree of fatigues use numeral mark respectively
For 0-4, it is 30 seconds per segment data length, the breathing and heartbeat signal under every kind of degree of fatigue gathers 100 groups of data respectively, extracts
Go out the breath signal mean frequency value of every segment data, breath signal frequency mean-square value, breath signal amplitude equalizing value, breath signal amplitude
Mean-square value, heartbeat signal mean frequency value, heartbeat signal amplitude equalizing value, heartbeat signal amplitude mean-square value are used as extreme learning machine algorithm
Input sample, by sample learning training until meeting in the range of error requirements, obtain exporting weights.In Single Chip Microcomputer (SCM) system pair
Extreme learning machine enters line algorithm realization, determines training output weights, in driver's test, gathers radar signal, be filtered
Signal characteristic is extracted, is classified by extreme learning machine, for classification results, give driver different promptings.Fig. 5 to Fig. 9 is
Collection driver breathing tired in various degree and heartbeat signal variation diagram, as seen from the figure, are recognized by neural network machine
Can be good at distinguishing degree of fatigues at different levels.
However, the thoracic cavity fine motion displacement scope that human normal breathing and heart beating cause is only 4-15mm, and prior art
In the application in dual-use field, the resolution of mobile object is at least 0.1 meter to middle doppler radar module;Meanwhile, just
The breathing of ordinary person and palmic rate are respectively 0.15~0.4Hz and 0.83~1.5Hz, frequency spectrum closely, in the time domain it is difficult to
Breath signal and heartbeat signal are distinguished.By formula (4) it is recognized that while the thoracic cavity fine motion that human normal is breathed and heart beating causes
Displacement scope is less, as long as choosing suitable Doppler radar operating frequency, can be good at detecting thoracic cavity micro-tremor signal;Though
So the frequency of breath signal and heartbeat signal closely, as long as selecting suitable sample frequency, can still distinguish breath signal
And heartbeat signal, as weak output signal and frequency separation are not it is obvious that how filtering interference signals extract useful data signal
It is the key for solving present invention problem.
In order to solve above-mentioned technical problem, referring to Figure 10, the FB(flow block) of Radar Signal Processing in the present invention is shown, wrapped
Include following steps:
Step S11:Continuous wave radar signal is launched to torso model by doppler radar sensor;
Step S12:Echo-signal and transmitting concussion frequency signal are carried out obtaining reaction human body after Frequency mixing processing detection
Breathing and the low frequency signal of heart beating change;
Step S13:Impedance matching is carried out to doppler radar sensor outfan and filters the direct current in low frequency signal point
Amount;
Step S14:Signal amplification will be carried out through the signal after the process of step S3;
Step S15:Process is filtered to its input signal by the band filter of 0.1Hz-10Hz;
Step S16:Frequency filtering will be carried out through the signal after the process of step S5 using digital filtering technique to exhale so as to obtain
Inhale signal and heartbeat signal.
Wherein, in step s 11, the doppler radar sensor adopts working frequency range many for the microwave of 10.525GHz
General Le radar detedtor probe sensor HB100 modules.In prior art, Doppler radar operating frequency range be 2~
75GHz, the present invention with reference to radar resolution, penetrate the factors such as barrier ability, volume size and power consumption, choose operating frequency
Doppler radar sensor for 10.525GHz.HB100 microwave modules are former using Doppler radar (Doppler Radar)
The microwave detector for moving object of reason design, is mainly used in automatic door control switch, safety and protection system, ATM automated tellers
The places such as the automatic video recording control system of machine, train automatic signal.HB100 is the 10.525GHz microwave Doppler thunders of standard
Reach detector, internal by FET medium DRO microwaves concussion source (10.525GHz), power divider, transmitting antenna, reception antenna,
The circuits such as frequency mixer, cymoscope are constituted, and which is 35mA in continuous direct current supply MODE of operation electric current, and gross output is less than
15mW.Transmitting antenna is outwardly directed transmitting microwave, is reflected when running into object, and echo is received by reception antenna, then arrives mixed
Clutch is mixed with the wave of oscillation, and the low frequency signal after mixing, detection has reacted the speed of object movement.General using prior art
Detecting module, greatly reduces cost and development difficulty.Prior art, generally detects human motion using HB100 modules,
The frequency of the signal is directly amplified and is detected to the low frequency signal which exports, so as to calculate human body according to frequency values
Translational speed, usual investigative range is more than 20 meters.However, in the application of the present invention, the breast that human normal breathing and heart beating cause
Fine motion displacement scope in chamber is only 4-15mm, the intensity of various noise signals considerably beyond useful signal, therefore, using tradition
HB100 module applications method cannot detect physiological signal.Therefore the present invention proposes a kind of suitable for bio-signal acquisition
Three-level filtering method, so as to realizing the detection of breath signal and heartbeat signal, the method application circuit described in detail below sets
Meter principle.
In step s 13, band connection frequency is adopted to filter for the passive RC filter of 0.1Hz-150Hz straight in low frequency signal
Flow component.
In step s 13, impedance matching is carried out to doppler radar sensor outfan using voltage follower.Voltage with
It is used for carrying out input signal voltage follow with device, passive filter is used for filtering the DC component in input signal;Human body breast
Fine motion change in chamber causes doppler radar sensor output signal change amplitude range 1-20mV, with amplitude is low, noise is big, band
The features such as load capacity difference, voltage follow is carried out on input signal and eliminates output impedance impact, improve driving force;Radar signal
For radiofrequency signal, the spurious signal in space is excessive, the amplifier saturation of rear end can be caused even to damage, in order to prevent due to direct current
Component causes amplifier saturation, filters DC component using passive filter.
Further, input signal is amplified by active band-pass filter and eliminates differential mode noise;Radar signal
After difference amplifier, common mode disturbances noise can be eliminated well.But also substantial portion of noise is with difference
The form of mould enters late-class circuit.These noises are dry comprising power supply noise when starting, DC baseline drift noise and power frequency
Disturb noise.So need to select suitable wave filter to be filtered the signal after primary amplification, in order to overcome passive filtering electricity
The shortcoming of road consumption signal energy, using the active power filtering being made up of amplifier and resistance-capacitance network, improves filtering performance.Relative
For passive filtering, due to the addition for having amplifier, active filter can not only carry out power back-off, moreover it is possible to while filtering
Signal is amplified, while amplifier can also play the effect of buffering and isolation.In conjunction with breathing and heartbeat signal frequency and
Human body forcing frequency, the present invention adopt active low-pass filter and high pass filter to constitute band logical of the frequency for 0.1Hz-10Hz
Wave filter.
According to wave filter amplitude-frequency and the difference of phase-frequency characteristic, it is broadly divided into according to active filter transmission characteristic following several
Class:
Butterworth filter:Within passband, the amplitude of amplitude frequency curve is most flat, by passband to stopband attenuation steepness compared with
Slow, phase-frequency characteristic is nonlinear, is most flat amplitude filter.
Chebyshev filter:In passband, with equal ripple.Butterworth of the cut-off frequency decay steepness ratio with exponent number
The steeper phase response of characteristic be non-linear, but than than Butterworth for poor.
Bessel filter:Time-delay characteristics are most flat, and amplitude-frequency characteristic most flat region is less, delay from passband to stopband attenuation
Slowly.The amplitude-frequency characteristic of Bessel filter is all poorer than Butterworth or Chebyshev filter.
Elliptic function filter:There is equal ripple in passband and stopband.Elliptic function filter is compared with other classes
The wave filter of type has the cut-off frequency decay steepness of steepest.But its time-delay characteristics are good not as first three.
Design requirement wave filter amplitude frequency curve of the present invention is as flat as possible in passband, and special with good intermediate zone
Property.On the basis of more above-mentioned wave filter actual performance, final choice, butterworth filter are all-pole filters,
So in n rank all-pole filters, when amplitude-frequency characteristic of the opinion at w=0, then Butterworth filter is most straight, therefore bar
Special Butterworth wave filter is referred to as maximally flat wave filter and there is maximally-flat in passband, and the phase characteristic ratio of Butterworth filter is same
The comparison that the linear relationship of all good phase shift of the Chebyshev of exponent number, anti-Chebyshev and elliptic function filter and frequency affects
Little, it is possible to achieve preferably signal filtering effect and less signal attenuation, it is adaptable to noise in radar breathing heartbeat signal
Remove.
In step S15, constituted using second order butterworth high pass filter and quadravalence Butterworth LPF
Wave filter group is amplified filtering to signal.
In step s 16, the separation of breath signal and heartbeat signal is realized using digital filtering technique, can adopt FIR
Any one in wave filter, iir filter or zero phase iir filter.FIR filter, the design of iir filter, will be
Realize describing in detail in the system of the inventive method, will not be described here.
In order to overcome the defect of FIR filter and iir filter, the present invention to propose a kind of zero phase iir filter, ginseng
See Figure 11, show the flowchart of zero phase iir filter, specifically include following steps:
Step S161:Feature according to breath signal and heartbeat signal separately designs breath signal iir filter and heart beating
Signal iir filter;
Step S162:Input signal is carried out signal sampling and is stored as digital signal sequences;
Step S163:The digital signal sequences are separately input to breath signal iir filter and heartbeat signal IIR filtering
Device carries out first time Filtering Processing;
Step S164:The signal exported through above-mentioned first time Filtering Processing is executed the upset of first time time domain;
Step S165:Step S64 output signal is again inputted into breath signal iir filter and heartbeat signal IIR filters
Ripple device carries out second Filtering Processing;
Step S166:The signal exported through above-mentioned second Filtering Processing is executed second time domain upset, so as to obtain
Filtered breath signal and heartbeat signal;
Step S167:Frequency spectrum is carried out obtaining respectively to filtered breath signal and heartbeat signal after FFT so as to reality
Existing breath signal and the separation of heartbeat signal.
In a preferred embodiment, the digital filter of above-mentioned employing is realized by program.
The lower system architecture for realizing the inventive method described in detail below, referring to Figure 12 and Figure 13, show and realizes this
The system block diagram of bright contactless bio-signal acquisition method, the system include doppler radar sensor, power module, signal
Pretreatment module, difference amplifier, active band-pass filter, breathing and heartbeat signal separation module and MCU module, wherein, electricity
Source module is used for system power supply;Doppler radar sensor is used for launching continuous wave radar signal to torso model and receiving echo
Signal processed after output-response human body respiration and heart beating change low frequency signal, low frequency signal is successively through Signal Pretreatment mould
After block, difference amplifier, active band-pass filter, breathing and heartbeat signal separation module and MCU module signal processing, MCU moulds
Block obtains human body respiration signal and heartbeat signal.
In order to improve the accuracy of detection of system, the undulatory property for taking into full account voltage is needed in power module design, and
Interference of the powerful electric current to system during starting.Therefore need to choose Width funtion input voltage stabilizing chip, radar signal output is very faint, electricity
Power supply ripple and noise problem is paid particular attention in source module, as system not only includes digital circuits section, is also included
The analog portions such as A/D conversions, low level signal amplification, need to isolate digital power and analog power, therefore separately design radar power supply, fortune
Discharge source, digital power and ADC reference power supplies.
Referring to Figure 14, show the circuit theory diagrams of radar power supply in power module, including the first power interface P1, first
Electric fuse F1, the first transient diode TVS1, the first diode D1, the 6th electrochemical capacitor C6, the 7th electric capacity C7, the second electric capacity
C2, the 5th power supply chip U5, the 14th electric capacity C14, the 15th tantalum electric capacity C15, wherein, the crus secunda of power interface P1 and first
One end of electric fuse F1 is connected, one end and the first diode of the other end of the first electric fuse and the first transient diode TVS1
The anode of D1 is connected, the anode of the negative terminal and the 6th electrochemical capacitor C6 of the first diode D1, one end of the 7th electric capacity C7, the 5th
The 5th pin of power supply chip, the 8th pin are connected, and first pin of power interface P1 is another with the first transient diode TVS1's
End, the negative terminal of the 6th electrochemical capacitor C6, the other end of the 7th electric capacity C7, the 6th pin of the 5th power supply chip U5 and the 7th pin
And three-prong is connected with simulation ground terminal jointly, the 4th pipe of first pin and the 5th power supply chip U5 of the second electric capacity C2
Foot is connected, and the other end of the second electric capacity C2 is connected with first pin and the second pin of the 5th power supply chip U5, and the tenth
First foot of four electric capacity C14 is connected with the anode of the 15th tantalum electric capacity C15, first pin and second of the 5th power supply chip U5
Pin is connected, and the other end of the 14th electric capacity C14 is connected with simulation ground terminal jointly with the negative terminal of the 15th tantalum electric capacity C15.
In foregoing circuit, the 5th power supply chip U5 sampling LT1763CS8-5 export 5V power supplys and give radar chip power supply,
The chip is a low noise, low voltage difference micropower regulator.It is 20 μ VRMS in 10Hz-100KHz output noises, Width funtion is defeated
Enter scope 1.8V to 20V, with 1 μ A of low-down standby current, internal have excessively stream and overheat protective function, with Switching Power Supply
Compare, with Ripple Noise little the characteristics of.Overcurrent protection is carried out to circuit using MF-R09009 at power interface end, and is held
At mouthful, a TVS pipe in parallel, plays a very good protection to power supply overvoltage pulse, and power end one diode of series connection prevents electricity
Source reversal connection, shields to rear class whole system.For reducing ripple interference, in each power supply chip plus a high frequency decoupling
Electric capacity, adds a high-frequency bypass capacitor beside each electrochemical capacitor.
As single supply amplifier of powering can reduce low frequency characteristic, single supply amplifier input/output signal scope can reduce,
Amplifier becomes more sensitive to internal and external error source, while in low pressure single supply device, gain accuracy can also drop
Low, the present invention considers and passes through experimental verification, and final sampling selects dual power supply to power to amplifier.Referring to Figure 15, it show
The circuit theory diagrams of amplifier power supply in power module, including the 13rd electric capacity C13, the 3rd power supply chip U3, the 18th electric capacity
C18, the 4th resistance R4, the 5th resistance R5, the 16th electric capacity C16, the 17th electric capacity C17, the 19th electric capacity C19, the 6th resistance
R6,3rd resistor R3, the 20th electrochemical capacitor C20, the 21st electric capacity C21, first resistor R1, the 4th power supply chip U4,
One electric capacity C1, the second inductance L2, the second diode D2, the 11st electrochemical capacitor C11, the 12nd electric capacity C12, wherein, the 13rd
First foot of electric capacity C13 one end and the 3rd power supply chip U3, the 3rd foot, the 5th foot are connected, one end of the 18th electric capacity and the
4th foot of three power supply chip U3 is connected, one end of the 16th electric capacity C16 and one end of the 17th electric capacity C17, the 3rd power supply
Tenth foot of chip, the 11st foot are connected, and one end of the 4th resistance R4 is connected with the 9th foot of the 3rd power supply chip, and the 5th
One end of resistance R5 is connected with the other end of the 4th resistance R4, the octal of the 3rd power supply chip U3, the 13rd electric capacity C13's
The other end of the other end and the 18th electric capacity C18, the other end of the 5th resistance R5, the other end of the 16th electric capacity C16, the 17th
The other end of electric capacity C17 is common to be connected with simulation ground;One end of 19th electric capacity C19 one end and 3rd resistor R3, the 6th electricity
Resistance one end of R6, the 3rd feet of the 4th power supply chip U4 are connected, the other end of 3rd resistor R3 and the 20th electrochemical capacitor C20's
Anode, the crus secunda of the 4th power supply chip U4 are connected, one end of the 21st electric capacity C21 and the 4th of the 4th power supply chip U4 the
Foot is connected, and one end of first resistor R1 is connected with one end of anistree, the first electric capacity C1 of the 4th power supply chip U4, and second
5th foot of the anode of diode D2 and the 4th power supply chip U4, the negative terminal of the 11st electric capacity C11, one end phase of the 12nd electric capacity
Connection, the negative terminal of the second diode D2 are connected with the 7th foot of the 4th power supply chip U4, one end of the second inductance L2, and the 19th
The other end of the other end of electric capacity C19 and the 6th resistance R6, the 20th electrolysis negative terminal of C20, the other end of the 21st C21,
The other end of the first electric capacity C1, the other end of the second inductance L2, the anode of the 11st electric capacity C11, the 12nd electric capacity C12 another
End jointly with simulation be connected.
In foregoing circuit, the 3rd power supply chip U3 is adopted using LP38798SDX_ADJ and the 4th power supply chip U4
TPS6735 voltage stabilizing chips, so as to realize exporting positive and negative 5V power supplys supply amplifier, wherein positive 5V power supplys give A/D chip power supplies simultaneously.
LP38798SDX_ADJ is a Width funtion input 3.0V-20V, is 5 μ VRMS, TPS6735 in 10Hz-100KHz output noises
Input voltage range 4V-6.2V, quiescent dissipation reach 1 μ A.So amplifier power supply required precision can be met.
Referring to Figure 16, the circuit theory diagrams of digital power in power module are shown, including the 3rd electric capacity C3, the first power supply
Chip U1, the first inductance L1, second resistance R2, the 8th electric capacity C8, the 9th electric capacity C9, wherein, the 3rd electric capacity C3 one end and first
The crus secunda of power supply chip U1, the 3rd foot are connected, the octal and the tenth of second resistance R2 one end and the first power supply chip U1
Foot, one end of the first inductance L1, one end of the 8th electric capacity, one end of the 9th electric capacity are connected, the other end of the first inductance L1 with
9th foot of the first power supply chip U1 is connected, the 4th foot of the other end of the 3rd electric capacity C3 and the first power supply chip U1, the 9th
Foot, the tenth foot, the 7th foot, the other end of second resistance R2, the other end of the 8th electric capacity, the 9th electric capacity the other end jointly with number
It is connected word.
First power supply chip U1 adopts Ti chip TPS62177DGCR chips, to single-chip microcomputer and wireless module NRF24L01
Power supply.Chip input voltage scope 4.7V-28V, up to 500mA, in a sleep mode, quiescent current only has input current
4.8 μ A, there are overtemperature protection, short-circuit protection etc. in inside.
Referring to Figure 17, show the circuit theory diagrams of ADC base modules in power module, including the 4th electric capacity C4, the 5th
Electric capacity C5, the second reference power supply chip U2, the tenth electric capacity C10, wherein, one end of the 4th electric capacity C4 and the one of the 5th electric capacity C5
End, the second reference power supply chip U2 crus secundas are connected, one end of the tenth electric capacity C10 and the 6th feet of the second reference power supply chip U2
Be connected, the other end of the other end and the 5th electric capacity C5 of the 4th electric capacity C4, the 4th foot of the second reference power supply chip U2, the tenth
The other end of electric capacity jointly with simulation be connected.
Second reference power supply chip U2 adopts 16 Precision A/D C transducers, digital output to change 1LSB, corresponding simulation electricity
Buckling turns to 76 μ V.Therefore higher reference voltage source is needed, ADR445 reference voltage chips have ultra-low noise, high accuracy and low
Temperature drifting performance.Power source change peak-to-peak value only has 2.25 μ V, can meet data acquisition system.
Referring to Figure 18, a kind of circuit theory diagrams of embodiment of signal pre-processing module of the present invention are shown, including:Second
Radar module P2, the 13rd resistance R13, the 33rd electric capacity C33, the 9th integrated transporting discharging U9, the 26th resistance R26, second
19 electric capacity C29, the 25th resistance R25, the 19th resistance R19, the 34th electrochemical capacitor C34, wherein, the second radar mould
Block P2 adopts HB100 modules, one end, the 9th integrated transporting discharging U9 of the 3rd foot and the 13rd resistance R13 of the second radar module P2
Crus secunda be connected, the crus secunda of the second radar module P2 is connected with one end of the 33rd electric capacity C33, the 26th electricity
Resistance one end is connected with the 4th foot of the 9th integrated transporting discharging U9, and the of the other end of the 26th resistance and the 9th integrated transporting discharging U9
One foot, one end of the 29th electric capacity C29 are connected, and the one of the other end and the 25th resistance R25 of the 29th electric capacity C29
End, one end of the 19th resistance R19, the anode of the 34th electrochemical capacitor are connected, the 3rd foot of the 3rd radar module P3 with
The other end of the 33rd electric capacity, the crus secunda of the 9th integrated transporting discharging U9, the other end of the 25th resistance R25, the 34th
The negative terminal of electrochemical capacitor jointly with simulation be connected.
The principle of foregoing circuit is as follows, and as radar signal output impedance is high, carrying load ability is low, in order to impedance is easier
Coupling, front end constitute voltage follower using TLV2631 and not only provide high input impedance and low output impedance.Simultaneously
An isolation buffer effect is played, impact of the signal processing to microwave front-end is reduced, it is ensured that the signal to noise ratio of input signal,
Wave filter can more easily be designed when design for rear class anti-aliasing.And radar emission electromagnetic wave is in fixed object
When, electromagnetic wave echo will not produce Doppler frequency, and its echo-signal is embodied in the signal for receiving straight at zero frequency
In flow component, additionally, radar is radiofrequency signal, the spurious signal in space is excessive, and the amplifier saturation of rear end can be caused even to damage
Bad, in order to prevent causing amplifier saturation due to DC component, it is necessary to filter DC component.In order to be further ensured that signal has
There is high signal to noise ratio, outfan design frequency is being followed for 0.1Hz-150Hz passive RC filters, due to radar breathing heart beating letter
Number frequency is higher than 0.1Hz, and design RC Frequency points will be less than 0.1Hz, and the selection of RC resistance is also required to pay special attention to, if choosing
The input resistance for taking is excessive, then at this time the thermal noise of resistance will be very big, can exceed the input voltage noise level of amplifier,
Amplify interference to rear class larger, so big input capacitance will be chosen as far as possible, then big input capacitance, leakage current are larger,
The direct saturation of rear class amplifying circuit can be caused.So electric capacity needs to choose the less ceramic disc capacitor of leakage current herein.
Further, difference amplifier is used for being amplified input signal and eliminating common-mode noise;Radar signal is passed through
During primary amplification, centre has been mingled with much noise.If primary amplifier amplification is excessive, the full of signal is easily caused
With.On the other hand in order to reduce the impact of signal source, it is necessary to improve the input impedance of amplifier, for radar signal interference main
Common mode disturbances are derived from, primary amplifier Main Function is to eliminate common-mode noise.Differential Input mode is adopted in the present invention, in reality
In the system of border, noise is mostly in the form of common mode.For Differential Input, common-mode noise can be effectively eliminated, so as to
A big chunk noise in signal can be removed.
For integrated transporting discharging, critically important performance indications are exactly common mode rejection ratio CMRR.Which is defined as follows:
Wherein Avd and Avs represent amplifier respectively to difference mode signal and the amplification of common-mode signal.Of the invention a kind of excellent
Select in embodiment, using instrument amplifier.Compare with common integrated transporting discharging, instrument amplifier has higher
Common mode rejection ratio.The CMRR of physiology amplifier typically requires 60dB-80dB, the concrete instrument from Analog Device companies
The CMRR of instrument amplifier AD627 reaches 83dB.AD627 provides flexible user and selects, by a non-essential resistance, it is possible to arrange
Gain, maximum programming gain can reach 1000, be a rail-to-rail low-power consumption instrument amplifiers, with very high cmrr,
Have very wide supply district (± 18V), when being operated in dual power supply, can rail to rail output, be signal amplify ideal
Select.When working at low supply voltages, rail to rail output stage makes dynamic range reach maximum.Ultralow power consumption, is suitable for
Application scenario in portable low power-consumption equipment.
Referring to Figure 19, a kind of circuit theory diagrams of embodiment of difference amplifier of the present invention are shown, including:24th
Resistance R24, the 36th electric capacity C36, the 39th electric capacity C39, the 29th resistance R29, the 12nd integration instrument put U12,
37 electric capacity C37, the 38th electric capacity C38, the 18th resistance R18, the 24th electric capacity C24, the 25th electric capacity C25,
18th resistance R18, the 38th electric capacity C38, wherein, the 24th resistance R24 one end and the one of the 31st electric capacity C31
End, the 12nd integration instrument are put the 3rd foot of U12, the 36th electric capacity C36 one end and are connected, the 36th electric capacity C36 other ends
With the 12nd integration instrument put the crus secunda of U12, one end of the 39th electric capacity C39, the 29th resistance one end be connected,
The octal that U12 is put with the 12nd integration instrument in one end of 18 resistance R18 is connected, the other end of the 18th resistance R18 and
12 integration instruments are put first foot of U12 and are connected, the 24th electric capacity C24 one end is connected with the 25th electric capacity C25 one end,
12nd integration instrument is put the 7th feet of U12 and is connected, the 37th electric capacity C37 one end and the 38th electric capacity C38 one end, the 12nd
Integration instrument is put the 4th foot of U12 and is connected, the other end of the 39th electric capacity C39 other ends and the 29th resistance R29, second
The other end of 14 electric capacity C24, the other end of the 25th electric capacity C25, the other end of the 39th electric capacity C39, the 38th
The other end of electric capacity C38 jointly with simulation be connected.Thus, AD627 Output Voltage Formulas:VO=(5+ (200K Ω/R18))
Vi, realize that signal amplifies.
Referring to Figure 20, a kind of circuit theory diagrams of embodiment of active band-pass filter of the present invention are shown, including:Second
Ten resistance R20, the 30th resistance R30, the 9th resistance R9, the 27th resistance R27, the 16th resistance R16, the 17th resistance
R17, the 7th resistance R7, the 21st resistance R21, the 22nd resistance R22, the 8th resistance R8, the 26th electric capacity C26,
26 electric capacity C26, the 27th electric capacity C27, the 32nd electric capacity C32, the 22nd electric capacity C22, the 35th electric capacity
C35, the 23rd electric capacity C23, electric capacity C, the 8th integrated transporting discharging U8, the tenth integrated transporting discharging U10, the 11st integrated transporting discharging U11,
Wherein, the 26th electric capacity C26 one end is connected with one end of the 27th electric capacity C27, one end of the 9th resistance R9, and the 20th
The other end of seven electric capacity C27 is connected with the 3rd foot of the 8th integrated transporting discharging U8, one end of the 20th resistance R20, the 9th resistance
First foot of the other end of R9 and the 8th integrated transporting discharging U8, the 27th resistance R27 one end, one end phase of the 16th resistance R16
Connection, the other end of the 27th resistance R27 are connected with the 4th foot of the 8th integrated transporting discharging U8, one end of the 30th resistance R30
Connect, the other end of the 16th resistance R16 and the 32nd electric capacity C32 one end, one end of the 7th resistance R7, the 17th resistance R17
One end is connected, the 17th resistance R17 is connected with the 4th foot of the tenth integrated transporting discharging U10, one end of the 22nd electric capacity C22
Connect, the other end of the other end and the 22nd electric capacity C22 of the 7th resistance R7, first foot of the tenth integrated transporting discharging U10, the 20th
One resistance R21 one end is connected, the other end of the 21st resistance R21 and the 35th electric capacity C35 one end, the 22nd resistance
R22 one end, one end of the 8th resistance R8 are connected, the of the other end of the 22nd resistance R22 and the 11st integrated transporting discharging U11
Four feet, one end of the 23rd electric capacity C23 are connected, the other end of the other end and the 23rd electric capacity C23 of the 8th resistance R8,
First foot of the 11st integrated transporting discharging U11 is connected, the 20th resistance R20 other ends and the 30th resistance other end, the 30th
The two electric capacity C32 other ends, the 35th electric capacity C35 other ends jointly with simulation be connected.
In foregoing circuit, the cascade of (MFB) low pass filter is fed back using two second order multiterminal and constitute fourth order low-pass wave filter.
For single second order multiterminal feed back (MFB) low pass filter, can be obtained according to Kirchhoff's theorem and negative feedback amplifier characteristic:
Wherein K be filter gain, ωcFor filter cutoff frequency, B and C is normalization coefficient.
Normalization coefficient B=1.414, C=1 can be obtained according to unlimited gain multiple feedback circuit topological structure, advised by experience
C is then selected32It is similar to 10/fc, by design objective cut-off frequency fc=10Hz, can obtain C32=1uF, filter gain distinguish 1 He
10, low-pass filter circuit device parameters are as shown in table 2.Simulation analysis can obtain the response of low pass filter amplitude-frequency characteristic, its 3dB cut-off frequency
For 8.237Hz, design requirement is met.Specifically related to parameter is as shown in the table.
2 low-pass filter circuit component parameter type selecting of table
Voltage controlled voltage source circuit of high pass filter design principle is, using RC filter circuits and in-phase proportion amplifying circuit group
Into second order voltage controlled voltage source high pass filter, the wave filter has the characteristics of input impedance is high, and output impedance is low.Butterworth is high
The transmission function of bandpass filter is such as
Wherein K be filter gain, ωcFor filter cutoff frequency.
According to design objective, cut-off frequency fc=0.1Hz, filter gain K=10, in f=0.1fcWhen, it is desirable to amplitude declines
Subtract more than 30dB, make R9=R20=R, C26=C27=C, fc=1/ (2 π RC).High-pass filtering circuit component parameter such as 3 institute of table
Show.Simulation result for voltage controlled voltage source high pass filter amplitude-frequency response, its 3dB cut-off frequency be 0.099Hz, pass-band performance meet set
Meter requires that physical circuit device parameters are as shown in table 3 below.
3 high-pass filtering circuit component parameter type selecting of table
Further, breathing and heartbeat signal separation module include voltage movement circuit, analog-digital converter and digital filtering
Device.As amplifying circuit adopts the amplitude of oscillation of dual power supply amplifier, signal to become big, also there is positive negative level in output signal, unavoidably
Make troubles for rear class ADC converter sampling, so need by voltage movement circuit by signal level move ADC conversion
The signal input scope that device is allowed.
Referring to Figure 21, the circuit theory diagrams of voltage movement circuit are shown, including:28th electric capacity C28, the 30th electricity
Hold C30, the tenth resistance R10, the 14th resistance R14, the 12nd resistance R12, the 23rd resistance R23, the 28th resistance
R28, the 11st resistance R11, the 15th resistance R15, the 6th integrated transporting discharging U6, the 7th integrated transporting discharging U7, the 3rd diode D3,
Four diode D4, wherein, one end of the tenth resistance R10 one end and the 14th resistance R14, one end of the 28th electric capacity C28, the
3rd foot of six integrated transporting discharging U6 is connected, the 4th feet of the 6th integrated transporting discharging U6 and the 6th the first feet of integrated transporting discharging U6, the 12nd
One end of resistance R12 is connected, one end, the 7th integrated transporting discharging U7 of the other end and the 11st resistance R11 of the 12nd resistance R12
The 3rd foot be connected, one end of the 23rd resistance R23 one end and the 28th resistance R28, the of the 7th integrated transporting discharging U7
Four feet are connected, the first foot of the other end and the 7th integrated transporting discharging U7 of the 28th resistance R28, the one of the 15th resistance R15
End is connected, one end of the other end and the 30th electric capacity C30 of the 15th resistance R15, the anode of the 3rd diode D3, the four or two
The negative terminal of pole pipe D4 is connected, the other end, the 11st resistance of the other end and the 28th electric capacity C28 of the 14th resistance R14
The other end of R11, the other end of the 30th electric capacity C30, the 4th diode D4 anode jointly with simulation be connected.
In foregoing circuit, radar signal sampling amplifier OPA188 after band-pass filter constitutes calculus of differences electricity
Road, in the positive input superposition constant voltage source of amplifier, constitutes voltage movement circuit, and wherein voltage source adopts amplifier
TLV2631 constitutes voltage follower and produces benchmark 2.5V voltage sources.WhereinSo removed by level
Move and output negative level signal can be moved positive level.Output signal adds two diodes of D3, D4, and anti-stop signal is excessive to fortune
Put and cause to damage, also ensure output signal in the range of ADC converter input voltages.
Analog-digital converter is used for for analog quantity being converted into discrete digital quantity, the system design radar signal output signal frequency
Rate is far below 20Hz, and sample frequency is set to 50Hz, and conversion speed is relatively low, it is possible to use the AD converter of common switching rate.
Radar signal amplify output comprising breathing and heartbeat signal, be ensure following digital Filtering Processing can be good at separate breathing and
Heartbeat signal, this are accomplished by selecting high-resolution and multichannel AD converter.
Referring to Figure 22, the circuit theory diagrams of analog-digital converter are shown, including:43rd electric capacity C43, the 42nd electricity
Hold C42, the 44th electric capacity C44, the 48th electric capacity C48, the 49th electric capacity C49, the 40th electric capacity C40, the 35th
Resistance R35, the 32nd resistance R32, the 13rd AD conversion chip U13, wherein, one end and the 4th of the 43rd electric capacity C43
One end of 12 electric capacity C42, the 9th foot of the 13rd AD conversion chip U13 are connected, one end of the 44th electric capacity C44 and the
Tenth foot of 13 AD conversion chips U13 is connected, one end of the 48th electric capacity C48 one end and the 49th electric capacity C49, the
One end of 35 resistance, the 13rd foot of the 13rd AD conversion chip U13 are connected, the 40th electric capacity C40 one end and the tenth
16th foot of three AD conversion chips U13 is connected, one end of the 32nd resistance R32 and the 13rd AD conversion chip U13
First foot is connected, the other end of the other end and the 42nd electric capacity C42 of the 43rd electric capacity C43, the 44th electric capacity
The other end, the 11st foot of the 13rd AD conversion chip U13, the 12nd foot, the other end of the 48th electric capacity C48, the 40th
The other end of nine electric capacity jointly with simulation be connected.The other end of the 40th electric capacity C40 be digitally connected.
Wherein, using Maxim MAX1167 analog-digital converter, the chip is low-power consumption, multichannel, 16 Approach by inchmeal
Pattern number converter (ADC), in 10kps, electric current only 185 μ A.There is internal reference and outside reference is available and band
The interface for having a high speed SPI/QSPI/ compatible.MAX1167 is using single+5V analog powers work and electric with independent numeral
Source, it is allowed to directly with the+2.7V extremely Digital Logic interfaces of+5.5V.MAX1167 external reference voltages source is high-precision AD R445,
There is very high degree of stability.The excellent dynamic properties of MAX1167 and low-power consumption, it is sufficient to meet wanting for current system A/D converter
Ask.
Digital filter is separated to breath signal and heartbeat signal in frequency domain using digital filtering technique.In the present invention
In a kind of preferred implementation, digital filter is using appointing in FIR filter, iir filter or zero phase iir filter
A kind of.The design principle of three kinds of digital filters is described in detail in detail separately below.
FIR (Finite Impulse Response) wave filter is that have limit for length's unit impulse response wave filter, and it can be
Ensure that there is while any amplitude-frequency characteristic strict linear phase-frequency characteristic, while its unit sample respo is time-limited, because
And wave filter is stable system.Breathing, heartbeat signal in due to physiological signal, energy are concentrated mainly near zero-frequency, adopt
It must is fulfilled for traditional digital filter claimed below:
(1) breathing, the frequency band range of heartbeat signal are concentrated mainly on 0.1Hz-4Hz, and therefore the bandwidth of wave filter must be non-
Often narrow, to detect that energy concentrates on the echo signal of low-frequency range;
(2) in order to filter noise jamming and noise outside useful signal frequency band range, in frequency domain, the intermediate zone of wave filter
Sinking speed is very fast, to obtain steeper intermediate zone, reduces the wave rear of wave filter as far as possible.
In the present invention, the design objective of physiological signal wave filter is as shown in table 4 below.
Table 4:Physiological signal wave filter design objective
Two kinds of Direct Method of Design of FIR filter are adding window Fourier space method and frequency sampling method.In design filtering
During device, after selecting the type of digital filter, next will estimate to meet the filtering required for given Filter specification
The exponent number of device.In order to reduce the complexity for calculating, filter order should be elected as and obtain smallest positive integral more than or equal to the estimated value.
The value of the type and length of window N of window function w (n) is depended on the performance of filter that window function metht is designed.?
In filter design procedure, after selecting the type of digital filter, next will estimate needed for the given Filter specification of satisfaction
The filter order that wants.For reducing the complexity for calculating, filter order should be elected as and be obtained most more than or equal to the estimated value
Small integer.Some scholars are proposed from the minimum equation that the index direct estimation filter order of numbers below wave filter is N
Such as Kaiser equations:If normalization passband border angular frequencyp, normalization stopband border angular frequencys, peaked passband ripple δp,
And peak value stopband ripple δs.Kaiser equations:
Wherein, frequencies omegapAnd ωsIt is referred to as passband edge frequency and stopband edge frequency.δpAnd δsReferred to as passband and resistance
The error capacitance of band is peak waviness.
And peaked passband ripple quantity αp=-20lg (1- δp) dB, minimum stop-band attenuation αs=-20lg (δs)dB.
If sample frequency is ft, fp and fs is passband and stopband edge frequency, then the normalization border in units of radian
Angular frequency can be expressed as:
It is possible thereby to the length of window of practical filter is estimated according to Kaiser, can then proceed in intermediate zone and stopband
Attenuation, selects window function form.The selection of window function should meet:In the case where ensureing that stopband attenuation meets requirement, to the greatest extent
Measure the narrow window function of selection main lobe to obtain steeper intermediate zone;Reduce the relative amplitude of the maximum secondary lobe of window spectrum as far as possible to reduce ripple
Stricture of vagina peak value.Performance indications of the table 5 for various window functions.
5 window function performance indications of table
Breathing can be calculated according to Kaiser equations and heartbeat signal window function length N smallest positive integral values are respectively:227
With 302.Can meet according to stopband maximum gain and be close to the window function of satisfaction and have Hanning window and Hamming window, due to breath signal and
Heartbeat signal is very close in the spectral peak of frequency domain, it is therefore desirable to choose the high window function of a frequency resolution.Hanning window and
Hamming window belongs to raised cosine window, is characterized in that secondary lobe is revealed few.The two comparatively speaking, it is peaceful that the main lobe of Hamming window is slightly narrower than the Chinese
Window, and the first side lobe attenuation speed of Hamming window is faster than Hanning window, above-mentioned 2 points of frequency resolutions for all causing Hamming window are better than
Hanning window, therefore from Hamming window as filter window function.
Iir digital filter is referred to as recursion filter, using in recursion type structure, i.e. structure carry feedback control loop.IIR
Filter operation structure generally by time delay, be multiplied by coefficient and the elementary operation such as be added and constitute, Direct-type, positive accurate can be combined into
Type, cascade connection type, four kinds of versions of parallel connection type, all have feedback circuit.For iir digital filter, the most frequently used design handss
Section is that the design objective of digital filter is changed into Design of Analog Filter index, so that it is determined that meeting the simulation of these indexs
The transmission function of wave filter, then retells the transmission function that it is converted to required digital filter.Its advantage is available with
Some classical analog filter forms are rapidly completed design.Conventional analog filter has Butterworth (Butterworth)
Wave filter, Chebyshev (Chebyshev) wave filter, ellipse (Ellipse) wave filter, Bezier (Bessel) wave filter etc..
Digital filter and analog filter are tied in a hundred and one ways, and the conversion between them is the conversion of s planes and z-plane, turn
The basic mode that changes is exactly Impulse invariance procedure and Bilinear transformation method.Elliptic filter, it are designed using elliptic method
The analog filter of low pass, then using conversion method obtain numeral high pass, low pass, band logical and band hinder wave filter.?
In the design of analog filter, the design of elliptic filter is a kind of the most complicated method in several filter design methods,
But the exponent number of the wave filter that it designs is minimum, and its intermediate zone is narrow.Elliptic filter is compared other kinds of
Wave filter, the passband and stop band ripple for having minimum under the conditions of exponent number identical are identical with the fluctuation of stopband in passband.
Using elliptic filter, minimum exponent number can be obtained, realize that given wave filter technology index, elliptic filter are needed
The amount of calculation that wants is minimum.Matlab filter design toolbox FDATOOL are based on, filter parameter ibid one saves design parameter one
In the case of cause, the elliptic filter exponent number minimum for extracting breath signal only needs 8 ranks, for extracting the oval filter of heartbeat signal
Ripple device exponent number minimum only needs to 14 ranks, it can be seen that operand is far smaller than FIR filter exponent number.
Referring to Figure 23 and Figure 24, it show and is respectively adopted FIR filter and iir filter is filtered and separates breath signal
Time domain and frequency domain comparison diagram, from the point of view of experimental result, in the signal contrast of time domain and frequency domain, FIR filter and iir filter
Breath signal can be efficiently separated out, FIR filtered signals phase-frequency characteristic is good, easily realizes linear phase, but required filtering
Device exponent number is high, and computing memory element is more, and signal delay is larger.Iir filter realizes same design index parameter, with wave filter
Exponent number is few, and required computing memory element is few, and the features such as operand is few, but filtered signal has severe phase distortion.
For the pluses and minuses of above two filtering method, the present invention is optimized on the basis of IIR filtering methods and improves
Propose zero phase iir filter afterwards, signal phase distortion is completely eliminated so as to reach.
The ultimate principle of zero phase iir filter is as follows:IIR filtering is separately designed according to breathing and heartbeat signal first
Device, then makes the positive output for obtaining filtering for the first time by wave filter of signal sequence, then by the output sequence of first time filtering
Row carry out time domain upset, and the sequence after time domain is overturn carries out secondary filtering by same wave filter, defeated after secondary filtering
Go out, positive time serieses and flip-flop transition sequence can be so utilized by phase phase shift during wave filter
Mutually offset, so as to realize the zero phase-shift of filter result.Hypothesis filter function is H (z), and the z of list entries becomes and turns to X (z), then
Zero-phase filtering process can be expressed as follows:
Y1(ejω)=X (ejω)H(ejω);
Y2(ejω)=e-jω(N-1)Y1(e-jω);
Y3(ejω)=Y2(ejω)H(ejω);
Y4(ejω)=e-jω(N-1)Y3(e-jω);
There is above formula to derive to obtain, finally entering output can be expressed as:
Y(ejω)=X (ejω)|H(ejω)|2
It is possible thereby to realize that zero phase-shift is filtered, from formula it can be seen that x sequences square are multiplied with filter function, therefore
The exponent number of wave filter can be doubled, and because square multiplication, compares compared to other filtering, the amplitude of signal can decrease.
Referring to Figure 25 and 26, breath signal and heartbeat signal time-domain diagram after zero-phase filtering is shown, Figure 27 is breathing letter
Number and heartbeat signal separation frequency domain figure, it can be seen that zero-phase filtering one side signal amplitude has portion than primary signal
Decay, the exponent number of another aspect wave filter is divided also to double, but for hundreds of rank that FIR filters calculating exponent number, rank
Several or very little, amount of calculation can be substantially reduced, furthermore due to signal time domain truncation, signal boundary being caused to lose during filtering
Very, after for breath signal using 8 rank wave filter, reusing zero-phase filtering exponent number can increase to 16 ranks, signal both sides
Distorted signals, 16 point datas of each loss.But generally speaking amplitude fading is not it is obvious that signal boundary loss in both sides is to whole letter
Number affect not being very big that the extraction of amplitude of respiration frequency is not had a significant impact, signal characteristic can be effectively extracted.
In a preferred embodiment, digital filter is realized by the program in the MCU module.
The explanation of above example is only intended to help and understands the method for the present invention and its core concept.It should be pointed out that right
For those skilled in the art, under the premise without departing from the principles of the invention, the present invention can also be carried out
Some improvement and modification, these improvement and modification are also fallen in the protection domain of the claims in the present invention.
The foregoing description of the disclosed embodiments, enables professional and technical personnel in the field to realize or using the present invention.
Multiple modifications of these embodiments will be apparent for those skilled in the art, as defined herein
General Principle can be realized without departing from the spirit or scope of the present invention in other embodiments.Therefore, the present invention
The embodiments shown herein is not intended to be limited to, and is to fit to and principles disclosed herein and features of novelty phase one
The most wide scope for causing.
In a preferred embodiment, also include for acquired breath signal and heartbeat signal being sent to vehicle-mounted computer
The step of, human body respiration signal wirelessly will be obtained by wireless communication module and heartbeat signal is sent to vehicle mounted electric
Brain.Wireless communication module is connected with the MCU module, for the MCU module is obtained human body respiration signal and heart beating letter
Number vehicle-mounted computer is sent to, further, wireless communication module adopts 2.4G wireless module NRF24L01.Deposited by vehicle-mounted computer
Storage and the situation of process patient respiratory and heart beating change, improve physiology by the process of vehicle-mounted computer big data and the function of storing and believe
Number accuracy of detection, and can show in real time.
The explanation of above example is only intended to help and understands the method for the present invention and its core concept.It should be pointed out that right
For those skilled in the art, under the premise without departing from the principles of the invention, the present invention can also be carried out
Some improvement and modification, these improvement and modification are also fallen in the protection domain of the claims in the present invention.
The foregoing description of the disclosed embodiments, enables professional and technical personnel in the field to realize or using the present invention.
Multiple modifications of these embodiments will be apparent for those skilled in the art, as defined herein
General Principle can be realized without departing from the spirit or scope of the present invention in other embodiments.Therefore, the present invention
The embodiments shown herein is not intended to be limited to, and is to fit to and principles disclosed herein and features of novelty phase one
The most wide scope for causing.
Claims (10)
1. a kind of fatigue drive of car detection method based on Multi-source Information Fusion, it is characterised in that comprise the following steps:
Step S1:By persistently carrying out to physiological driver's signal intensity installed in the doppler radar unit for driving interior
Detection is simultaneously obtained and stores physiological signal in real time;
Step S2:Variable quantity of the steering wheel angle information in Preset Time is obtained, and is started to when previous according to the variable quantity
Acquired physiological signal in the Duan Lianxu moment carries out data processing;
Step S3:Judge whether fatigue driving occur according to physiological signal change.
2. the fatigue drive of car detection method based on Multi-source Information Fusion according to claim 1, it is characterised in that
In step S3, by the way of neural network machine study, judge whether fatigue driving occur, wherein, know in neural network machine
The data template of various fatigue state grades is prestored in other module, and the physiological signal input neural network machine of collection is known
Other module carries out data analysiss so as to judging fatigue state grade.
3. the fatigue drive of car detection method based on Multi-source Information Fusion according to claim 2, it is characterised in that god
Extreme learning machine model is adopted through net machine identification module.
4. the fatigue drive of car detection method based on Multi-source Information Fusion according to claim 2, it is characterised in that each
The data template for planting fatigue state grade passes through the breath signal that will be obtained and heartbeat signal input neutral net learning base
Determine in the expert analysis mode method of facial video.
5. the fatigue drive of car detection method based on Multi-source Information Fusion according to claim 1, it is characterised in that
In step S1, further comprising the steps:
Step S11:Continuous wave radar signal is launched to torso model by doppler radar sensor;
Step S12:Echo-signal and transmitting concussion frequency signal are carried out obtaining reaction human body respiration after Frequency mixing processing detection
Low frequency signal with heart beating change;
Step S13:Impedance matching is carried out to doppler radar sensor outfan and filters the DC component in low frequency signal;
Step S14:Signal after processing through step 3 is carried out signal amplification;
Step S15:Process is filtered to its input signal by the band filter of 0.1Hz-10Hz;
Step S16:Frequency filtering is carried out using digital filtering technique prize through the signal after the process of step S25 to believe so as to obtain breathing
Number and heartbeat signal;
In step S3, whether the change according to the breath signal and heartbeat signal judges driver in fatigue driving shape
State.
6. the fatigue drive of car detection method based on Multi-source Information Fusion according to claim 5, it is characterised in that
In step S15, band logical is realized by quadravalence Butterworth LPF and second order butterworth high pass filter
Wave filter.
7. contactless bio-signal acquisition method according to claim 5, it is characterised in that in step S16, adopt
Any one in FIR filter, iir filter or zero phase iir filter realizes the separation of breath signal and heartbeat signal.
8. contactless bio-signal acquisition method according to claim 5, it is characterised in that zero phase iir filter
Realize that step is as follows:
Step S161:Feature according to breath signal and heartbeat signal separately designs breath signal iir filter and heartbeat signal
Iir filter;
Step S162:Input signal is carried out signal sampling and is stored as digital signal sequences;
Step S163:The digital signal sequences are separately input to breath signal iir filter and heartbeat signal iir filter enters
Row first time Filtering Processing;
Step S164:The signal exported through above-mentioned first time Filtering Processing is executed the upset of first time time domain;
Step S165:Step S64 output signal is again inputted into breath signal iir filter and heartbeat signal iir filter
Carry out second Filtering Processing;
Step S166:The signal exported through above-mentioned second Filtering Processing is executed second time domain upset, so as to be filtered
Breath signal afterwards and heartbeat signal;
Step S167:Filtered breath signal and heartbeat signal are carried out obtaining frequency spectrum after FFT respectively so as to realizing exhaling
Inhale the separation of signal and heartbeat signal.
9. contactless bio-signal acquisition method according to claim 5, it is characterised in that in step step S11, institute
State the microwave Doppler radar detedtor probe sensor that doppler radar sensor adopts working frequency range for 10.525GHz
HB100 modules.
10. contactless bio-signal acquisition method according to claim 5, it is characterised in that in step s 13, adopt
Impedance matching is carried out to doppler radar sensor outfan with voltage follower, band connection frequency is adopted for 0.1Hz-150Hz's
Passive RC filter filters the DC component in low frequency signal.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610890731.7A CN106491156A (en) | 2016-10-13 | 2016-10-13 | A kind of fatigue drive of car detection method based on Multi-source Information Fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610890731.7A CN106491156A (en) | 2016-10-13 | 2016-10-13 | A kind of fatigue drive of car detection method based on Multi-source Information Fusion |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106491156A true CN106491156A (en) | 2017-03-15 |
Family
ID=58294781
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610890731.7A Pending CN106491156A (en) | 2016-10-13 | 2016-10-13 | A kind of fatigue drive of car detection method based on Multi-source Information Fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106491156A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107374603A (en) * | 2017-08-23 | 2017-11-24 | 辽宁工业大学 | A kind of safe driving decision-making system and decision method |
CN107886679A (en) * | 2017-11-26 | 2018-04-06 | 南通尚力机电工程设备有限公司 | A kind of real-time driving fatigue monitoring system |
CN108622096A (en) * | 2017-03-17 | 2018-10-09 | 电装国际美国公司 | Vehicular system and vehicle control device for controlling vehicle |
CN109190718A (en) * | 2018-11-07 | 2019-01-11 | 苏州热工研究院有限公司 | A kind of nuclear power plant's marine growth detecting early-warning Multi-source Information Fusion algorithm |
CN109893732A (en) * | 2019-02-28 | 2019-06-18 | 杭州智瑞思科技有限公司 | A kind of mechanical ventilation patient-ventilator asynchrony detection method based on Recognition with Recurrent Neural Network |
CN110435662A (en) * | 2019-07-30 | 2019-11-12 | 上海思致汽车工程技术有限公司 | A kind of automobile driver and passenger monitoring system |
CN110450784A (en) * | 2019-07-30 | 2019-11-15 | 深圳普捷利科技有限公司 | A kind of driver status monitoring method and system based on fmcw radar |
CN110667595A (en) * | 2019-10-14 | 2020-01-10 | 深圳迈睿智能科技有限公司 | Monitoring method and monitoring system for physical state of vehicle driver |
CN111166357A (en) * | 2020-01-06 | 2020-05-19 | 四川宇然智荟科技有限公司 | Fatigue monitoring device system with multi-sensor fusion and monitoring method thereof |
CN111209816A (en) * | 2019-12-28 | 2020-05-29 | 杭州电子科技大学 | Non-contact fatigue driving detection method based on regular extreme learning machine |
TWI733293B (en) * | 2019-12-30 | 2021-07-11 | 關貿網路股份有限公司 | Driving occupational evaluation system and evaluation device |
CN115137941A (en) * | 2022-06-21 | 2022-10-04 | 北京中科心研科技有限公司 | Emergency awakening method and system based on fatigue recognition model and electrical stimulation |
TWI785378B (en) * | 2020-09-03 | 2022-12-01 | 雲云科技股份有限公司 | Optical image physiological monitoring system with radar detection assistance |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101746269A (en) * | 2010-01-08 | 2010-06-23 | 东南大学 | Fatigue driving fusion detection method based on soft computing |
CN104057896A (en) * | 2013-03-20 | 2014-09-24 | 厦门歌乐电子企业有限公司 | Vehicle-mounted reminding device |
WO2015160272A1 (en) * | 2014-04-14 | 2015-10-22 | Novelic D.O.O. | Mm-wave radar driver fatigue sensor apparatus |
-
2016
- 2016-10-13 CN CN201610890731.7A patent/CN106491156A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101746269A (en) * | 2010-01-08 | 2010-06-23 | 东南大学 | Fatigue driving fusion detection method based on soft computing |
CN104057896A (en) * | 2013-03-20 | 2014-09-24 | 厦门歌乐电子企业有限公司 | Vehicle-mounted reminding device |
WO2015160272A1 (en) * | 2014-04-14 | 2015-10-22 | Novelic D.O.O. | Mm-wave radar driver fatigue sensor apparatus |
Non-Patent Citations (2)
Title |
---|
李佳晖等: "基于雷达式非接触生命参数监测系统的安全带设计", 《科技创新与应用》 * |
胡治等: "非接触呼吸与心跳监护装置的研制", 《中国医疗器械杂志》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108622096A (en) * | 2017-03-17 | 2018-10-09 | 电装国际美国公司 | Vehicular system and vehicle control device for controlling vehicle |
CN107374603A (en) * | 2017-08-23 | 2017-11-24 | 辽宁工业大学 | A kind of safe driving decision-making system and decision method |
CN107886679A (en) * | 2017-11-26 | 2018-04-06 | 南通尚力机电工程设备有限公司 | A kind of real-time driving fatigue monitoring system |
CN109190718A (en) * | 2018-11-07 | 2019-01-11 | 苏州热工研究院有限公司 | A kind of nuclear power plant's marine growth detecting early-warning Multi-source Information Fusion algorithm |
CN109190718B (en) * | 2018-11-07 | 2021-02-12 | 苏州热工研究院有限公司 | Marine organism detection early warning multi-source information fusion method for nuclear power plant |
CN109893732A (en) * | 2019-02-28 | 2019-06-18 | 杭州智瑞思科技有限公司 | A kind of mechanical ventilation patient-ventilator asynchrony detection method based on Recognition with Recurrent Neural Network |
CN109893732B (en) * | 2019-02-28 | 2021-04-06 | 杭州智瑞思科技有限公司 | Mechanical ventilation man-machine asynchrony detection method based on recurrent neural network |
CN110435662A (en) * | 2019-07-30 | 2019-11-12 | 上海思致汽车工程技术有限公司 | A kind of automobile driver and passenger monitoring system |
CN110450784A (en) * | 2019-07-30 | 2019-11-15 | 深圳普捷利科技有限公司 | A kind of driver status monitoring method and system based on fmcw radar |
CN110667595A (en) * | 2019-10-14 | 2020-01-10 | 深圳迈睿智能科技有限公司 | Monitoring method and monitoring system for physical state of vehicle driver |
CN111209816B (en) * | 2019-12-28 | 2023-10-24 | 杭州电子科技大学 | Non-contact fatigue driving detection method based on regular extreme learning machine |
CN111209816A (en) * | 2019-12-28 | 2020-05-29 | 杭州电子科技大学 | Non-contact fatigue driving detection method based on regular extreme learning machine |
TWI733293B (en) * | 2019-12-30 | 2021-07-11 | 關貿網路股份有限公司 | Driving occupational evaluation system and evaluation device |
CN111166357A (en) * | 2020-01-06 | 2020-05-19 | 四川宇然智荟科技有限公司 | Fatigue monitoring device system with multi-sensor fusion and monitoring method thereof |
TWI785378B (en) * | 2020-09-03 | 2022-12-01 | 雲云科技股份有限公司 | Optical image physiological monitoring system with radar detection assistance |
CN115137941A (en) * | 2022-06-21 | 2022-10-04 | 北京中科心研科技有限公司 | Emergency awakening method and system based on fatigue recognition model and electrical stimulation |
CN115137941B (en) * | 2022-06-21 | 2024-05-03 | 北京中科心研科技有限公司 | Emergency awakening method and system based on fatigue recognition model and electric stimulation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106491156A (en) | A kind of fatigue drive of car detection method based on Multi-source Information Fusion | |
CN106297194A (en) | A kind of fatigue drive of car monitoring system | |
CN106264502B (en) | A kind of contactless bio-signal acquisition method | |
CN106264501A (en) | A kind of bio-signal acquisition system based on continuous wave Doppler radar | |
Xiao et al. | Wearable heart rate monitoring intelligent sports bracelet based on Internet of things | |
CN109480787A (en) | A kind of contactless sleep monitor equipment and sleep stage method based on ULTRA-WIDEBAND RADAR | |
Ali et al. | Evaluation of electrocardiogram signals classification using CNN, SVM, and LSTM algorithm: A review | |
CN103083013A (en) | Electrocardio signal QRS complex wave detection method based on morphology and wavelet transform | |
CN103690169B (en) | Respiratory information detecting method | |
CN103337132A (en) | Tumble detection method for human body based on three-axis acceleration sensor | |
Shokouhmand et al. | Efficient detection of aortic stenosis using morphological characteristics of cardiomechanical signals and heart rate variability parameters | |
CN105997043A (en) | Pulse frequency extraction method based on wrist type wearable device | |
CN112998689B (en) | Non-contact multi-person real-time respiration condition evaluation system, device and storage medium | |
Dong et al. | Accurate detection of Doppler cardiograms with a parameterized respiratory filter technique using a K-band radar sensor | |
CN107361753A (en) | Health state monitoring method based on pulse wave characteristic point | |
CN206239402U (en) | A kind of fatigue drive of car real-time detecting system based on DSP | |
CN113273978B (en) | Ultra-wideband radar-based human body respiration and heartbeat frequency detection method | |
Shi et al. | Contactless person identification using cardiac radar signals | |
Xing et al. | Dwatch: A reliable and low-power drowsiness detection system for drivers based on mobile devices | |
CN109199366A (en) | Electrocardiosignal QRS complex detection method based on variation mode decomposition | |
Liu et al. | Non-contact human fatigue assessment system based on millimeter wave radar | |
CN205758555U (en) | A kind of detection device of human body respiration | |
Shintomi et al. | Effectiveness of the heartbeat interval error and compensation method on heart rate variability analysis | |
Jing et al. | Harmonics and intermodulation products-based fuzzy logic (HIPBFL) algorithm for vital sign frequency estimation using a UWB radar | |
CN204274481U (en) | The evaluating system that a kind of depression degree quantizes |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170315 |