CN106933146A - Electrocar pedestrian's caution sound method for designing, caution sound control system and method - Google Patents

Abstract

The invention discloses a kind of electrocar pedestrian caution sound method for designing, differential responses according to people to different frequency sections sound, design substantial amounts of sample sound, and propose using the objective psychoacoustic parameter of sound quality and subjective evaluation result, set up by the BP neural network forecast model of genetic algorithm optimization, the forecast model can provide guidance for the screening of follow-up sample sound, simplify the screening process of sound, cost-effective；Actual operating state according to vehicle sets up the intelligence control system of multi-state, electric motor car/hybrid vehicle is set to be capable of the suitable sound of selection of intelligence under different operating conditions and send, also according to the multiplication factor of the intelligent selection gain circuitry of the size of ambient noise, sent warning acoustic energy is set more reasonably to warn pedestrian and not result in the noise pollution of environment.The present invention discloses a kind of pure electric or hybrid pedestrian caution sound method for designing and caution sound control system and its control method.

Description

Electrocar pedestrian's caution sound method for designing, caution sound control system and method

Technical field

The invention belongs to field of automobile safety, be related to it is a kind of for electrocar when running at a low speed to goers outside vehicle with police The automatic sounding system being shown as, and in particular to a kind of electrocar pedestrian caution sound method for designing, caution sound control system and Control method.

Background technology

Due to the consumption and the increase of fuel price of fossil fuel, promote China Automobile Industry focuses on new energy On automobile.Nearest a survey report of National Highway Traffic safety management bureau shows that electric motor car and traditional vehicle are knocked down The traffic accident of other road users such as pedestrian or bicycle occurs mainly in the low-speed region of road, contrast display, electric motor car With hybrid electric vehicle traditional internal-combustion engines vehicle is significantly larger than in the probability that low regime occurs traffic accident.Electric motor car and mixing are dynamic Power car, in low speed (i.e. speed is less than 30km/h) traveling, because motor radiated noise is smaller, is produced as new-energy automobile The outer sound of raw overall car when car is crossed, is difficult to be found, if honking suddenly than relatively low, it is easier to cause pedestrian by In alarmed and produce traffic accident, therefore it is to ensure pedestrains safety, alarming sound need to be added, points out pedestrian and other road user energy Enough vehicles for avoiding traveling in time, so as to reduce road traffic accident.

For prior art, pedestrian's caution system is all directly to send sound according to speed travel situations, is sent Sound major part be all that directly collection traditional vehicle sound is played out, be not related to the design of sound and selection, and just existing For caution system, sounding is directly carried out when sounding condition is reached without considering carrying out practically situation and the environment back of the body of vehicle The influence of scape noise, the sound that is sent is single, can not provide the operation conditions of vehicle to pedestrian well, not be to reach The purpose for warning pedestrian is exactly that noise pollution is caused to environment, and equal designed experiment is excessively single, it is impossible to good mould Intend actual conditions when vehicle is on the way travelled, it is impossible to which whether good detecting system can meet the requirement of design.

The content of the invention

The peace existed when the technical problems to be solved by the invention are directed to electric motor car or hybrid vehicle runs at a low speed Full hidden danger, prior art sound design and selection, the problems such as control system is excessively single, there is provided electrocar pedestrian's caution sound Method for designing, when electric motor car or hybrid electric vehicle run at a low speed, by the selection preprepared sample sound of intelligence, and root Play the intensity of sound to warn pedestrian or other road users according to the selection of the size intelligence of ambient noise, to remind it to have Effect avoids the vehicle of traveling.

Differential responses of the present invention according to people to different frequency sections sound, design substantial amounts of sample sound, and propose profit With the objective psychoacoustic parameter of sound quality and subjective evaluation result, set up and mould is predicted by the BP neural network of genetic algorithm optimization Type, the forecast model can provide guidance for the screening of follow-up sample sound, simplify the screening process of sound, cost-effective；Its The secondary actual operating state according to vehicle sets up the intelligence control system of multi-state, makes electric motor car/hybrid vehicle in difference It is capable of the suitable sound of selection of intelligence under operating condition and sends, also, in the controls also according to the big of ambient noise The multiplication factor of small intelligent selection gain circuitry, makes sent warning acoustic energy more reasonably warn pedestrian and not result in The noise pollution of environment；Finally, the outer Analysis of The Acoustic Fields of driving and real train test are entered to being mounted with intelligentized pedestrian's caution system, with The validity and reliability of checking system.

In order to solve the above technical problems, the present invention adopts the following technical scheme that realization：

The present invention provides a kind of electrocar pedestrian caution sound method for designing, it is characterised in that comprise the following steps：

Step one, using the outer voice signal sample of sound collection equipment collecting vehicle：Many same level traditional vehicles are gathered respectively With the outer voice signal of car of the pure electric vehicle under different operating modes；

Step 2, the outer voice signal of all cars gathered to the step one using wave filter extract the sound in special frequency channel Message number；

Step 3, the voice signal extracted to the step 2, are synthesized by different proportional components：In the step The composition of electric motor car voice signal, the special frequency channel that will be extracted are removed on the outer voice signal of the rapid one traditional vehicle car for being gathered Voice signal synthesized with gained reference voice signal with amplitude in different ratios；

Step 4, all sample sounds to the step one and step 3 acquisition enter under conditions of certain intensity of sound Row subjective assessment is tested；

Step 5, the calculating objective psychoacoustic parameter of sound quality and A weighted sound pressure levels；The objective psychoacoustic parameter of sound quality Including：Loudness, roughness, sharpness, shake degree, AI indexes；

Step 6, the voice signal to subjective assessment highest scoring in the step 4 enter the outer Analysis of The Acoustic Fields of driving, and sentence The disconnected requirement for whether reaching sound field coverage and intensity；

Step 7, set up neutral net valuation prediction models：Subjective assessment score and the step according to the step 4 The rapid five objective psychoacoustic parameter values being calculated, the BP neural network sound quality that foundation genetic algorithm optimization is crossed is evaluated pre- Survey model；

Step 8, set up electrocar pedestrian's caution sound control system.

Present invention simultaneously provides a kind of electrocar pedestrian caution sound control system, including：Micro-processor MCV, sound are broadcast Electric discharge road, storage circuit, hum measurement circuit, CAN interface circuit and power circuit；Power circuit is whole control System power supply, storage circuit, sound playing circuit, hum measurement circuit, CAN interface circuit respectively with microprocessor MCU is connected；

Micro-processor MCV is obtained vehicle speed and travel conditions and is obtained by hum measurement circuit by CAN The ambient noise in running environment is taken, according to the gain factor of the control amplifying circuit of the size intelligence of ambient noise, Ran Houxuan Select the sample sound in storage circuit and played out by sound playing circuit；

Sound playing circuit includes the D/A converter, amplifying circuit, the loudspeaker that are sequentially connected, and it is responsible for playing microprocessor The sample sound of device MCU screenings；

Storage circuit includes FLASH storage chips and RAM storage chips, and FALSH storage chips are responsible for storing what is filtered out Scoring highest voice data, RAM storage chips are responsible for the temporary and broadcasting sound of the voice data selected by Micro-processor MCV The storage of sound data；

CAN interface realizes level conversion with vehicle CAN bus interface, and speed or motor are read by Micro-processor MCV Rotating speed and hybrid vehicle type of drive information judges whether to need to open caution control system, upon power-up of the system, The sound to be played will be selected according to running state of the vehicle；

Hum measurement circuit is measured using noise testing sensor to the ambient noise residing for vehicle, and will be surveyed Amount result is supplied to Micro-processor MCV, to control the gain range of amplifying circuit in sound playing circuit；

Power circuit uses DC/DC dc source chips, for each circuit of system provides different voltages.

Present invention simultaneously provides a kind of control method of electrocar pedestrian caution sound control system：

1) Micro-processor MCV utilizes the status information that vehicle is read with vehicle CAN bus interface circuit, using ambient noise The size of the ambient noise measured by measurement sensor, with reference to all kinds of sample sounds being previously stored in memory, by D/A Converter exports simulated audio signal, and amplified circuit is exported by loudspeaker, so as to select sample sound and gain size, passed through The difference of institute's playing sound samples and the difference of acoustic gain carry out real-time table up to the operation conditions of vehicle；

3) linear relationship set up between speed and voice signal time domain：GES is read by CAN, works as speed During less than 10km/h, Micro-processor MCV is chosen 10km/h scoring highest sample sounds and is played out, and speed is in 10km/h- When between 20km/h, the 20km/h scoring highest sample sounds that Micro-processor MCV chooses storage are played out, and speed exists When between 20km/h-30km/h, the sample sound that Micro-processor MCV chooses the 30km/h storages of storage is played out；During reversing, Micro-processor MCV plays the sample sound during 30km/h of storage；

4) ambient noise measured by hum measurement sensor is utilized, the different intelligent according to background noise level The grade of gain in selection amplifying circuit：Ambient noise is divided into Three Estate, respectively less than 55dB, 55dB to 65dB and big In 65dB, according to different background noise levels, different gain factors are selected.

Compared with prior art the beneficial effects of the invention are as follows：

1st, form first by inquiry understands the acceptable alarming sound form of people, and is determined by experiment different people Then group recycles matlab to carry out the design of rational sample sound the sensitivity of different frequency sections sound；

2nd, with sound quality theory in vibration noise as foundation, by sound quality with the subjective assessment of irritated degree grade classification Experiment, gives a mark to substantial amounts of sample sound, selects fraction highest sample sound as the sample sound for needing, it is theoretical according to According to strong；

3rd, the present invention is prototype to calculate the objective psychoacoustic parameter and subjective evaluation result of gained, sets up GA-BP god Through assessing network model, the evaluation result of model reaches 0.93 with the coefficient correlation of actual subjective assessment value, with weight very high Renaturation, will provide Appreciation gist for follow-up sound design, and subjective assessment is carried out without establishment officer again, save a large amount of Manpower and materials；

4th, the present invention can carry out sounding, closing to reality vehicle fortune according to the selection alarming sound of different speed situation intelligence Row situation, facilitates the operation conditions of pedestrian determination vehicle；

5th, the ambient noise of environment is taken into full account in the present invention, using the back of the body in hum measurement sensor measuring environment Scape noise, so that the gain level of the distribution sound of intelligence, is at utmost reduced and made to environment by the alarming sound for being added Into pollution；

6th, the present invention is by the reliability of test of many times checking system, and by normal road with the anti-of normal pedestrian Should be used as feedback to verify reliability during caution system real work, more closing to reality situation, be that the validity of system is carried For the guidance of science.

Brief description of the drawings

The present invention is further illustrated below in conjunction with the accompanying drawings：

Fig. 1 is the overview flow chart of electrocar pedestrian caution sound method for designing of the present invention；

Fig. 2 is the schematic diagram of collection traditional vehicle of the present invention and pure electric vehicle pass-by noise recording location；

Fig. 3 is of the present invention to measure designed sound in sound herein with sound level meter at the outer minimum range scope of car The experiment schematic diagram of size of arbitrarily downgrading；

Fig. 4 is the structure chart of the neural network model of performance appraisal of sound designed by foundation of the present invention；

Fig. 5 is predicting the outcome for the neutral net valuation prediction models of the sample sound designed by foundation of the present invention With the percentage error figure between actual subjective assessment marking result；

Fig. 6 is predicting the outcome for the neutral net valuation prediction models of the sample sound designed by foundation of the present invention With the dependency diagram between actual subjective assessment marking result；

Fig. 7 is that electrocar pedestrian caution sound control system hardware of the present invention is constituted and theory diagram；

Fig. 8 is the course of work flow chart of electrocar pedestrian caution sound control system of the present invention；

Fig. 9 is the ambient noise sample circuit of electrocar pedestrian caution sound control system of the present invention；

Figure 10 is DAC conversions and the power amplification circuit of electrocar pedestrian caution sound control system of the present invention；

Figure 11 is the global design scheme block diagram of electrocar pedestrian caution sound control system of the present invention.

Specific embodiment

The present invention is explained in detail below in conjunction with the accompanying drawings：

Spirit of the invention is to make traveling to solve electrocar excessively quiet car external environment when running at a low speed Vehicle is difficult the problem being found, to reduce traffic accident, it is necessary to add alarming sound to point out pedestrian and other road users The vehicle of traveling is avoided in time, it is ensured that safe passing.Simultaneously present invention is alternatively directed to different crowd to sound in different frequency sections The difference of sensitivity carries out the design of alarming sound and also introduces the concept of sound quality carrying out the selection of sound, by sound quality Objective psychoacoustic parameter calculate and subjective agitation degree evaluation selection correspondence speed under scoring highest sample sound as electricity Alarming sound in motor-car correspondence vehicle speed intervals, and the nerve set up between objective psychoacoustic parameter and subjective agitation degree score value Assessing network forecast model, the predicted value of model has correlation very high with actual score value, will be the design of follow-up sound Selection provides theoretical direction, so as to save substantial amounts of manpower and material resources.And the control system designed by the present invention being capable of basis The suitable sample sound of selection of different speed intelligents as alarming sound, also with the ability of real-time detection ambient noise, can With the multiplication factor of the control gain circuitry of the different intelligent according to ambient noise magnitude, it is ensured that alarming sound is same prompting pedestrian's When can not produce noise pollution.

One, electrocar pedestrian caution sound methods for designing of the present invention, equipment and computational methods used is as follows：

1. the sound collection equipment described in is microphone, as Mike, is recorded using special recording software；

2. the wave filter described in is the fdesign+design bandpass filters using MATLAB designs, is set using MATLAB The digital filter of meter can very easily call and change design parameter therein, can very easily select to filter frequency range；

2. the sound level described in is calculated as small wild precision sound level meter；

3. the valuation prediction models described in are using the valuation prediction models of GA-BP neural networks：(1) nerve is determined The network number of plies；(2) the implicit number of plies is determined；(3) neural network topology structure is determined；

4. by Micro-processor MCV, sound playing circuit, storage circuit, (FLASH storages, RAM are deposited the control system described in Storage), the composition such as CAN interface circuit and power circuit.

Two, electrocar pedestrian caution system control methods of the present invention

It is as follows the step of electrocar pedestrian caution sound method for designing of the present invention refering to Fig. 1 to Fig. 9：

Step 1：Using the sound collection equipment outer voice signal sample of collection traditional vehicle car；

Refering to Fig. 2,5 same level traditional vehicles of collection operating mode different with pure electric vehicle is got off outer voice signal, selected The operating mode selected respectively 10km/h operating modes, 20km/h operating modes and 30km/h operating modes；

The outer voice signal sample of traditional vehicle car is gathered described in technical scheme in step 1, is comprised the following steps that：

(1) environmental selection is recorded in semianechoic room, using the recording microphone of specialty, and is recorded by professional software Sound, the form of storage is .wav；

(2) recording microphone wind reference by location measure exhaust noise method, select on longitudinal direction of car center line, with away from Any position on circular arc, drawn circular arc or so 45 ° of camber lines is drawn from the outstanding 0.5m of Chinese herbaceous peony, microphone, the height ginseng of microphone is placed Acceleration noise measuring method is examined, is selected in the altitude range of 0.9m-1.2m,；

(3) record the pass-by noise signal of five traditional vehicles and a pure electric vehicle respectively in semianechoic room rotating hub, record The noise signal of system includes：10km/h operating modes, 20km/h operating modes and 30km/h operating modes.It is 5s in the middle of the sound interception that will be recorded Part, the voice signal that will have been intercepted is numbered record, to facilitate subsequent treatment.

Step 2：The voice signal in the outer voice signal special frequency channel of car is extracted using wave filter；

In the case of refering to great mass of data and document, the alarming sound shape that people are easily accepted by is collected using the form of survey Formula, show that people are more prone to the voice signal of similar traditional vehicle by substantial amounts of data analysis.Meanwhile, experiment display is different Crowd is different in the reaction of different frequency composition to same sample sound, as a result show elderly population to 600Hz near Sound it is more sensitive, it is stronger to the ability of discovery of the neighbouring sound of 2500Hz with normal-hearing people, and near 1000Hz Sound noise pollution easily is caused to environment.According to content described above, the sound frequency section extracted is respectively 500- 700Hz、900-1100Hz、2400-2600Hz；

The voice signal in the outer voice signal special frequency channel of car is extracted using wave filter in step 2 described in technical scheme, Comprise the following steps that：

(1) design bandpass filter, it is contemplated that higher using frequency, with reference to the performance of various wave filters, utilize Fdesign+design designs digital filter in matlab, designed wave filter meets design requirement, can obtain very well Effect.Filter stop bend decays to 80dB, and passband ripple is 3dB.

(2) voice signal, it is necessary to extract is filtered to all voice signals for gathering using designed wave filter Frequency separation be 500-700Hz, 900-1100Hz, 2400-2600Hz.And by each car correspondence speed under filtered letter Number composition is stored.

Step 3：To the voice signal for being extracted, synthesized by different proportional components；

Know that different crowd is different to the sensitivity of each frequency band sound, what is gathered according to investigation and experiment Removed on traditional vehicle voice signal on the component base of electric motor car voice signal, the voice signal of the special frequency channel that will be extracted Synthesized with gained reference voice signal with amplitude in different ratios, obtained substantial amounts of voice signal sample

The voice signal for being extracted is synthesized, specifically by different proportional components in step 3 described in technical scheme Step is as follows：

(1) due to being that alarming sound is added on pure electric vehicle, although sound very little or almost when electric motor car runs at a low speed There is no sound, but the influence of the acoustic constituents of this part still should be considered when alarming sound is added to it, it is also possible to Referred to as ambient noise, it is therefore desirable to adopted when electric motor car correspondence speed is removed in the traditional vehicle original sound composition for being gathered The acoustic constituents for collecting, obtain new sample sound data, referred to as reference voice signal；

(2) on a reference voice signal for speed with a vehicle, will be right with 0.1,0.2,0.3,0.4,0.5 times The sound filtered in resulting 500-700Hz and 2400-2600Hz frequency ranges under vehicle and speed is answered to be added to reference sound intensity message On number, and sound in 0.1,0.2,0.3,0.4,0.5 times of 900-1100Hz frequency ranges is subtracted on the basis of reference voice signal Signal.

(3) by the treatment of a series of sample sound, 40 sample sounds will be obtained under each speed, and altogether can Obtain 120 sample sound signals.

Step 4：The all of sample sound obtained to the step one and step 3 is under conditions of certain intensity of sound Carry out subjective assessment experiment；

In Appraising subject quantity, the evaluation test of most of psychologic acoustics, the subjective assessment person of more than 20 just can be with Draw more accurately acoustic assessment result.Based on this principle, it is subjective as sound quality that the present invention chooses 30 testing crews Appraising subject, M-F is 2.3 in Appraising subject:1, between 24~56 years old, there is the test evaluation people of driving experience at the age The quantity of member is in the great majority.By being familiar with test site environment in one's power, and explain the training such as implication of experiment process and appraised index Composition of personnel jury of acting as an assessor afterwards carries out sound quality subjective assessment to designed sample sound, with irritated degree in subjective assessment Grade is evaluation index, carries out subjective assessment experiment；

All sample sounds described in technical scheme acquired in step 4 pair enter under conditions of certain sound intensity level Row subjective assessment is tested, and is comprised the following steps that：

(1) jury member is chosen：30 evaluation personnels are chosen in experiment altogether, and these members are by driver, acoustics expert and many It is engaged in technical staff and the student composition of automobile vibration and noise operation year, is respectively provided with certain evaluation experience, wherein men and women point Not Wei 21 people and 9 people, the age, hearing was normal between 24~56 years old, evaluation environment be subjective assessment room.

(2) sound quality evaluation index is determined：This evaluation test with irritated degree grade as evaluation index, using grade scoring method Each sample sound is given a mark.Sound quality agitation degree is divided into 11 grades according to international standard, it is as shown in table 1 below.Evaluate Person hears different sound successively, and clicks on corresponding digital button in scoring device respectively according to the subjective feeling of oneself.

The subjective assessment agitation degree grade table of comparisons of table 1.

It is very bad	It is bad	It is very poor	Difference	It is dissatisfied	It is acceptable	It is satisfied	Preferably	It is good	Very well	It is fabulous
											1	2	3	4	5	6	7	8	9	10	11

(3) coefficient correlation between each estimator is calculated：Must be carried out using the evaluation result obtained by grade scoring method Data detection, calculates the coefficient correlation between each estimator, with the validity of inspection data.Use SPSS applied statistics softwares pair The grade scoring result of all estimators is analyzed, and calculates their coefficient correlations between any two, then by each estimator Arithmetic mean of instantaneous value is taken with the coefficient correlation between other estimators, average correlation coefficient is tried to achieve, boldface as shown in table 2 is related Property it is relatively low, rejected, then remaining evaluation result is given averagely, you can subjective assessment sound quality value.

Coefficient correlation is evaluated between the estimator of table 2.

Estimator	1	2	3	4	5	6	7	8
									Coefficient correlation	0.805	0.763	0.821	0.633	0.749	0.779	0.584	0.814
Estimator	9	10	11	12	13	14	15	16
									Coefficient correlation	0.790	0.765	0.809	0.619	0.759	0.782	0.831	0.827
Estimator	17	18	19	20	21	22	23	24
									Coefficient correlation	0.793	0.605	0.784	0.816	0.807	0.737	0.791	0.761
Estimator	25	26	27	28	29	30
									Coefficient correlation	0.812	0.794	0.817	0.775	0.781	0.802

Step 5：Calculate the objective psychoacoustic parameter of sound quality and A weighted sound pressure levels；

The objective psychoacoustic parameter of described sound quality includes：Loudness, roughness, sharpness, shake degree, AI indexes, and Calculate A weighted sound pressure levels；

The objective psychoacoustic parameter of sound quality and A weighted sound pressure levels, specific steps are calculated described in technical scheme in step 5 It is as follows：(1) loudness is calculated：

People is described to the subjective sensation of sound intensity with loudness, and the size of loudness represents the sonority of sound.It It is one of most important psychologic acoustics objective parameter, unit is Song (sone).Frequency is the pure tone of 1kHz, if its acoustic pressure is 40dB, then then its loudness is 1sone.

In formula：N ' is characterized loudness, and unit is sone, and E is sound stimulation, E_TQIt is absolute subaudible excitation, E₀It is base Excitation under the quasi- sound intensity；

Integral and calculating is carried out on total Bark domains to N ', you can obtain total loudness N：

(2) roughness is calculated

Roughness is sound modulation amplitude size, the reflection of modulating frequency low feature high, is to characterize sound degree of modulation Parameter, unit is asper.Using the modulating frequency f of noise_modWith the differential Δ L of excitation in individual feature band_E(z) (namely Sound shelters depth) calculate roughness, i.e.,：

In formula, R represents roughness, and unit is asper, f_modIt is modulating frequency, Δ L_EZ () is the change of acoustical signal drive(r) stage Change amount, is defined as：

In formula：N′_max(z) and N '_minZ () represents characteristic loudness maximum and minimum value respectively；

(3) sharpness is calculated

Sharpness reflects the ear-piercing degree of voice signal, and as a rule, sharpness more high sound is more ear-piercing, and people are to sound The sensation of sound is more worried, and sound quality is also poorer, and it is acum that its unit is.Definition centre frequency be 1kHz, with a width of 160Hz's The sharpness of 60dB narrow-band noises is 1acum.

Sharpness S is calculated using Zwicker models, its Mathematical Modeling is based on Scale Model of Loudness, and mathematical formulae is as follows：

In formula, k is weight coefficient (in general k takes 0.11)；N is total loudness value；N ' (z) is represented in z Bark domains Characteristic loudness；In the sharpness computation model of Zwicker, g (z) is weight coefficient of the voice signal in different Bark domains, Its expression formula is：

(4) shake degree is calculated

Shake degree is the subjective sensation for describing human ear to slow modulation sound, and its size represents the fluctuating of sound intensity Degree, unit is vacil.It is the pure tone of 1kHz for frequency, if its sound pressure level is 60dB, by the frequency modulation(PFM) of 4Hz, amplitude Modulation rate be 100% after, then its shake degree be 1vacil.

Zwicker shake degree computation models are：

Wherein, F is shake degree, f₀Represent modulation fundamental frequency (f₀Take 4Hz)；f_mod、ΔL_EThe definition and calculating of (z) and roughness It is identical.

(5) speech articulation

Speech intelligibility index is also AI indexes (articulation index), instead for determining voice annoyance level Reflect due to the parameter that lamprophonia degree when noise reason causes talk changes.

Speech articulation computation model is：

In formula, W (f) is weighted coefficient, and N (f) reaches the standard grade for ambient noise, and its expression formula is UL (f)=H (f)+12dB, LL F () is that ambient noise is offline, its expression formula is LL (f)=UL (f) -30dB.

(6) A weighteds sound pressure level

A sound levels are defined as with sound level meter or with the measuring instrument equivalent with this, by the sound letter of A weighting networks measurement Number noise grade, unit with dB (A) represent, the objective psychoacoustic parameter value in selected part and subjective assessment score value are such as Shown in table 3 below.

The subjective and objective result of the sample of table 3

Step 6：The outer Analysis of The Acoustic Fields of driving is entered to the highest voice signal that scored in each group, and judges whether to reach sound field covering The requirement of scope and intensity；

Refering to Fig. 3, black round dot is the position for measuring the small wild precision sound level meter used by sound pressure level in figure, in semianechoic room Inside enter the outer Analysis of The Acoustic Fields of driving, the coverage and intensity of sound of sound field in the angular region of 45 degree of the axis both sides in Main Analysis Chinese herbaceous peony portion Situation of change；And judge whether sound field reaches covering requirement in the range of audibility for ensureing pedestrains safety distance, if do not had Reaching requirement then answers return to step 4 to adjust the reference intensity of sound；

Enter the outer Analysis of The Acoustic Fields of driving, specific step described in technical scheme in step 6 to the highest voice signal that scored in each group It is rapid as follows：

(1) according to Motor Vehicle Braking Procedure, the reaction time for seeing pedestrian from driver brakes completely to automobile, this period The distance that interior automobile is gone should be required minimum safe distance S, braking Distance Calculation formula during according to automobile brake, calculate most Lower security is apart from S：

In formula, S represents braking distance, τ₁Time of driver's reaction is represented, 0.3~1.0s, τ ' is typically taken₂+τ″₂Referred to as make Dynamic device action time, u_a0It is initial speed, 30km/h, a is taken when calculating herein_bmaxIt is maximum braking deceleration.

(2) three sample sounds of subjective assessment highest scoring in 3 groups of sample sounds are utilized into electric motor car in semianechoic room Inside play out, loudspeaker are positioned over the position similar to transmission car engine during broadcasting；

(3) with sound level meter measure front side axis or so 45° angle with minimum safe distance S as radius drawn by it is any on camber line A weighted sound pressure levels at a little, and keep a record；

(4) A weighted sound pressure level of any one same level traditional vehicle at S distances is measured in same position, and records these A weighted sound pressure levels at point；

(5) compare the A sound level differences measured twice in (3), (4), answer return to step 4 to adjust reference sound intensity if difference is excessive Compressive Strength, when difference of them is little, then it represents that selected sound can reach the level of the outer sound of equal traditional vehicle car, also just have There is good warning pedestrian；

Step 7：Set up neutral net valuation prediction models；

Refering to the objective psychoacoustic parameter that Fig. 4, subjective assessment institute's score value and step 5 according to step 4 are calculated Value, the BP neural network sound quality valuation prediction models that foundation genetic algorithm optimization is crossed, the topological structure of neutral net is 6- 1- hidden layer neurons 1 in 11-1, Fig. 4,2- hidden layer neurons 2, j- hidden layer neurons j, 11- hidden layer neuron 11,- input layer to hidden layer connection weight,Connection weight of-the hidden layer to output layer；

Neutral net valuation prediction models are set up described in technical scheme in step 7, is comprised the following steps that：

(1) neural network topology structure is determined, it is that 11, i.e. its structure are 6- to select hidden layer neuron number through repeatedly training 11-1, determines the related parameter values in genetic algorithm, using genetic algorithm come Optimized BP Neural Network, sets up GA-BP nerve nets Network model；

(2) by 104 samples in 120 sample sounds as training sample, remaining 16 sample sounds are as inspection Sample is tested, gained testing result reaches 96.3% with the correlation of actual appraisal result after training is finished, it was demonstrated that set up network Accuracy, can be used to predict the subjective assessment score of follow-up sample sound, Fig. 5 is the mistake between predicted value and actual score value Difference percentage, Fig. 6 represents the correlation between predicted value and actual score value, the GA-BP neutral net evaluation and forecas set up The weights and threshold value of model are as shown in table 4 below；

The weights and threshold value of table 4GA-BP valuation prediction models

Step 8：Intelligentized control system is set up according to control program；

Refering to Fig. 7, Control System Design scheme is determined according to vehicle Real-road Driving Cycle, i.e., when speed is less than 30km/h System is in opening, and when speed is more than 30km/h, system is closed automatically, and system is in and opens shape during reversing State.Also, system can be according to the sample sound to be played of selection of vehicle speed range intelligence, while can also be made an uproar according to background The multiplication factor of the control gain circuitry of the different intelligent of sound intensity；

After selected sound is determined, the specific work process of control system is done into one with reference to accompanying drawing 7 and Fig. 8 The explanation of step.If Fig. 7 is the hardware composition part of system, MCU receives the car status information provided from CAN, MCU The running status of the vehicle according to vehicle, when speed is less than 30km/h, would automatically turn on system, and according to specific speed selection The sample sound stored in FLASH and RAM, while ambient noises of the MCU according to measured by hum measurement sensor is determined The gain conditions of amplifying circuit -6.All modules carry out the distribution of voltage by power circuit -9.Whole work process such as Fig. 8 institutes Show, i.e. the running state of the vehicle that MCU is provided according to CAN transceiver and in-car CAN, from memory FLASH and RAM The middle suitable sample sound of selection, the road environment noise intensity situation obtained according to hum measurement sensor, by MCU Be transferred to amplifying circuit to select suitable gain factor carries out gain to intensity of sound, and sound playing circuit is then passed through again In D/A converter, amplifying circuit, loudspeaker carry out the broadcasting of sound, to reach the effect of warning pedestrian.

Intelligentized electrocar pedestrian caution sound control system is set up according to control program in step 8 described in technical scheme System, comprises the following steps that：

(1) the automobile mounted intelligent pedestrian's caution control system of pure electric/hybrid is broadcast by Micro-processor MCV, sound The compositions such as electric discharge road, storage circuit, CAN interface circuit and power circuit, as shown in Figure 7 to 10, power circuit is whole Individual control system is powered, and storage circuit, sound playing circuit, CAN interface circuit are connected with Micro-processor MCV respectively；Deposit Storing up electricity road includes FLASH storage chips and RAM storage chips；Sound playing circuit includes the D/A converter being sequentially connected, amplifies Circuit and loudspeaker, D/A converter are connected with Micro-processor MCV.

Wherein：

1) Micro-processor MCV uses TMS320F28xx series DSPs, and vehicle speed and travel conditions are obtained by CAN And the ambient noise in running environment is obtained by hum measurement sensor, according to the control of the size intelligence of ambient noise The gain factor of amplifying circuit, then the sample sound in intelligent selection storage circuit and is broadcast by sound playing circuit Put；

2) sound playing circuit：Including D/A converter, amplifying circuit (adjustable gain), loudspeaker, it is responsible for playing DSP sieves The sample sound of choosing；

3) storage circuit：The program space is extended out using FALSH storage chips, the scoring highest sound that responsible storage is filtered out Sound data, data space is extended out using RAM storage chips, the temporary and broadcasting of the voice data being responsible for selected by DSP The storage of voice data；

4) CAN interface：Using CAN transceiver, level conversion is realized with vehicle CAN bus interface, read by DSP Fast or motor speed (hybrid vehicle also needs to read type of drive) information is picked up the car to judge whether to need to open intelligent row People's caution system, upon power-up of the system, by according to running state of the vehicle come the sound to be played of intelligent selection；

5) hum measurement circuit：The ambient noise residing for vehicle is measured using noise testing sensor, and Measurement result is supplied to DSP, to control the gain range of amplifying circuit in sound playing circuit；

6) power circuit：Using DC/DC dc source chips, for each chip of system provides different voltages.

(2) refering to Fig. 8, DSP is made an uproar using the status information with vehicle CAN bus interface circuit reading vehicle using background The size of the ambient noise measured by sounding quantity sensor, with reference to all kinds of sample sounds being previously stored in memory, passes through D/A converter exports simulated audio signal, and amplified circuit is exported by loudspeaker, so as to intelligently select sample sound and gain Size, can realize being expressed in real time by the difference of institute's playing sound samples and the difference of acoustic gain the operation shape of vehicle Condition；

(3) linear relationship set up between speed and voice signal time domain, i.e., read GES by CAN, when When speed is less than 10km/h, the selection 10km/h scoring highest sample sounds of DSP intelligence are played out, and speed is in 10km/h- When between 20km/h, the 20km/h scoring highest sample sounds of the selection storage of DSP intelligence are played out, and speed is in 20km/ When between h-30km/h, the sample sound of the 30km/h storages of the selection storage of DSP intelligence is played out.During reversing, DSP is played The sample sound during 30km/h of storage；

(4) using the ambient noise measured by hum measurement sensor, according to the different intelligent of background noise level Selection amplifying circuit in gain grade.Ambient noise is divided into Three Estate, respectively less than 55dB, 55dB to 65dB and More than 65dB, according to different background noise levels, vehicle mounted intelligentized pedestrian caution system will select different gain factors, with Reaching can not only warn pedestrian to avoid the purpose of driving vehicle in time under any environmental condition, to reduce the hair of traffic accident It is raw, the harm to pedestrian is directly reduced, while sent sound can also be made also will not indirectly to be reduced as ambient noise The infringement physically and mentally healthy to people, this function be by the MCU in such as Fig. 7, hum measurement sensor, amplifying circuit come Realize.

Step 9：System to establishing carries out real train test.

System that should be to being set up after whole system is set up and finished carries out train experiment, carrys out the validity of checking system And reliability, the content of train experiment includes：Traveling finds experiment, turning perceives experiment, reversing experiment and roadway experiment.

Real train test is carried out to the system that establishes in step 9 described in technical scheme, is comprised the following steps that：

System that should be to being set up after whole system is set up and finished carries out real train test, carrys out the validity of checking system And reliability, the content of real train test includes：Traveling finds experiment, turning perceives experiment, astern trial and actual road test.

(1) traveling finds that experiment is that, to verify the distinguishability of audio direction, this experiment is carried out on special test site, Testing crew judged by sound automobile be by which direction running come；

(2) it is that testing crew judges that vehicle is to continue with straight-line travelling or turns at crossing by sound to turn and perceive experiment Curved traveling, this experiment is also to be carried out on special test site；

(3) astern trial is that testing crew judges that vehicle moves forward or reversing back traveling by sound, secondary Experiment is carried out on special test site；

(4) actual road test is that checking is equipped with time vehicle of intellectuality pedestrian's caution system when normal road is travelled by pedestrian It was found that possibility, this experiment is carried out in the more complicated block of road condition, in-car have special personnel to observe and remember Record the performance situation of goers outside vehicle.87 pedestrians are recorded in experiment altogether, wherein there are 72 potential energies enough accurately to recognize that vehicle is gone forward side by side Avoidance is gone, discrimination has reached 82.7%, it was demonstrated that the feasibility of set meter systems.

Claims (10)

1. a kind of electrocar pedestrian caution sound method for designing, it is characterised in that comprise the following steps：

Step one, using the outer voice signal sample of sound collection equipment collecting vehicle：Many same level traditional vehicles and one are gathered respectively The outer voice signal of car of the pure electric vehicle under different operating modes；

Step 2, the outer voice signal of all cars gathered to the step one using wave filter are extracted the sound in special frequency channel and believed Number；

Step 3, the voice signal extracted to the step 2, are synthesized by different proportional components：In the step one The composition of electric motor car voice signal, the sound of the special frequency channel that will be extracted are removed on the outer voice signal of traditional vehicle car for being gathered Message number is synthesized with amplitude in different ratios with gained reference voice signal；

Step 4, all sample sounds to the step one and step 3 acquisition are led under conditions of certain intensity of sound See evaluation test；

Step 5, the calculating objective psychoacoustic parameter of sound quality and A weighted sound pressure levels；The objective psychoacoustic parameter bag of sound quality Include：Loudness, roughness, sharpness, shake degree, AI indexes；

Step 6, the voice signal to subjective assessment highest scoring in the step 4 enter the outer Analysis of The Acoustic Fields of driving, and judgement is The no requirement for reaching sound field coverage and intensity；

Step 7, set up neutral net valuation prediction models：Subjective assessment score and the step 5 according to the step 4 The objective psychoacoustic parameter value being calculated, the BP neural network sound quality evaluation and foreca mould that foundation genetic algorithm optimization is crossed Type；

Step 8, set up electrocar pedestrian's caution sound control system.

2. according to the electrocar pedestrian's caution sound method for designing described in claim 1, it is characterised in that the step one is specific Step is as follows：

1) environmental selection is recorded in semianechoic room, and the form of storage is .wav；

2) method that the reference by location of recording microphone wind measures exhaust noise, selects on longitudinal direction of car center line, with apart from car Front overhang 0.5m draws any position on circular arc, drawn circular arc or so 45 ° of camber lines, places microphone, and the elevation references of microphone add Fast noise measuring method, selects in the altitude range of 0.9m-1.2m；

3) the pass-by noise signal of five traditional vehicles and a pure electric vehicle is recorded respectively in semianechoic room rotating hub, recording is made an uproar Acoustical signal includes：10km/h operating modes, 20km/h operating modes and 30km/h operating modes；It is the portion of 5s in the middle of the sound interception that will be recorded Point, the voice signal that will have been intercepted is numbered record.

3. according to the electrocar pedestrian's caution sound method for designing described in claim 1, it is characterised in that the step 2 is specific Step is as follows：

1) bandpass filter is designed, using fdesign+design designs digital filter in matlab, filter stop bend decay It is 80dB, passband ripple is 3dB.

2) sound, it is necessary to extract is filtered to all voice signals that the step one is gathered using designed wave filter The frequency separation of signal is 500-700Hz, 900-1100Hz, 2400-2600Hz, and by each car after filtering under correspondence speed Signal component stored.

4. according to the electrocar pedestrian's caution sound method for designing described in claim 1, it is characterised in that the step 3 is specific Step is as follows：

1) remove what is collected during electric motor car correspondence speed in the traditional vehicle original sound composition that the step one is gathered Acoustic constituents, obtain new sample sound data, referred to as reference voice signal；

2) on a reference voice signal for speed with a vehicle, will correspondence car with 0.1,0.2,0.3,0.4,0.5 times The sound in 500-700Hz and 2400-2600Hz frequency ranges obtained by being filtered under type and speed is added on reference voice signal, And voice signal in 0.1,0.2,0.3,0.4,0.5 times of 900-1100Hz frequency ranges is subtracted on the basis of reference voice signal.

5. according to the electrocar pedestrian's caution sound method for designing described in claim 1, it is characterised in that the step 4 is specific Step is as follows：

1) jury member is chosen, evaluation environment is subjective assessment room；

2) sound quality evaluation index is determined：With irritated degree grade as evaluation index, each sample sound is entered using grade scoring method Row marking；Sound quality agitation degree is divided into 11 grades according to international standard, estimator hears different sound successively, and according to The subjective feeling of oneself is given a mark in scoring device respectively；

3) coefficient correlation between each estimator is calculated：Using SPSS applied statistics software to the grade scoring knot of all estimators Fruit is analyzed, and calculates their coefficient correlations between any two, then by the phase relation between each estimator and other estimators Number takes arithmetic mean of instantaneous value, tries to achieve average correlation coefficient, then remaining evaluation result is given averagely, you can obtain the sound product of subjective assessment Matter value.

6. according to the electrocar pedestrian's caution sound method for designing described in claim 1, it is characterised in that the step 5 is specific Step is as follows：

1) loudness is calculated：

$N^{\′}=0.08{\left(\frac{E_{TQ}}{E_0}\right)}^{0.28}\[{(0.5+0.5\frac{E}{E_{TQ}})}^{0.23}-1\]$

In formula：N ' is characterized loudness, and unit is sone, and E is sound stimulation, E_TQIt is absolute subaudible excitation, E₀On the basis of sound Excitation under strong；

Integral and calculating is carried out on total Bark domains to N ', you can obtain total loudness N：

$N=\underset{0}{\overset{24Bark}{\∫}}{N}^{\′}\left(z\right)dz\left(sone\right)$

2) roughness is calculated：

$R=0.3f_{\mathrm{mod}}\underset{0}{\overset{24Bark}{\∫}}{\mathrm{\ΔL}}_E\left(z\right)dz\left(asper\right)$

In formula, R represents roughness, and unit is asper, f_modIt is modulating frequency,

${\mathrm{\ΔL}}_E\left(z\right)=20{\log }_{10}\left(\frac{{N^{\′}}_{\max }\left(z\right)}{{N^{\′}}_{\min }\left(z\right)}\right)$

In formula：N′_max(z) and N '_minZ () represents characteristic loudness maximum and minimum value respectively；

3) sharpness is calculated

$S=k\frac{{\∫}_0^{24Bark}{N}^{\′}\left(z\right)\·z\·g\left(z\right)dz}{N}\left(acum\right)$

In formula, S is sharpness, and k is weight coefficient (k takes 0.11)；N is total loudness value；N ' (z) represents the spy in z Bark domains Levy loudness；

$g\left(z\right)=\left\{\begin{array}{cc}1& z\≤16\\ 0.0625\·e^{0.1733\·z}& z>16\end{array}\right.$

4) shake degree is calculated

$F=0.008\frac{{\∫}_0^{24Bark}{\mathrm{\ΔL}}_E\left(z\right)dz}{(f_{\mathrm{mod}}/f_0)+(f_0/f_{\mathrm{mod}})}\left(vacil\right)$

Wherein, F is shake degree, f₀Represent modulation fundamental frequency；f_mod、ΔL_EZ the definition and calculating of () are identical with roughness；

5) speech articulation

Speech articulation computation model is：

$AI=\Σ\frac{W\left(f\right)D\left(f\right)}{30}$

$\left\{\begin{array}{cc}D\left(f\right)=0,& N\left(f\right)>UL\left(f\right)\\ D\left(f\right)=UL\left(f\right)-N\left(f\right),& LL\left(f\right)<N\left(f\right)<UL\left(f\right)\\ D\left(f\right)=30,& N\left(f\right)<LL\left(f\right)\end{array}\right.$

In formula, W (f) is weighted coefficient, and N (f) reaches the standard grade for ambient noise, and its expression formula is UL (f)=H (f)+12dB, and LL (f) is Ambient noise is offline, and its expression formula is LL (f)=UL (f) -30dB；

6) A weighteds sound pressure level

A sound levels are defined as with sound level meter or with the measuring instrument equivalent with this, the voice signal measured by A weighting networks Noise grade, unit is dB (A).

7. according to the electrocar pedestrian's caution sound method for designing described in claim 1, it is characterised in that the step 6 is specific Step is as follows：

1) according to Motor Vehicle Braking Procedure, the reaction time for seeing pedestrian from driver brakes completely to automobile, vapour in this period The distance of Che Suohang should be required minimum safe distance S, braking Distance Calculation formula during according to automobile brake, calculate minimum peace Full distance S：

$S={\mathrm{\&tau;}}_1{u}_{a0}+\frac{1}{3.6}({{\mathrm{\&tau;}}^{\&prime;}}_2+\frac{{{\mathrm{\&tau;}}^{\&prime;\&prime;}}_2}{2})u_{a0}+\frac{{u^2}_{a0}}{25.92a_{b\max }}$

In formula, S represents braking distance, τ₁Time of driver's reaction is represented, 0.3~1.0s, τ ' is taken₂+τ″₂Referred to as brake effect Time, u_s0It is initial speed, 30km/h, a is taken when calculating herein_bmaxIt is maximum braking deceleration；

2) several sample sounds of subjective assessment highest scoring in sample sound are broadcast using electric motor car in semianechoic room Put, loudspeaker are positioned over the position similar to transmission car engine during broadcasting；

3) with sound level meter measure front side axis or so 45° angle with minimum safe distance S as radius drawn by any point on camber line The A weighted sound pressure levels at place, and keep a record；

4) A weighted sound pressure level of any one same level traditional vehicle at S distances is measured in same position, and is recorded at these points A weighted sound pressure levels；

5) the relatively step 3), step 4) in the A sound level differences that measure twice, answer return to step 4 to adjust if difference is excessive Reference acoustic pressure intensity, when difference of them is little, then it represents that selected sound can reach the level of the outer sound of equal traditional vehicle car.

8. according to the electrocar pedestrian's caution sound method for designing described in claim 1, it is characterised in that the step 7 is specific Step is as follows：

Determine neural network topology structure, it is that 11, i.e. its structure are 6-11-1 to select hidden layer neuron number through repeatedly training, really Determine the related parameter values in genetic algorithm, using genetic algorithm come Optimized BP Neural Network, set up GA-BP neural network models.

9. according to the electrocar pedestrian's caution sound method for designing described in claim 1, it is characterised in that the step 8 is set up Electrocar pedestrian's caution sound control system include：Micro-processor MCV, sound playing circuit, storage circuit, ambient noise are surveyed Amount circuit, CAN interface circuit and power circuit；Power circuit is powered for whole control system, and storage circuit, sound are played Circuit, hum measurement circuit, CAN interface circuit are connected with Micro-processor MCV respectively；

Micro-processor MCV is obtained vehicle speed and travel conditions and is obtained by hum measurement circuit and gone by CAN The ambient noise in environment is sailed, according to the gain factor of the control amplifying circuit of the size intelligence of ambient noise, then selection is deposited Sample sound in storing up electricity road is simultaneously played out by sound playing circuit；

Sound playing circuit includes the D/A converter, amplifying circuit, the loudspeaker that are sequentially connected, and it is responsible for playing Micro-processor MCV The sample sound of screening；

Storage circuit includes FLASH storage chips and RAM storage chips, and FALSH storage chips are responsible for storing the scoring for filtering out Highest voice data, RAM storage chips are responsible for the temporary of the voice data selected by Micro-processor MCV and play sound number According to storage；

CAN interface realizes level conversion with vehicle CAN bus interface, and speed or motor speed are read by Micro-processor MCV And hybrid vehicle type of drive information judges whether to need to open caution control system, upon power-up of the system, by root The sound to be played is selected according to running state of the vehicle；

Hum measurement circuit is measured using noise testing sensor to the ambient noise residing for vehicle, and measurement is tied Fruit is supplied to Micro-processor MCV, to control the gain range of amplifying circuit in sound playing circuit；

Power circuit uses DC/DC dc source chips, for each circuit of system provides different voltages.

10. according to the electrocar pedestrian's caution sound method for designing described in claim 9, it is characterised in that the electrocar The control method of pedestrian's caution sound control system is：

1) Micro-processor MCV utilizes the status information that vehicle is read with vehicle CAN bus interface circuit, using hum measurement The size of the ambient noise measured by sensor, with reference to all kinds of sample sounds being previously stored in memory, is changed by D/A Device exports simulated audio signal, and amplified circuit is exported by loudspeaker, so as to select sample sound and gain size, passes through to be broadcast The difference of the difference and acoustic gain of putting sample sound carrys out real-time table up to the operation conditions of vehicle；

3) linear relationship set up between speed and voice signal time domain：GES is read by CAN, when speed is less than During 10km/h, Micro-processor MCV is chosen 10km/h scoring highest sample sounds and is played out, and speed is in 10km/h-20km/h Between when, Micro-processor MCV choose storage 20km/h scoring highest sample sound play out, speed is in 20km/h- When between 30km/h, the sample sound that Micro-processor MCV chooses the 30km/h storages of storage is played out；During reversing, microprocessor Device MCU plays the sample sound during 30km/h of storage；

4) using the ambient noise measured by hum measurement sensor, the selection of the different intelligent according to background noise level The grade of gain in amplifying circuit：Ambient noise is divided into Three Estate, respectively less than 55dB, 55dB to 65dB and is more than 65dB, according to different background noise levels, selects different gain factors.