A kind of DAS (Driver Assistant System) based on environment voice recognition and positioningTechnical field
The invention belongs to voice recognitions to assist driving field, and in particular to one kind is based on environment voice recognition and positioningDAS (Driver Assistant System).
Background technique
The auxiliary driving technology of automobile is quickly grown at present, advanced driving assistance system (Advanced DriverAssistance System, abbreviation ADAS) it is more and more widely used on automobile.It is continuous due to automobile quantityIncrease, the driving environment that driver is faced also becomes increasingly complex.
Nowadays in driving, it is frequently encountered the sound that driver ignores some keys outside vehicle due to absent mindedSound, for example, police car, ambulance, fire fighting truck police whistle sound or other vehicles tucket, cause driver not make anti-in timeIt reflects, the driving for causing the passage for hindering special vehicle or influencing other vehicles.
The important information in surrounding environment mainly is captured using radar or camera in current ADAS system, intoRow relevant treatment and the driving behavior for reminding or assisting driver, increase the comfortableness and security of car steering.But it is based onSonic transducer carries out that the system research that Context awareness auxiliary drives is seldom, especially for sound extraordinary in environment identification in turnThe system driven is assisted, there is very big research space.
There is special vehicle when driver encounters around and give way not in time, hinders the rescue of special vehicle；Or it waitsStart not in time when traffic lights, influences the passage of rear car；Or it is not careful to rear car observation when lane change, the safety for influencing both sides is drivenIt sails, if the police whistle sound of above-mentioned special vehicle can be identified by voice recognition device, the tucket of rear car prompt, and then remind and driveMember, assists its driving behavior, and elimination driver is ignored these crucial ambient sounds due to absent minded and made to drivingAt influence, the safety of driving can be greatly improved, also allow driver that can more comprehensively understand the driving environment of surrounding.
Summary of the invention
DAS (Driver Assistant System) proposed by the present invention based on the voice recognition of environment special type and positioning, can accurately identifyThe ambient sound of several keys in daily environment, and driver is reminded, its driving behavior is assisted, the safety of driving is increasedProperty.
The present invention achieves the above technical objects by the following technical means.
A kind of DAS (Driver Assistant System) based on environment voice recognition and positioning characterized by comprising
Sound acquisition module, for acquiring ambient sound when driving, and by collected voice input preprocessing moduleIt is handled；
Preprocessing module, for being denoised to the voice signal of acquisition, end-point detection, framing, windowing process, and will be pre-Treated, and signal is input to identification module is identified in real time；
Sound recognition module, including important environment audio database and pattern matching unit, the audio databaseIn be stored with environment sound, pattern matching unit can carry out characteristic parameter extraction to sound, and in audio databaseData carry out pattern match, obtain recognition result；
ECU electronic control module controls the positioning that auditory localization module carries out Sounnd source direction according to recognition result；Sound source is fixedPosition module feeds back to ECU electronic control mould for judging the relative position direction in sound source and vehicle, and by positioning resultBlock；
As a result output module, for exporting voice recognition result and positioning result.
In above scheme, the sound acquisition module includes being mounted on 10 sound biography around vehicle body, in array arrangementSensor, the sonic transducer top are equipped with silencing cotton.
In above scheme, preprocessing module can be denoised the voice signal of acquisition using the method for wavelet threshold.
In above scheme, police car, the police whistle sound of ambulance and fire fighting truck and automobile are included in the audio databaseTucket；The extracted sound characteristic parameter of pattern matching unit is mel-frequency cepstrum coefficient, and the pattern match link is adoptedIt is realized with the algorithm based on support vector machines.
In above scheme, auditory localization module is to judge audio direction based on acoustic sensor array method, according to sonic transducerTime difference and phase difference of the array acquisition to same sound judge that sound source forms circle in the three-dimensional space centered on automobileIn specific orientation.
In above scheme, the result output module is auditory tone cues module or display module.
In above scheme, the display module is hud new line display module, and hud new line display module is controlled by ECU, energyIt is enough that the result of voice recognition and auditory localization is projected on front windshield.
In above scheme, if detecting target sound, ECU electronic control module controls the work of auditory localization module, butIf obtained recognition result is tucket, the maximum decibel value that 10 sonic transducers collect this tucket is first compared, ifMaximum decibel value is less than 40dB, then ECU electronic control module does not execute subsequent step, leaves the big tucket of decibel value, and recognizeIt is important ambient sound for it.
It further include the vehicle speed sensor being connected with ECU and steering wheel angle sensor, ECU is according to sound in above schemeIt is that sound identifies and positions as a result, in conjunction with current vehicle speed and whether be in lane-change state, provide the suggestion of driver behavior, auxiliary drivesMember carries out operation judges.Hud new line display module can be by the speed of current driving, if judgement and sound in lane-change stateThe result identified and positioned combines, and driver behavior suggestion is provided in projection, and auxiliary driver carries out operation judges.
In above scheme, the result output module further includes steering wheel vibration module, is controlled by ECU, when identifying meshWhen marking environment sound, ECU control steering wheel is vibrated, and carries out different numbers and characteristic according to different recognition resultsVibration, to remind driver to pay attention to environment situation.
Beneficial effects of the present invention:
(1) present invention can be realized the identification to important ambient sounds several in daily driving, and carry out to driverRemind, assist the driving behavior of driver, traffic safety can be increased, avoid driver due to it is absent minded cause it is not rightSpecial vehicle avoids in time, or the tucket of ignorance surrounding vehicles leads to the normal pass for hindering other vehicles.
(2) the sound characteristic parameter that the present invention uses is mel-frequency cepstrum coefficient, can preferably be characterized several importantThe characteristics of ambient sound, and the pattern matching unit based on support vector machines can quickly and efficiently obtain recognition result, increaseThe real-time of system.
(3) sound source position of target sound can be accurately positioned out in the auditory localization module that the present invention uses, and passes through vehicle bodyAround 6 region divisions intuitively show driver, driver's more intuitive understanding environment can be made, and according to working asPreceding speed and whether give driver in lane change state and accordingly prompt, assists the driver behavior of driver.
(4) the hud new line display module and steering wheel vibration module that the present invention uses can not stop driver normalIt travels sight and does not need to make it more intuitively recognize relevant information in the case where driver's rotary head, improve the safety of drivingProperty.
(5) present invention has also combined the driving status of vehicle, current vehicle speed and whether is in lane-change state, provides driving and graspsThe suggestion of work, auxiliary driver carry out operation judges.
(6) when identifying target driving ambient sound, steering wheel is controlled by steering wheel vibration module and is vibrated, andThe vibration of different numbers and characteristic is carried out, according to different recognition results to remind driver to pay attention to environment situation.
Detailed description of the invention
Fig. 1 is the frame diagram of the DAS (Driver Assistant System) of the present invention based on environment voice recognition and positioning.
Fig. 2 is the flow chart of the auxiliary driving method based on environment voice recognition and positioning.
Fig. 3 is the styles figure of hud new line display projection.
Specific embodiment
Present invention will be further explained with reference to the attached drawings and specific examples, but protection scope of the present invention is simultaneouslyIt is without being limited thereto.
Fig. 1 show the best reality of the DAS (Driver Assistant System) of the present invention based on environment voice recognition and positioningApply example, including sound acquisition module, preprocessing module, sound recognition module, auditory localization module, ECU electronic control module,Hud new line display module, steering wheel vibration module, vehicle speed sensor module and steering wheel angle sensor module.
Sound acquisition module, for acquiring ambient sound when driving, and by collected voice input preprocessing moduleIt is handled.Including being mounted on 10 sonic transducers around vehicle body, arranging in array, it is equipped with and disappears at the top of the sonic transducerSound cotton can be effectively reduced the interference that driving apoplexy is made an uproar to target sound, improve the signal-to-noise ratio of signal.
Preprocessing module, for being denoised to the voice signal of acquisition, end-point detection, framing, windowing process, and will be pre-Treated, and signal is input to identification module is identified in real time.Wavelet threshold is used to the voice signal denoising of acquisitionMethod, can effectively filter out unnecessary interference sound, improve accuracy of identification.
Sound recognition module, including important environment audio database and pattern matching unit, the audio databaseIn be stored with environment sound, the tucket comprising police car, ambulance, the police whistle sound of fire fighting truck and automobile.Pattern matchUnit can carry out characteristic parameter extraction to sound, and carry out pattern match with the data in audio database, obtain identification knotFruit.The extracted sound characteristic parameter of pattern matching unit is mel-frequency cepstrum coefficient, can several important drivings of Efficient CharacterizationThe feature of ambient sound reaches higher accuracy of identification.The pattern match link uses the algorithm based on support vector machinesIt realizes, rapidly and accurately realizes the identification of sound.
It is two-way communication between ECU electronic control module and sound source locating module, it is fixed sound source can be controlled according to recognition resultThe operation of position module carries out the positioning of Sounnd source direction, to judge the relative position direction in sound source and vehicle, and receives sound sourceThe positioning result of locating module.
If detecting target sound, ECU electronic control module controls the work of auditory localization module, carries out Sounnd source directionPositioning.But if obtained recognition result is tucket, the maximum point that 10 sonic transducers collect this tucket is first comparedShellfish value, if maximum decibel value is less than 40dB, ECU electronic control module does not execute subsequent step, can filter out so at a distanceWoofer sound, leave the big tucket of decibel value, and think it for important ambient sound.
Auditory localization module is to judge audio direction based on acoustic sensor array method, is collected together according to acoustic sensor arrayThe time difference of one sound and phase difference judge that sound source forms the specific side in circle in the three-dimensional space centered on automobilePosition.
Hud new line display module is controlled by ECU, and ECU controls hud new line display module for voice recognition and auditory localizationAs a result it is projected on front windshield with designed styles, driver is allowed intuitively to recognize the letter of driving environmentBreath, and the sight of normal driving is not blocked.Circle centered on automobile is divided into 6 pieces of regions, corresponding automobile just beforeSide, left front, right front, left back, right back and dead astern.The corresponding region in the specific orientation for the sound that will identify that is prominentIt has been shown that, intuitively shows driver.Fig. 3 show the display sample of the display module hud new line display module, knows at this timeNot Chu the right back of automobile have the police whistle sound of ambulance, corresponding right rear region will be highlighted, and remind the attention of driver.At this timeSpeed is 50km/h, and steering wheel angle is less than 2 degree, driver behavior suggestion that ECU is provided are as follows: please keep lane, pay attention to evacuation.
Steering wheel vibration module is controlled by ECU, and when identifying target driving ambient sound, ECU controls steering wheel and carries outIt vibrates, and carries out the vibration of different numbers and characteristic according to different recognition results, to remind driver to pay attention to environment situation.
Above-mentioned steering wheel vibration module, hud new line display module belong to result output module, for exporting voice recognitionAnd positioning result as a result.In addition to the above results way of output, result prompt can also be carried out by the way of auditory tone cues.OrPerson's auditory tone cues are presented with image, vibration uses simultaneously.
Vehicle speed sensor and steering wheel angle sensor are connected with ECU, vehicle speed sensor and steering wheel angle sensorDriving status for real-time detection vehicle.Whether ECU is according to voice recognition with positioning as a result, in conjunction with current vehicle speed and being inLane-change state, provides the suggestion of driver behavior, and auxiliary driver carries out operation judges.Hud new line display module can be by current lineThe speed of vehicle, if the judgement in lane-change state combines with the result of voice recognition and positioning, provides and drives in projectionSuggestion for operation is sailed, auxiliary driver carries out operation judges.
Fig. 2 gives the specific steps and process of technical solution of the present invention, comprising:
(1) ambient sound when being driven a vehicle by the acoustic sensor array acquisition being arranged in around body of a motor car.
(2) collected voice input preprocessing module is handled, including denoising, end-point detection, framing and adding window.Wherein denoising method is realized using wavelet threshold denoising, Selection of Wavelet Basis db3 small echo, and Decomposition order is set as 7 layers, denoisingMode selects wavelet soft-threshold denoising.
(3) pretreated signal identification module is input to be identified in real time.Sound recognition module includes importantEnvironment audio database and pattern matching unit.It is concluded in audio database and contains common police car, ambulance and disappearThese four sound are summarized as the important sound event of environment by the police whistle sound of anti-vehicle and the tucket of automobile, and as knowledgeOther object.Pattern matching unit can carry out characteristic parameter extraction to sound, and carry out mode with the data in audio databaseMatching, obtains recognition result.Extracted sound characteristic parameter is mel-frequency cepstrum coefficient and empirical mode decomposition energy ginsengNumber, can several important environment sound of Efficient Characterization feature, reach higher accuracy of identification.The pattern match link is adoptedIt is realized with the algorithm based on support vector machines, can rapidly and accurately realize the identification of sound.
(4) if recognition result show that, there are one of several critical environments sound of target, ECU controls auditory localization moduleCarry out the positioning of Sounnd source direction.Auditory localization module is the unit that audio direction is judged based on acoustic sensor array method.According to soundSensor array collects time difference and the phase difference of same sound, judges sound source group in the three-dimensional space centered on automobileAt the specific orientation in circle.
It is two-way communication between ECU electronic control module and sound source locating module, can control the operation of auditory localization moduleAnd receive the positioning result of auditory localization module.If detecting target sound, ECU controls the work of auditory localization module, but such asThe recognition result that fruit obtains is tucket, the maximum decibel value that 10 sonic transducers collect this tucket is first compared, if mostBig decibel value is less than 40dB, then ECU does not execute subsequent step, can filter out the woofer sound of distant place in this way, leave decibelIt is worth big tucket, and thinks it for important ambient sound.
(5) positioning result is transferred to ECU by auditory localization module, ECU control hud new line display module by recognition result andPositioning result is shown in together on the windshield immediately ahead of driver, while ECU combination vehicle speed sensor and steering wheel sensingThe data of device judge current vehicle condition, provide driver behavior suggestion in hud comes back and shows.ECU also controls steering wheel vibration moduleDriver is reminded, achievees the purpose that auxiliary drives.
In the step (5), ECU is connected with vehicle speed sensor and steering wheel angle sensor, according to voice recognition andPositioning as a result, in conjunction with current vehicle speed and whether be in lane-change state, provide the suggestion of driver behavior, driver assisted to graspWork judges.
The ECU integrated information and particular content of the driver behavior suggestion provided is as follows:
The vibration mode of steering wheel has following several:
When recognition result is three kinds of police whistle sounds, if steering wheel angle, less than 2 degree, steering wheel is with weak vibration mode vibration twoIt is secondary, if steering wheel angle be greater than 2 degree, steering wheel with strong vibration mode vibration twice；When recognition result is honk, ifFor steering wheel angle less than 2 degree, steering wheel is primary with the vibration of weak vibration mode, if steering wheel angle is greater than two degrees, steering wheel is with strongVibration mode vibration is primary.
The embodiment is a preferred embodiment of the present invention, but present invention is not limited to the embodiments described above, notIn the case where substantive content of the invention, any conspicuous improvement that those skilled in the art can make, replacementOr modification all belongs to the scope of protection of the present invention.