CN111798831B - Sound particle synthesis method and device - Google Patents

Abstract

The invention belongs to the technical field of digital signal processing and discloses a sound particle synthesis method and device. The method includes collecting original sound samples, constructing a sound particle sample library based on the original sound samples, adding an index of sound particles, inputting information according to vehicle conditions, and based on The index of sound particles reads the sound particles in the sound particle sample library, constructs transition audio based on the sound particles read by the index, splices adjacent sound particles through the transition audio, and synthesizes the target sound sample; the device includes a collection unit, sound particle samples Library building unit, indexing unit, sound particle reading unit, transition audio building unit, target sound sample synthesis unit. The invention solves the problem in the prior art that sound particles will produce a "step" phenomenon at the splicing point, resulting in "pops" being mixed into the synthesized audio sound, making the synthesized sound not continuous and natural enough. It can ensure the coherence of the synthesized sound and satisfy the requirements. Requirements for target sound quality.

Description

A sound particle synthesis method and device

Technical field

The present invention relates to the technical field of digital signal processing, and in particular to a sound particle synthesis method and device.

Background technique

The automobile industry has entered a steady period of connotative development. When the NVH (Noise, Vibration, Harshness) problems of various automobile systems are improved, people pay more attention to the impact of noise on the subjective feelings of passengers. , highlighted as the transformation of “from silence to sound quality”. ASD (active sound design) is a method that uses active sound to satisfy the "sound fun" experience of drivers and passengers and improve the target requirements of in-car sound quality. At present, ASD technology is mainly researched from three parts: sound design, sound synthesis algorithm and active sound control system.

Among them, the sound synthesis algorithm is the key to ensuring the coherence of the target sound playback. The current sound synthesis algorithm for the car mainly includes the harmonic algorithm and the particle algorithm. The harmonic algorithm is based on setting the input signal (such as rotation speed) and synthesizing the order sound. The functional relationship between amplitude and frequency is used to synthesize order sound; the particle algorithm cuts the recorded sound samples into small sound particles, reads the corresponding samples according to the working condition input, and splices the sound particles to play The sound changes with the working conditions.

Compared with the harmonic algorithm, the sound particle algorithm can highly restore the sound quality of the original sound. In order to meet the diversified in-car sound needs of customers, the sound particle algorithm has become one of the most effective sound synthesis solutions. The coherence of sound particles during the splicing process is the key to ensuring the quality of synthetic sound. However, usually sound particles will produce a "step" phenomenon at the splicing point, resulting in "pops" in the synthesized audio sound, making the synthesized sound insufficient. Continuous and natural.

Contents of the invention

By providing a method and device for synthesizing sound particles, the embodiments of the present application solve the problem of the "step" phenomenon in the existing technology where sound particles are spliced together, resulting in "pops" being mixed into the synthesized audio sound, making the synthesized sound not continuous enough. Natural question.

The embodiment of the present application provides a sound particle synthesis method, which includes the following steps:

Step 1. Collect original sound samples;

Step 2. Construct a sound particle sample library based on the original sound sample, and add the index of the sound particle;

Step 3: According to the vehicle condition input information, read the sound particles in the sound particle sample library based on the index of the sound particles;

Step 4: Construct transitional audio based on the sound particles read by the index; splice adjacent sound particles through the transitional audio to synthesize the target sound sample.

Preferably, step 1 is implemented by: selecting a target car model and recording the audio signal corresponding to the target car model as the original sound sample;

Wherein, the audio signal includes but is not limited to engine sound information.

Preferably, step 2 is implemented by: locating and intercepting mark points and extracting sound particles according to the original sound sample, and constructing the sound particle sample library;

Wherein, the first and last phases of the sound particles in the sound particle sample library are both 0.

Preferably, in step 2, the method of adding an index of sound particles is to: extract the sound particle fragments in the sound particle sample library and place them in order; number each sound particle, and Make an index array;

In step 3, the index number of the corresponding sound particle is obtained according to the vehicle condition input information, and the sound particles in the sound particle sample library are read according to the index number; the vehicle condition input information includes but is not limited to engine speed information, vehicle speed information, torque information.

Preferably, in step 4, based on the sound particles read by the index, a transitional audio is constructed to splice adjacent sound particles; wherein, the audio signal is compared with the functional expression of the Hilbert transform to obtain the audio signal function expression;

The phase information and frequency information of the transition audio are obtained through the Hermitian interpolation algorithm, and the amplitude information is obtained through the interpolation algorithm;

The transition audio is constructed according to the audio signal function expression and the phase information, frequency information, and amplitude information of the transition audio.

Preferably, the audio signal function expression y(t) is as follows:

y(t)=A(t)sin(Φ(t))

In the formula, A(t) is the instantaneous amplitude of the audio signal, Φ(t) is the instantaneous phase of the audio signal, is the derivative of the instantaneous phase of the audio signal, F(t) is the instantaneous frequency of the audio signal, C is the integration constant, and t2-t1 is the duration of the audio signal.

Preferably, the implementation method of obtaining the phase information of transition audio through the Hermitian interpolation algorithm is:

Assume that the adjacent sound particles are sound particle i and sound particle i+1 respectively, then the transition audio used to splice the sound particle i and the sound particle i+1 is recorded as transition audio i;

The instantaneous phase curve of the synthesized sound obtained after splicing is continuously differentiable;

Wherein, the time interval corresponding to the sound particle i is t _i ~ t _i+1 , and the instantaneous phase of the sound particle i at time t _i+1 is Φ(t _i+1 ); the sound particle i+1 The corresponding time interval is t _i+2 ~ t _i+3 , and the instantaneous phase of the sound particle i+1 at time t _i+2 is Φ(t _i+2 );

The time interval corresponding to the transition audio i is ti ₊₁ to ti ₊₂ ; the left endpoint of the transition audio i is the audio signal of the sound particle i at time t _i+1 , and the transition audio i The right endpoint is the audio signal of the sound particle i+1 at time t _i+2 ;

The integral constant of the instantaneous phase of the sound particle i at time t _i+1 is C1; the integral constant of the instantaneous phase of the sound particle i+1 at time t _i+2 is C2;

Preset _ti is 0, C1=0, according to the instantaneous phase Φ(t _i+1 ) of the left endpoint of the transition audio i and the first-order derivative of the instantaneous phase of the left and right endpoints and/> The second derivative of the instantaneous phase of the left and right endpoints/> and/> The expression Φ(t) for solving the phase of transition audio i using the quartic Hermite interpolation method is obtained as follows:

Φ(t)＝at ⁴ +bt ³ +ct ² +dt+e

In the formula, a, b, c, d, and e are the polynomial coefficients to be found;

Using the first interpolation condition, the first linear equation system is obtained as:

Among them, rank(A|y)=rank(A)=5, then the polynomial coefficient of Hermitian interpolation has a unique solution, and the expression of the transition audio phase Φ(t) can be obtained, and C2=2π*Φ can be obtained. (t _i+2 );

When sin(C2) obtained from C2 obtained according to the first interpolation condition is not 0, the second interpolation condition is used to replace the first interpolation condition, and the second linear equation set is used to replace the first linear equation set; The second system of linear equations is expressed as:

Among them, rank(A|y)=rank(A)<5, then the interpolation coefficient vector X has countless solutions, from which an expression Φ of the transition audio phase satisfying sin(C1)=0 and sin(C2)=0 is obtained. (t).

Preferably, the frequency information of the transition audio satisfies the following constraints:

The frequency of the transition audio is between the frequencies of the left and right endpoints;

The frequency of the transition audio is monotonic.

Preferably, the amplitude of the transition audio satisfies the continuity condition; according to the amplitudes A(t _i+1 ) and A(t _i+2 ) of the sound particle i and the sound particle i+1 at the splicing point , interpolate to obtain the amplitude information of the transition audio.

An embodiment of the present application provides a sound particle synthesis device, including:

A collection unit is used to collect and obtain original sound samples;

A sound particle sample library construction unit is used to obtain a sound particle sample library based on the original sound sample;

An indexing unit, used to obtain the index of sound particles based on the sound particle sample library;

A sound particle reading unit is used to input information according to the vehicle condition and read the sound particles in the sound particle sample library based on the index of the sound particles;

A transition audio building unit, used to obtain transition audio based on the sound particles read by the index;

A target sound sample synthesis unit is used to splice adjacent sound particles through the transition audio to synthesize the target sound sample;

The sound particle synthesis device is used to implement the above sound particle synthesis method.

One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

In the embodiment of this application, original sound samples are first collected, then a sound particle sample library is constructed based on the original sound samples, and the index of the sound particle is added. Then, information is input according to the vehicle condition, and the sound particle sample library is read based on the index of the sound particle. Sound particles, and then construct transition audio based on the sound particles read by the index, and splice adjacent sound particles through the transition audio to synthesize the target sound sample. The present invention realizes the natural splicing of sound particles by constructing transitional audio to splice adjacent sound particles, solves the "step" phenomenon that occurs in the splicing process of existing sound particles, avoids the occurrence of "pops" in synthesized sounds, and ensures that the synthesized sounds are coherence to meet the requirements for target sound quality.

Description of the drawings

In order to explain the technical solution in this embodiment more clearly, the drawings needed to be used in the description of the embodiment will be briefly introduced below. Obviously, the drawing in the following description is an embodiment of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is a flow chart of a sound particle synthesis method provided by an embodiment of the present invention;

Figure 2 is a schematic diagram of a sound particle synthesis method provided by an embodiment of the present invention;

Figure 3 is a schematic diagram of splicing adjacent sound particles through "transition audio" in a sound particle synthesis method provided by an embodiment of the present invention;

Figure 4 is a spectrogram of a synthetic sound that directly splices adjacent sound particles;

Figure 5 is a spectrogram of a synthetic sound that uses "transition audio" to splice adjacent sound particles into a bridge.

Among them, 1-sound particle 1, 2-transition audio 1, 3-sound particle 2, 4-transition audio 2, 5-sound particle 3.

Detailed ways

This embodiment provides a sound particle synthesis method, which includes the following steps:

Step 1. Collect original sound samples.

Specifically, a target car model is selected, and the audio signal corresponding to the target car model is recorded as the original sound sample. Wherein, the audio signal includes but is not limited to engine sound information.

Step 2: Construct a sound particle sample library based on the original sound sample, and add the index of the sound particle.

Specifically, according to the original sound sample, the interception mark point is located and the sound particles are extracted, and the sound particle sample library is constructed. Wherein, the first and last phases of the sound particles in the sound particle sample library are both 0.

One implementation method of adding the index of sound particles is to: extract the sound particle fragments in the sound particle sample library and place them in order; number each sound particle and create an index array.

Step 3: According to the vehicle condition input information, read the sound particles in the sound particle sample library based on the index of the sound particles.

Specifically, the index number of the corresponding sound particle is obtained according to the vehicle condition input information, and the sound particles in the sound particle sample library are read according to the index number. The vehicle condition input information includes but is not limited to engine speed information, vehicle speed information, and torque information.

Step 4: Construct transitional audio based on the sound particles read by the index; splice adjacent sound particles through the transitional audio to synthesize the target sound sample.

Specifically, based on the sound particles read by the index, a transition audio is constructed to splice adjacent sound particles. Among them, the audio signal is compared with the function expression of the Hilbert transform to obtain the audio signal function expression. The phase information, frequency information, and amplitude information of the transition audio are obtained through the Hermitian interpolation algorithm. The transition audio is constructed according to the audio signal function expression and the phase information, frequency information, and amplitude information of the transition audio.

Wherein, the audio signal function expression y(t) is as follows:

y(t)=A(t)sin(Φ(t))

In the formula, A(t) is the instantaneous amplitude of the audio signal, Φ(t) is the instantaneous phase of the audio signal, is the derivative of the instantaneous phase of the audio signal, F(t) is the instantaneous frequency of the audio signal, C is the integration constant, and t2-t1 is the duration of the audio signal.

The implementation method of obtaining the phase information of the transition audio through the Hermitian interpolation algorithm is as follows: assuming that the adjacent sound particles are sound particle i and sound particle i+1 respectively, then it is used to splice the sound particle i and the The transition audio of sound particle i+1 is recorded as transition audio i; the instantaneous phase curve of the synthesized sound obtained after splicing is continuously differentiable.

Wherein, the time interval corresponding to the sound particle i is t _i ~ t _i+1 , and the instantaneous phase of the sound particle i at time t _i+1 is Φ(t _i+1 ); the sound particle i+1 The corresponding time interval is t _i+2 ~ t _i+3 , and the instantaneous phase of the sound particle i+1 at time t _i+2 is Φ(t _i+2 ). The time interval corresponding to the transition audio i is ti ₊₁ to ti ₊₂ ; the left endpoint of the transition audio i is the audio signal of the sound particle i at time t _i+1 , and the transition audio i The right endpoint is the audio signal of the sound particle i+1 at time t _i+2 . The integral constant of the instantaneous phase of the sound particle i at time t _i+1 is C1; the integral constant of the instantaneous phase of the sound particle i+1 at time t _i+2 is C2.

Preset _ti is 0, C1=0, according to the instantaneous phase Φ(t _i+1 ) of the left endpoint of the transition audio i and the first-order derivative of the instantaneous phase of the left and right endpoints and/> The second derivative of the instantaneous phase of the left and right endpoints/> and/> The expression Φ(t) for solving the phase of transition audio i using the quartic Hermite interpolation method is obtained as follows:

Φ(t)＝at ⁴ +bt ³ +ct ² +dt+e

In the formula, a, b, c, d, and e are the polynomial coefficients to be found.

Using the first interpolation condition, the first linear equation system is obtained as:

Among them, rank(A|y)=rank(A)=5, then the polynomial coefficient of Hermitian interpolation has a unique solution, and the expression of the transition audio phase Φ(t) can be obtained, and C2=2π*Φ can be obtained. (t _i+2 ).

When sin(C2) obtained from C2 obtained according to the first interpolation condition is not 0, the second interpolation condition is used to replace the first interpolation condition, and the second linear equation set is used to replace the first linear equation set; The second system of linear equations is expressed as:

Among them, rank(A|y)=rank(A)<5, then the interpolation coefficient vector X has countless solutions, from which an expression Φ of the transition audio phase satisfying sin(C1)=0 and sin(C2)=0 is obtained. (t).

Since the instantaneous frequency of the audio signal is the derivative of the instantaneous phase of the audio signal, frequency information can be obtained based on the phase information. In addition, the frequency information of the transition audio satisfies the following constraints: the frequency of the transition audio is between the frequencies of the left and right end points; the frequency of the transition audio is monotonic.

The amplitude of the transition audio satisfies the continuity condition; according to the amplitudes A(t _i+1 ) and A(t _i+2 ) of the sound particle i and the sound particle i+1 at the splicing point, interpolation is obtained Amplitude information of the transition audio. Specifically, linear interpolation can be used here. According to the amplitude information of the two end points, the amplitude information of the transition audio is obtained by interpolation. The amplitude curve of the synthetic sound only needs to meet the continuity condition.

Corresponding to the above synthesis method, this embodiment also provides a sound particle synthesis device, including:

A collection unit is used to collect and obtain original sound samples;

A sound particle sample library construction unit is used to obtain a sound particle sample library based on the original sound sample;

An indexing unit, used to obtain the index of sound particles based on the sound particle sample library;

A sound particle reading unit is used to input information according to the vehicle condition and read the sound particles in the sound particle sample library based on the index of the sound particles;

A transition audio building unit, used to obtain transition audio based on the sound particles read by the index;

A target sound sample synthesis unit is used to splice adjacent sound particles through the transition audio to synthesize the target sound sample;

The sound particle synthesis device is used to implement the above sound particle synthesis method.

In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation modes.

This embodiment provides a sound particle synthesis method, as shown in Figures 1 and 2, which mainly includes the following steps:

Step 1. Collect original sound samples.

Collecting original sound samples includes: targeting customers' expectations for sound quality, selecting matching models on the market as target models, and recording their interior or engine sounds.

A specific recording method is to reasonably arrange a micro microphone in the engine compartment or car, then run the engine in neutral, slowly step on the accelerator, and let the engine speed slowly accelerate from idle to the highest speed the engine can reach, in order to obtain enough Long sound samples are used for sound sample library preparation.

Step 2. Build a sound particle sample library.

The preparation of the sound particle sample library includes: extraction of main harmonics, positioning and selection of sound sample interception mark points, and extraction of sound particles.

The main technical points include the design of the time-varying window function to meet the changes in the sound particle cutting size with the working conditions. Once the interception mark is determined from the interception mark point list, a time window centered on the interception mark will be used to extract the sound particles. It is possible to cut the recorded audio signal into sound particles whose first and last phases are both 0, and save them as a sound particle sample library.

Step 3. Index of sound particles.

The indexing of sound particles includes: extracting sound particle fragments and placing them in order, and numbering each sound particle to create an index array. And by obtaining input parameters such as engine speed, vehicle speed, and torque, the index number of the corresponding sound particle is calculated, and the sound particles in the sound sample library are read according to the index number.

Step 4. Construct transition audio and synthesize target sound samples.

It mainly includes constructing transition audio based on the sound particles read by the index; splicing adjacent sound particles through the transition audio to synthesize the target sound sample.

For example, sound particles 1, 2, and 3 are spliced with transition audio 1 and 2 to achieve sound synthesis, see Figure 3.

Specifically, step 4 mainly includes the following steps:

(1) Compare the audio signal to the functional expression of Hilbert transform.

Among them, Hilbert transform can analyze a narrow-band one-dimensional signal into a signal on a two-dimensional complex plane. The module and argument of the complex number represent the amplitude and phase of the signal. For a real signal x (t), its Hilbert transform is:

In the formula, * represents the convolution operation, and the corresponding analytical signal is:

Among them, for the narrow-band real signal x(t) with the carrier frequency f _s , the Hilbert transform can be used to solve the analytical signal, thereby obtaining the instantaneous amplitude a(t) and instantaneous phase of the signal. And the instantaneous phase derivative can be used to solve for the instantaneous frequency f(t). The specific expression is as follows:

Therefore, the analytical signal can be re-expressed as:

In the formula,

Based on the above analysis of the Hilbert transform, the present invention understands the audio signal expression and the analytical formula of the Hilbert transform by analogy. Suppose the audio signal function expression is y(t), therefore:

y(t)=A(t)sin(Φ(t))

In the formula, C is the integration constant, A(t) is the instantaneous amplitude (envelope) of the audio signal, Φ(t) is the instantaneous phase of the audio signal, F(t) is the instantaneous frequency of the audio signal, t2-t1 is The duration of the audio signal.

(2) Use the Hermitian interpolation algorithm to interpolate the instantaneous phase of the "transition audio".

The instantaneous frequency of the audio signal is the derivative of the instantaneous phase. Therefore, when time t=0, the corresponding phase is 0, then when adjacent sound particles are spliced, the integration constant C1 of the instantaneous phase of the first sound particle=0, and the frequency integration can be completed by the MATLAB function CUMSUM. However, when time t = t ₂ , the integral constant C2 of the instantaneous phase of the second segment of sound particles cannot be determined, and C2 should ensure that the instantaneous phase curves of adjacent sound particles are continuously differentiable after splicing.

Specifically, according to the solution method of the integral constant C2 of the instantaneous phase of the sound particles in the second section: in the form of "transition audio" connection, it is ensured that the instantaneous phase curves of the spliced sound particles are all continuously differentiable, and the "transition audio" connection is used. The instantaneous phase of "audio" can be interpolated by the Hermitian interpolation method.

Among them, according to the Hermitian interpolation method:

Assume that we have obtained the node a≤x ₀ <x ₁ ...<x _n ≤b, the function value h(x _j ) at the node and the derivative value h'(x _j ) at the point, then from this n +1 nodes interpolate a polynomial H 2n ₊₁ (x) of degree 2n+1 that approximates the original function through Hermite.

Specifically, the instantaneous phase solution method for interpolating the "transition audio" in the form of a "transition audio" connection and using the Hermitian interpolation method includes:

It is known that the value of the left endpoint of the interval to be interpolated, the first-order derivative of the two endpoints, and the second-order derivative of the two endpoints are known. Therefore, the fourth-order Hermitian interpolation method can be used to solve the instantaneous phase expression Φ(t). Suppose:

Φ(t)＝at ⁴ +bt ³ +ct ² +dt+e

In the formula, a, b, c, d, and e are the polynomial coefficients to be found, so a system of linear equations can be obtained:

Among them, rank(A|y)=rank(A)=5, the polynomial coefficient of Hermitian interpolation has a unique solution. At this time, the expression of the transition audio phase Φ(t) can be obtained, and C2=2π* can be obtained at the same time. Φ(t ₂ ).

In some cases, the C2 obtained by using the quartic Hermitian interpolation method may not guarantee sin(C2)=0. At this time, the interpolation conditions of the Hermitian interpolation coefficient equations can be appropriately relaxed. If a certain known condition is removed, the linear equations become:

Specifically, at this time rank(A|y)=rank(A)<5, then the interpolation coefficient vector X has countless solutions, and one can be obtained from the solutions that satisfies the requirements of sin(C1)=0 and sin(C2)=0 Solution (that is, obtain the expression Φ(t) of the transition audio phase).

Note that the initial phase of the phase signal of the second sound particle is 0, and it is necessary to ensure that C2=2π*Φ(t ₂ ) is an integer multiple of π, that is, Φ(t ₂ )∈N, N is an integer. Depend on Obtain the basic analysis U. When searching for appropriate basic solution coefficients, we can use optimization methods to make Φ(t ₂ )∈N as close to an integer as possible.

(3) According to the audio signal function expression, use the instantaneous phase obtained by interpolation to construct "transition audio".

Among them, the instantaneous phase of the "transition audio" is obtained through Hermitian interpolation, and the "transition audio" is constructed using the audio signal function expression y(t)=A(t)sin(Φ(t)) .

In addition, in order to ensure that the frequency of the transition section does not mutate, it is best to add appropriate constraints as follows:

Constraint 1: The frequency of the transition section (i.e., the transition audio) is between the frequency domains of the two end points (i.e., the two sound particles spliced at the left and right ends of the transition audio); Constraint 2: The frequency of the transition section is monotonic.

(4) Based on the constructed "transition audio", the target sound sample is synthesized.

As shown in Figure 3, based on the constructed transition audio 1 and 2, this is used as a bridge to connect sound particles 1 and 2, and to connect sound particles 2 and 3, respectively, to achieve the synthesis of sound particles.

That is, this embodiment provides a sound particle synthesis method. According to the recorded audio signal, the interception mark points are located and sound particle samples are extracted to construct a sound particle sample library. According to the sound particle sample library, in the sound synthesis stage, the sound signal is compared with the functional expression of the Hilbert transform, and the instantaneous phase and instantaneous frequency of the sound signal are interpolated through the Hermitian interpolation algorithm. Instantaneous amplitude, generating a "transition function". The "transition function" ensures that adjacent audio particles are continuously differentiable in instantaneous amplitude, frequency, and phase, and uses the "transition function" to synthesize "transition audio", which serves as a bridge to splice adjacent sound particles and synthesize target sound samples.

In order to verify the effectiveness and superiority of a sound particle synthesis method proposed by the present invention, the present invention constructs sound particles 1 and 2, and directly synthesizes these two sound particles and uses a splicing optimization algorithm to construct a "transition" "Audio" is synthesized (that is, synthesized using the method provided by the present invention), and the spectrum of the synthesized sound in two splicing methods is obtained.

As shown in Figure 4, two single-frequency sound particles are directly synthesized without using "transition audio". It can be clearly seen that at the splicing point, the sound wave has an impact at the splicing point. When the synthesized sound is played, an amazing sound is heard. Displeased "pop". As shown in Figure 5, the "transitional audio" constructed using the sound particle synthesis method proposed by the present invention connects two adjacent sound particles, which can effectively eliminate the impact at the splicing point, ensure the continuity of the sound wave, and avoid sound playback Stuttering occurs.

A sound particle synthesis method and device provided by embodiments of the present invention at least include the following technical effects:

(1) The present invention realizes the natural splicing of sound particles by constructing transitional audio to splice adjacent sound particles, solves the "step" phenomenon that occurs in the splicing process of existing sound particles, avoids the occurrence of "pops" in synthetic sounds, and ensures improves the coherence of synthesized sounds.

(2) Based on the physical meaning of the Hilbert transform and combined with the Hermitian interpolation algorithm, the present invention proposes a sound particle synthesis method. This method uses the SHC function to extract the main harmonics of the sound sample based on the recorded sound sample. Locate and intercept markers and extract sound samples to build a sound particle sample library. In the sound synthesis stage, the sound signal is compared with the Hilbert transform, and the instantaneous phase and frequency of the sound signal are interpolated through the Hermite interpolation algorithm. The instantaneous amplitude is obtained through the interpolation algorithm. , generating a "transition function" to ensure that adjacent audio particles are continuously differentiable in amplitude, frequency, and phase, and using the "transition function" to synthesize "transition audio" to splice adjacent sound particles to achieve natural splicing of sound particles.

(3) Compared with other sound synthesis algorithms (for example, the "death trajectory-rebirth trajectory" frequency tracking algorithm proposed by Wen Mcaulay R et al.), the sound particle synthesis method proposed by the present invention has a small amount of calculation and high efficiency. High, the optimization effect is better, it can effectively eliminate the "step" phenomenon produced during the splicing process, and ensure the coherence of the synthesized sound.

(4) The sound particle synthesis method proposed by the present invention has a wide range of applications, is simple, fast, and highly flexible, and ensures the sound quality requirements of the target sound in active sound design.

Finally, it should be noted that the above specific embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to examples, those of ordinary skill in the art will understand that the technical solutions of the present invention can be carried out. Modifications or equivalent substitutions without departing from the spirit and scope of the technical solution of the present invention shall be included in the scope of the claims of the present invention.

Claims (7)

1. A method of synthesizing sound particles, comprising the steps of:

step 1, collecting an original sound sample;

step 2, constructing a sound particle sample library according to the original sound sample, and adding indexes of sound particles;

step 3, according to the vehicle condition input information, reading the sound particles in the sound particle sample library based on the indexes of the sound particles;

step 4, constructing transition audio according to the sound particles read by the index; splicing adjacent sound particles through the transition audio frequency to synthesize a target sound sample;

in the step 4, adjacent sound particles of transition audio splicing are constructed according to the sound particles read by the index; the audio signal and the Hilbert transform function expression are subjected to analogy to obtain an audio signal function expression;

obtaining phase information and frequency information of transition audio through a Hermite interpolation algorithm, and obtaining amplitude information through the interpolation algorithm;

constructing the transition audio according to the audio signal function expression, and the phase information, the frequency information and the amplitude information of the transition audio;

the audio signal function expression y (t) is as follows:

y(t)＝A(t)sin(Φ(t))

wherein A (t) is the instantaneous amplitude of the audio signal, phi (t) is the instantaneous phase of the audio signal,Is audio frequencyThe derivative of the instantaneous phase of the signal, F (t) being the instantaneous frequency of the audio signal, C being the integration constant, t2-t1 being the duration of the audio signal;

the implementation mode of obtaining the phase information of the transition audio through the Hermite interpolation algorithm is as follows:

if adjacent sound particles are respectively sound particles i and sound particles i+1, the transition audio used for splicing the sound particles i and the sound particles i+1 is recorded as transition audio i;

the instantaneous phase curve of the synthesized sound obtained after splicing is continuous and conductive;

wherein the time interval corresponding to the sound particle i is t _i ～t _i+1 The sound particles i are at t _i+1 The instantaneous phase of the moment is phi (t _i+1 ) The method comprises the steps of carrying out a first treatment on the surface of the The time interval corresponding to the sound particle i+1 is t _i+2 ～t _i+3 The sound particle i+1 is at t _i+2 The instantaneous phase of the moment is phi (t _i+2 )；

The corresponding time interval of the transition audio i is t _i+1 ～t _i+2 The method comprises the steps of carrying out a first treatment on the surface of the The left end point of the transition audio i is t of the sound particle i _i+1 An audio signal at a moment, wherein the right end point of the transition audio i is the sound particle i+1 at t _i+2 An audio signal of time;

the sound particle i is at t _i+1 The integral constant of the instantaneous phase at the moment is C1; the sound particle i+1 is at t _i+2 The integral constant of the instantaneous phase at the moment is C2;

preset t _i Is 0, c1=0, according to the instantaneous phase Φ (t _i+1 ) First derivative of instantaneous phase of left and right endpointsAnd->Second derivative of left and right endpoint instantaneous phase +.>And->The expression Φ (t) for solving the phase of the transition audio i by four Hermite interpolation methods is obtained as follows:

Φ(t)＝at ⁴ +bt ³ +ct ² +dt+e

wherein a, b, c, d, e is a polynomial coefficient of the Hermite interpolation to be solved;

using a first interpolation condition, the first set of linear equations is expressed as:

wherein, rank (a|y) =rank (a) =5, the polynomial coefficient of the hermite interpolation has a unique solution, and the expression Φ (t) of the audio phase can be transited, and c2=2pi×Φ (t) _i+2 )；

When sin (C2) obtained by C2 obtained according to the first interpolation condition is not 0, replacing the first interpolation condition by adopting a second interpolation condition, and replacing the first linear equation set by adopting a second linear equation set; the second system of linear equations is expressed as:

where rank (a|y) =rank (a) < 5, there are innumerable solutions to the polynomial coefficients of the hermite interpolation, from which an expression Φ (t) is obtained that satisfies the transition audio phase of sin (C1) =0 and sin (C2) =0.

2. The sound particle synthesis method according to claim 1, wherein the implementation manner of step 1 is as follows: selecting a target vehicle type, and recording an audio signal corresponding to the target vehicle type as the original sound sample;

wherein the audio signal includes engine sound information.

3. The sound particle synthesis method according to claim 1, wherein the implementation manner of the step 2 is as follows: positioning a interception mark point and extracting sound particles according to the original sound sample, and constructing the sound particle sample library;

and the head-tail phases of the sound particles in the sound particle sample library are all 0.

4. The method of synthesizing sound particles according to claim 1, wherein in the step 2, the index of the added sound particles is implemented by: extracting sound particle fragments in the sound particle sample library, and sequentially placing the sound particle fragments; numbering each sound particle and making an index array;

in the step 3, obtaining the index number of the corresponding sound particle according to the vehicle condition input information, and reading the sound particle in the sound particle sample library according to the index number; the vehicle condition input information comprises engine rotation speed information, vehicle speed information and torque information.

5. The sound particle synthesis method according to claim 1, wherein the frequency information of the transition audio satisfies the following constraint:

the frequency of the transition audio frequency is between the frequencies of the left endpoint and the right endpoint;

the transition audio is monotonic in frequency.

6. The sound particle synthesis method according to claim 1, wherein the amplitude of the transition audio satisfies a continuous condition; according to the amplitude A (t _i+1 ) And A (t) _i+2 ) And interpolating to obtain the amplitude information of the transition audio.

7. An acoustic particle synthesizer, comprising:

the acquisition unit is used for acquiring an original sound sample;

the sound particle sample library construction unit is used for obtaining a sound particle sample library according to the original sound sample;

the index unit is used for obtaining indexes of the sound particles based on the sound particle sample library;

a sound particle reading unit for reading sound particles in the sound particle sample library based on the index of the sound particles according to the vehicle condition input information;

the transition audio construction unit is used for obtaining transition audio according to the sound particles read by the index;

the target sound sample synthesis unit is used for splicing adjacent sound particles through the transition audio frequency to synthesize a target sound sample;

the sound particle synthesizing apparatus is for realizing the sound particle synthesizing method according to any one of claims 1 to 6.