CN103854642B - Flame speech synthesizing method based on physics - Google Patents

Abstract

The present invention relates to the fields such as computer game, video display specially good effect, Engineering Simulation, for providing the automatic synthesis method of a kind of flame sound, provide key technology to support for acousto-optic emulation.For this; the present invention adopts the technical scheme that, flame speech synthesizing method based on physics, comprises the following steps: the visual simulation of flame: use blue kernel model to model flame; flame is simulated, and derives velocity field and other dependent field amount such as fuel firing rates of every frame；Speed divergence integral and calculating: calculated by the speed divergence integration of flame front encirclement part by fuel firing rate；Up-sampling based on fast fourier transform: use up-sampling based on FFT that the speed divergence integration tried to achieve is reconstructed；Acoustic pressure exports: the speed divergence integration derivation after up-sampling is obtained final acoustic pressure.Present invention is mainly applied to video, image procossing occasion.

Description

Flame speech synthesizing method based on physics

Technical field

The present invention relates to the fields such as computer game, video display specially good effect, Engineering Simulation, specifically, relate to fire based on physics Flame speech synthesizing method.

Technical background

Traditional flame simulating often only focuses on visual drafting [1], and have ignored rendering acoustically.Paint with vision As system, it is also the important component part of virtual reality sense of reality that audition renders.Even if flame simulating can produce life-like Scene true to nature, if but lacked rendering acoustically, it is untrue that user still can experience scene, thus causes bad Consumer's Experience.

Recording is one the most direct audition rendering intent.User needs to record one section of correspondence for each concrete application Sound.But owing to the recording of such form is unfavorable for that multiplexing, different application will be recorded again, people more start to make Use short recording.Mating concrete application by editing short recording, operate, synthesizing, this kind of method is collectively referred to as granularity Synthesis (Granular Synthesis).But recording have in two defect: first, record and the scene simulated need hands Dynamic synchronization, this is not only the work taken time and effort, and also is difficult to accomplish to synchronize completely；Secondly, can be mutual at some Scene in, most of interbehaviors also can not look-ahead, this also allow for recording method have in actual applications Bigger restriction.

In the last few years, along with the further development of computer graphics, the concern rendered for audition gradually increased, accordingly Achievement in research gradually highlight, researcher in succession propose vibrations solid [2] [3] [4], air force acoustics phenomenon [5] [6], The sound generation method of the phenomenons such as splash liquid [7] [8] and pressure break solid [9].But the generation for flame sound is the most relative Blank.Dobashi et al. [5] [6] proposes a kind of method generating whirlpool sound based on physics, and is applied to flame.Should Although method can solve the problems referred to above, but the main sound source of flame is not from vortex, say, that whirlpool sound is flame sound Middle small part.Jeffrey and Doug [10] proposes the side generating the produced sound that burns from flame main sound source Method.Although the method that they propose can generate believable flame sound, but its huge calculating expends and makes it unsuitable for again Miscellaneous scene.Even moving back and forth of simple flame scene, such as wax candle, also to expend tens of hours, even days of Operation time.Reduce the frame per second of simulation, the time of simulation can be substantially reduced, but the most also appreciable impact can generate the true of sound True feeling.

Flame is the important component part of scene of game, film specially good effect, simulated drill system (such as fire-fighting drill) etc..Flame Sense of reality simulation include that visual simulation and audition emulate two parts, but existing work biases toward the former research more；And listen Feeling that emulation can greatly strengthen the feeling of immersion of simulated scenario, it is not only a requisite part in game engine, is also many Key technology indispensable in film specially good effect.

[1]Bukowski R,Sequin C.Interactive simulation of fire in virtual Building environments.Proceedings of SIGGRAPH, 1997.35～44.

[2]O’Brien J F,Cook P R,Essl G.Synthesizing sounds from physically Based motion.Proceedings of SIGGRAPH, 2001.529～536.

[3]O’Brien J F,Shen C,Gatchalian C M.Synthesizing sounds from rigid- Body simulations.Proceedings of SIGGRAPH, 2002.175～181.

[4]Van D D,Pai D K.Foleyautomatic:Physically based sound effects for Interactive simulation and animation.Proceedings of SIGGRAPH, 2001.537～544.

[5]Dobashi Y,Yamamoto T,Nishita T.Real-time rendering of aerodynamic sound using sound textures based on computational fluid dynamics.ACM Transactions on Graphics, 2003,22 (3): 732～740.

[6]Dobashi Y,Yamamoto T,Nishita T.Synthesizing sound from turbulent fields using sound textures for interactive fluid simulation.Computer Graphics Forum, 2004,23 (3): 736～744.

[7]Zheng C.,James D.L.Harmonic fluids.ACM Transactions on Graphics, 2009,28 (3): 37:1～37:12.

[8]Moss W,Yeh H,Manocha D.Sounding liquids:Automatic sound synthesis From fluid simulation.ACM Transactions on Graphics, 2010,29 (3): 21:1～21:13.

[9]Zheng C,James D L.Rigid-body fracture sound with precomputed Soundbanks.ACM Transactions on Graphics, 2010,29 (3): 69:1～69:13.

[10]Chadwick J N,James D L.Animating fire with sound.ACM Transactions On Graphics, 2011,30 (4): 84:1～84:8..

Summary of the invention

For overcoming the deficiencies in the prior art, it is provided that the automatic synthesis method of a kind of flame sound, provide for acousto-optic emulation and close Key technology supports.To this end, the present invention adopts the technical scheme that, flame speech synthesizing method based on physics, including following step Rapid:

The visual simulation of flame: use blue kernel model to model flame, flame is simulated, and derives every frame Velocity field and other dependent field amount such as fuel firing rates；

Speed divergence integral and calculating: calculated by the speed divergence integration of flame front encirclement part by fuel firing rate；

Up-sampling based on fast fourier transform (Fast Fourier Transform, hereinafter referred to as FFT): use base The speed divergence integration tried to achieve is reconstructed by the up-sampling in FFT；

Acoustic pressure exports: the speed divergence integration derivation after up-sampling is obtained final acoustic pressure.

To flame be simulated particularly as follows:

The present invention uses blue kernel model to model flame, and this model hypothesis crosses gaseous fuel and combustion when gaseous fuel Can burn the most completely when burning the separating surface producing thing, release amount of heat and corresponding burning produce thing, and separating surface is referred to as Flame front, uses level set or similar method explicitly to follow the tracks of flame front, and gaseous fuel and burning are produced thing respectively Following Navier-Stokes equation group is used to be modeled:

$\frac{\∂u}{\∂t}=-(u\·\▿)u-\▿p/\mathrm{\ρ}+f,---\left(1\right)$

$\▿\·u=\mathrm{\Φ},---\left(2\right)$

Wherein, formula (1) is the equation of momentum, and u is speed, and t represents the time, and p is pressure, and ρ is density, and f represents external force, bag Include buoyancy or gravity；Formula (2) is mass equation, and Φ is an optional divergence source, obtains by specifying in advance or simulating Real time data calculate the value of Φ, when Φ is more than zero, velocity field at this point to external diffusion；Otherwise, when Φ is less than zero, speed Degree field is shunk inwards at this point；In addition to velocity field, density field is used for modeling burning and produces the state of smog, and temperature field is used for The heat that modeling burning produces；

Speed divergence integral and calculating particularly as follows:

Based on blue core it is assumed that heat release concentrates on flame front, and heat burst size and speed fuel are straight It is connected into direct ratio, the curve surface integral of the volume integral conversion rate flux that then heat can be discharged, thus obtain:

$\∫{\mathrm{qd}}^{3}x={\∫}_{S}u\·\mathrm{nds},---\left(3\right)$

Wherein, S is flame front, and u represents that speed, n are normal vectors, and q represents heat burst size；In order to avoid this operation, According to Gauss divergence formula, curve surface integral is changed into the volume integral of divergence, it may be assumed that

${\∫}_{S}u\·\mathrm{nds}={\∫}_V\▿\·\mathrm{udv},---\left(4\right)$

Wherein, V is any solid surrounded by flame front, based on describing sound source and the propagation wave equation of sound, And ignore time delay and range attenuation, obtaining acoustic pressure p (t) is:

$p\left(t\right)=\frac{d}{\mathrm{dt}}{\∫}_V\▿\·u(x,t)\mathrm{dv}\left(x\right).---\left(5\right)$

Up-sampling based on FFT particularly as follows:

Build up-sampling based on a FFT operation under at frequency domain by the way of zero padding to realize signal and strengthen, zero padding Operate with zero padding to extend frequency spectrum length, by zero padding, the frequency spectrum of an a length of L is extended to N L, as follows:

${\mathrm{ZeroPad}}_{N,L}\left(x\right)=\left\{\begin{array}{cc}x\left(\mathrm{\&omega;}\right),& \left|\mathrm{\&omega;}\right|<L/2\\ 0,& \mathrm{otherwise}\end{array}\right.,---\left(6\right)$

Wherein, L is former frequency spectrum length, and N is spectrum signal length after zero padding operation, and usually, N=kL, k are for expanding multiple； Divergence value div (t) tried to achieve specifically, is first converted to frequency domain s (ω) by FFT by up-sampling behaviour from spatial domain；Keep low frequency Part is constant, and HFS is carried out zero padding operation；Finally, by reverse Fourier transform (Inverse Fast Fourier Transform, be called for short IFFT) up-sampled after time-domain signal.

Acoustic pressure output particularly as follows:

In order to try to achieve acoustic pressure p (t), need divergence value div (t) is carried out derivation operations: use cube interpolating function to div T () carries out interpolation:

$\mathrm{div}\left(t\right)=\underset{n=0}{\overset{N}{\mathrm{\&Sigma;}}}\mathrm{div}\left(\mathrm{n\&Delta;t}\right)k[(t-\mathrm{n\&Delta;t})/\mathrm{\&Delta;t}],---\left(7\right)$

Wherein, $k\left(x\right)=\frac{1}{6}\left\{\begin{array}{cc}7{\left|x\right|}^3-12{\left|x\right|}^2+\frac{16}{3},& \left|x\right|<1\\ -\frac{7}{3}{\left|x\right|}^3+12{\left|x\right|}^2-20\left|x\right|+\frac{32}{3},& 1\&le;\left|x\right|<2\\ 0,& \mathrm{otherwise}\end{array}\right..$

Afterwards, its derivation is tried to achieve acoustic pressure:

$p\left(t\right)=\frac{d}{\mathrm{dt}}\mathrm{div}\left(t\right)=\underset{n=0}{\overset{N}{\mathrm{\&Sigma;}}}\mathrm{div}\left(\mathrm{n\&Delta;t}\right)\frac{d}{\mathrm{dt}}k[(t-\mathrm{n\&Delta;t})/\mathrm{\&Delta;t}],---\left(8\right).$

The present invention possesses following technique effect:

The present invention can automatically generate flame sound effect the most true to nature, can be used for computer game, video display specially good effect system The fields such as work.Additionally, in Engineering Simulation, the sense of reality of simulated environment and the property immersed can be strengthened.

The running environment of the present invention requires simple, and implementation cost is low, and specific requirement is as follows:

Support Windows XP, Windows vista, Windows7 environment, the configuration of more than 2G internal memory, more than 1G video memory Microcomputer.

Accompanying drawing explanation

Fig. 1 up-sampling based on FFT schematic diagram.

Fig. 2 illustrates the flame scene and the acoustical signal of synthesis utilizing the present invention to generate.

Fig. 3 flow chart of the present invention.

Detailed description of the invention

The present invention proposes a kind of flame sound automatic synthesis method based on physics, mainly comprises the steps that

The visual simulation of flame: use solution based on physics that flame is simulated, and derive the speed of every frame Field and other dependent field amount such as fuel firing rates.

Speed divergence integral and calculating: calculated by the speed divergence integration of flame front encirclement part by fuel firing rate.

Up-sampling based on fast fourier transform (Fast Fourier Transform, hereinafter referred to as FFT): use base The speed divergence integration tried to achieve is reconstructed by the up-sampling in FFT.

Acoustic pressure exports: the speed divergence integration derivation after up-sampling is obtained final acoustic pressure.

The concrete technical scheme of the present invention is as follows:

1) visual simulation of flame

The present invention uses blue kernel model (list of references: Nguyen D, Fedkiw R, Jensen H.Physically based modeling and animation of fire.ACM Transactions on Graphics,2002,21(3): 721～728.) flame is modeled.The meeting when gaseous fuel crosses gaseous fuel with the separating surface of the generation thing that burns of this model hypothesis Burning the most completely, release amount of heat and corresponding burning produce thing.This separating surface is referred to as flame front.To flame front Use level set or similar method explicitly to follow the tracks of, gaseous fuel and burning are produced thing and uses following Navier-respectively Stokes equation group is modeled:

$\frac{\&PartialD;u}{\&PartialD;t}=-(u\&CenterDot;\&dtri;)u-\&dtri;p/\mathrm{\&rho;}+f,---\left(1\right)$

$\&dtri;\&CenterDot;u=\mathrm{\&Phi;},---\left(2\right)$

Wherein, formula (1) is the equation of momentum, and u is speed, and t represents the time, and p is pressure, and ρ is density, and f represents external force, than Such as buoyancy or gravity.Formula (2) is mass equation, and Φ is an optional divergence source, can be by specifying in advance or simulation The real time data obtained is to calculate the value of Φ.When Φ more than zero time, velocity field at this point to external diffusion；Otherwise, when Φ is less than zero Time, velocity field is shunk inwards at this point.In addition to velocity field, density field can be used to model burning and produces the state of smog, temperature Degree field can be used to model the heat that burning produces.

2) speed divergence integral and calculating

It is based on blue core it is assumed that heat release concentrates on flame front.And heat burst size and speed fuel are straight It is connected into direct ratio, the curve surface integral of the volume integral conversion rate flux that then heat can be discharged, thus obtain:

$\&Integral;{\mathrm{qd}}^{3}x={\&Integral;}_{S}u\&CenterDot;\mathrm{nds},---\left(3\right)$

Wherein, S is flame front, and u represents that speed, n are normal vectors, and q represents heat burst size.Although above-mentioned conversion without Ask for heat burst size q again, but owing to needs calculate curve surface integral in flame front, the most inevitably need before flame End is separated into tri patch.Experiment shows, this is a suitable operation expending calculating resource, and operation will expend second level every time Time.In order to avoid this operation, curve surface integral is changed into the volume integral of divergence by the present invention according to Gauss divergence formula, That is:

${\&Integral;}_{S}u\&CenterDot;\mathrm{nds}={\&Integral;}_V\&dtri;\&CenterDot;\mathrm{udv},---\left(4\right)$

Wherein, V is any solid surrounded by flame front.Although this integration is still the geometry surrounded in flame front Internal calculating, but the most calculative be the divergence of speed, such that it is able to be obviously improved the calculating time.Based on describe sound source and Propagation wave equation (the list of references: Chrighton D G, Dowling A P.Modern Methods in of sound Analytical Acoustics.Berlin:Springer-Verlag, 1992.), and ignore time delay and range attenuation, Obtaining acoustic pressure p (t) is:

$p\left(t\right)=\frac{d}{\mathrm{dt}}{\&Integral;}_V\&dtri;\&CenterDot;u(x,t)\mathrm{dv}\left(x\right).---\left(5\right)$

3) up-sampling based on FFT

The data derived from flame simulating obtain speed divergence integration div (t) of every frame, but the information between frame and frame All lost.In order to improve speed further, need to be reached by the approach reducing simulation frame per second.But reduce frame per second can lead Cause the loss of some key frame informations thus cause generating the distortion of sound.

The present invention constructs up-sampling based on a FFT operation under at frequency domain and realizes signal and increase by the way of zero padding By force.Zero padding operates with zero padding to extend frequency spectrum length.By zero padding, the frequency spectrum of an a length of L can be extended to N > L, as follows:

${\mathrm{ZeroPad}}_{N,L}\left(x\right)=\left\{\begin{array}{cc}x\left(\mathrm{\&omega;}\right),& \left|\mathrm{\&omega;}\right|<L/2\\ 0,& \mathrm{otherwise}\end{array}\right.,---\left(6\right)$

Wherein, L is former frequency spectrum length, and N is spectrum signal length after zero padding operation.Usually, N=kL, k are for expanding multiple. Owing to such zero padding operation can be the up-sampling information that time-domain signal provides almost Perfect, the up-sampling thus built operates The frame information lost by reducing frame per second can be recovered as far as possible.Up-sampling operation is as it is shown in figure 1, divergence that first will try to achieve Value div (t) is converted to frequency domain s (ω) by FFT from spatial domain.In Fig. 1 (c), keep low frequency part constant, HFS is entered Row zero padding operates.Finally, obtained by reverse Fourier transform (Inverse Fast Fourier Transform is called for short IFFT) Time-domain signal after up-sampling.Comparison diagram 1(a) and Fig. 1 (d), after up-sampling operation, lose between each time step Frame information the most recovered.

Acoustic pressure exports

In order to try to achieve acoustic pressure p (t), need divergence value div (t) is carried out derivation operations.Owing to div (t) is only fixing Time step 0, △ t, 2 △ t ..., N △ t understands, and need to obtain continuous print time-domain signal by interpolation operation.The present invention uses Cube interpolating function (the list of references: Mitchell D P, Netravali that Mitchell and Netravali proposes A.Reconstruction filters in computer-graphics.Proceedings of SIGGRAPH, 1988.221～228.) div (t) is carried out interpolation:

$\mathrm{div}\left(t\right)=\underset{n=0}{\overset{N}{\mathrm{\&Sigma;}}}\mathrm{div}\left(\mathrm{n\&Delta;t}\right)k[(t-\mathrm{n\&Delta;t})/\mathrm{\&Delta;t}],---\left(7\right)$

Wherein, $k\left(x\right)=\frac{1}{6}\left\{\begin{array}{cc}7{\left|x\right|}^3-12{\left|x\right|}^2+\frac{16}{3},& \left|x\right|<1\\ -\frac{7}{3}{\left|x\right|}^3+12{\left|x\right|}^2-20\left|x\right|+\frac{32}{3},& 1\&le;\left|x\right|<2\\ 0,& \mathrm{otherwise}\end{array}\right..$

Afterwards, its derivation is tried to achieve acoustic pressure:

$p\left(t\right)=\frac{d}{\mathrm{dt}}\mathrm{div}\left(t\right)=\underset{n=0}{\overset{N}{\mathrm{\&Sigma;}}}\mathrm{div}\left(\mathrm{n\&Delta;t}\right)\frac{d}{\mathrm{dt}}k[(t-\mathrm{n\&Delta;t})/\mathrm{\&Delta;t}],---\left(8\right)$

The present invention relates to solving and vector calculus of a large amount of equation, use based on GPU (Graphics Processing Units) programming mode can improve computational efficiency.

Claims (5)

1. a flame speech synthesizing method based on physics, is characterized in that, comprise the following steps:

The visual simulation of flame: use blue kernel model to model flame, flame is simulated, and derives the speed of every frame Field and the dependent field amount including fuel firing rate；

Speed divergence integral and calculating: calculated by the speed divergence integration of flame front encirclement part by fuel firing rate；

Up-sampling based on fast fourier transform (Fast Fourier Transform, hereinafter referred to as FFT): use based on The speed divergence integration tried to achieve is reconstructed by the up-sampling of FFT；

Acoustic pressure exports: the speed divergence integration derivation after up-sampling is obtained final acoustic pressure.

2. flame speech synthesizing method based on physics as claimed in claim 1, is characterized in that, is simulated flame concrete For: using blue kernel model to model flame, this model hypothesis is crossed gaseous fuel when gaseous fuel and is produced thing with burning Can burn the most completely during separating surface, release amount of heat and corresponding burning produce thing, and separating surface is referred to as flame front, right Flame front uses level set explicitly to follow the tracks of, and gaseous fuel and burning is produced thing and uses following Navier-Stokes side respectively Journey group is modeled:

Wherein, formula (1) is the equation of momentum, and u is speed, and t represents the time, and p is pressure, and ρ is density, and f represents external force, including floating Power or gravity；Formula (2) is mass equation, and Φ is an optional divergence source, by the reality specified in advance or simulation obtains Time data calculate the value of Φ, when Φ is more than zero, velocity field at this point to external diffusion；Otherwise, when Φ is less than zero, velocity field Shrink inwards at this point；In addition to velocity field, density field is used for modeling burning and produces the state of smog, and temperature field is used for modeling The heat that burning produces.

3. flame speech synthesizing method based on physics as claimed in claim 1, is characterized in that, speed divergence integrating meter calculator Body is:

Based on blue core it is assumed that heat release concentrates on flame front, and heat burst size directly becomes with speed fuel Direct ratio, the curve surface integral of the volume integral conversion rate flux that then heat can be discharged, thus obtain:

∫qd³X=∫_SU nds, (3)

Wherein, S is flame front, and u represents that speed, n are normal vectors, and q represents heat burst size；According to Gauss divergence formula by song Line Integral changes into the volume integral of divergence, it may be assumed that

Wherein, V is any solid surrounded by flame front, based on describing sound source and the propagation wave equation of sound, and neglects Slightly time delay and range attenuation, obtaining acoustic pressure p (t) is:

。

4. flame speech synthesizing method based on physics as claimed in claim 1, is characterized in that, up-sampling based on FFT has Body is: building up-sampling based on a FFT operation under at frequency domain by the way of zero padding and realize signal and strengthen, zero padding operates Use zero padding to extend frequency spectrum length, by zero padding, the frequency spectrum of an a length of L is extended to N L, as follows:

Wherein, L is former frequency spectrum length, and N is spectrum signal length after zero padding operation, and N=kL, k are for expanding multiple；Up-sampling behaviour's tool Body is, first by FFT, divergence value div (t) tried to achieve is converted to from spatial domain frequency domain s (ω)；Keep low frequency part constant, right HFS carries out zero padding operation；Finally, by reverse Fourier transform (Inverse Fast Fourier Transform, Be called for short IFFT) up-sampled after time-domain signal.

5. flame speech synthesizing method based on physics as claimed in claim 1, is characterized in that, acoustic pressure output particularly as follows:

In order to try to achieve acoustic pressure p (t), need divergence value div (t) is carried out derivation operations: use cube interpolating function to div (t) Carry out interpolation:

Wherein,

Afterwards, its derivation is tried to achieve acoustic pressure:

。