Rotary sound source beam forming method for improving wavelet transform

Abstract

The invention provides a rotating sound source beam forming method based on improved wavelet transform, which comprises the following steps of: firstly, executing Morse wavelet transform on data received by each microphone, and converting a time domain signal into a time-frequency domain signal; then, parameters related to the Doppler effect are calculated, and a Doppler effect correction Green function is introduced; then, calculating a weight vector and a cross-spectrum matrix formed by an array beam; and finally, grid division is performed on a plane where a target signal is located, and beam forming is performed on each grid point by using the time-frequency array signal model on each wavelet domain so that sound imaging of a target sound source can be realized. Compared with a traditional beam forming method based on Fourier transform, the beam forming method based on improved wavelet transform has the advantage that the real-time imaging quality and background noise suppression are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of sound source localization technology, specifically to an improved wavelet transform method for rotating sound source beamforming. Background Technology

[0002] The rapid development of low-altitude, small, and slow-moving rotary-wing unmanned aerial vehicles (UAVs) in recent years has led to frequent incidents of unauthorized flights and the carrying of dangerous items, posing a serious threat to public safety. How to effectively detect and locate UAVs in complex low-altitude environments has become an urgent social problem. Acoustic imaging technology based on microphone arrays uses microphone arrays to collect audio data and selects appropriate beamforming algorithms to calculate the distribution of sound sources in the scanning plane. By fusing the sound source information with real-world images, the spatial distribution of sound sources is visualized, facilitating the monitoring, tracking, and location of equipment.

[0003] Beamforming-based sound source localization methods are constantly emerging. For signals with time-varying statistical functions, it is necessary to analyze the signal simultaneously in the time and frequency domains to obtain detailed signal characteristics. Commonly used time-frequency analysis methods include Short-Time Fourier Transform (STFT), Gabor Transform, and Wavelet Transform. Among them, Wavelet Transform has strong time-domain resolution for the high-frequency part of the signal and strong frequency-domain resolution for the low-frequency part of the signal, making it a powerful tool for processing non-stationary signals, such as high-speed rotating sound sources from low-altitude UAVs. This invention introduces an improved wavelet transform-based beamforming method for rotating sound sources.

[0004] The existing beamforming methods have the following problems:

[0005] 1. When locating a high-speed rotating UAV, the Doppler effect is observed in the noise of the fan and propeller; the rotation of the sound source relative to the observer causes a frequency shift, which in turn leads to a time-varying Green's function.

[0006] 2. The rotating sound source beamforming method based on improved wavelet proposed in this paper directly incorporates Morse wavelet and Doppler effect into Green's function to generate a time-frequency domain acoustic image. Summary of the Invention

[0007] The purpose of this invention is to provide an improved wavelet transform method for rotating sound source beamforming to solve the problems mentioned in the background art.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: an improved wavelet transform method for rotating sound source beamforming, comprising the following steps:

[0009] A1: First, perform Morse wavelet transform on the data received by each microphone to convert the time-domain signal into a time-frequency domain signal;

[0010] A2: Next, calculate the parameters related to the Doppler effect and introduce the Doppler effect to correct the Green's function;

[0011] A3: Then, calculate the weight vector and cross-spectral matrix of the array beamformation;

[0012] A4: Finally, the plane containing the target signal is divided into grids, and the time-frequency array signal model in each wavelet domain is used to perform beamforming on each grid point to achieve acoustic imaging of the target sound source.

[0013] Preferably, in A1, the Morse wavelet transform of the received audio data is performed as follows:

[0014] Traditional beamforming algorithms mainly use Fourier transform to transform signals from the time domain to the frequency domain. Their time and frequency resolutions are fixed and lack adaptability. Using continuous wavelet transform to process transient signals in the time and frequency domains can improve the resolution in both domains. Continuous wavelet transform will be used to construct a time-frequency array signal model.

[0015] For the nth sensor, its distance to the sound source is in It is the position vector of the nth sensor. Is the sound source at imaging time t i The position vector at time. The speed of sound is c0, and the time (i.e., time delay) required for sound to travel from the sound source to the sensor is τ. n for:

[0016] The generalized Morse wavelet is used to perform continuous wavelet transform on the signal acquired by the sensor. The Morse wavelet is defined in the frequency domain as:

[0017]

[0018] in Used for wavelet normalization, U(ω) represents the unit step function. The parameters for the wavelet form are controlled by γ, with γ = 3 and γ = 4. For the nth sensor signal P n Its continuous wavelet transform is:

[0019]

[0020] In the formula This represents the sensor sample at time t+τ′.

[0021] Preferably, in A2, the calculation of the Doppler effect parameters involves introducing a Doppler effect-corrected Green's function; the specific process is as follows:

[0022] Assuming a point source At time 0 Starting from a velocity of x0, the governing equation for the sound pressure field p is:

[0023]

[0024] Where c0 is the speed of sound, and δ is the Dirac function. Using spatial coordinates, and utilizing the relationship between velocity potential and sound pressure, Substituting, we get:

[0025]

[0026] The velocity potential solution based on the free-field Green's function is:

[0027]

[0028] In the formula when hour, It is time t i The source location at time satisfies Using the Jacobian property of the Dirac delta function:

[0029]

[0030] The solution for the velocity potential is:

[0031]

[0032] In the formula Let n be the associated location of the nth sensor. This indicates the nth sensor (located in the array). ) and source location In the instantaneous time t i The distance between them. This is caused by the Doppler effect. The sound pressure solution is obtained by taking the time derivative of the velocity potential:

[0033]

[0034] In the formula

[0035] The above derivation (related to the Doppler effect) and (Related to near-field amplification effect) and other parameters.

[0036] Preferably, in A3, the weight vector formed by the computational array beam is... And the cross-spectral matrix A; the specific process is as follows: after obtaining α and β, the array rebroadcast model corresponding to the point sound source is:

[0037]

[0038] Calculate the weight vector using the array propagation vector.

[0039] In the formula, ||·|| represents the l2 norm.

[0040] After wavelet transform, in the equation A function to obtain time t and wavelet scaling parameter s:

[0041]

[0042] The relationship between s and angular frequency ω is: In the formula, f0 is the peak frequency, and T s This represents the sampling time interval. Rewritten as Thus, the frequency is obtained. and time Wavelet transform output Y at point n :

[0043]

[0044] In the formula This indicates that due to the Doppler effect, a given frequency ω i The rotating source, the imaging frequency sensed by the nth sensor. Finally, the cross matrix A is obtained.

[0045]

[0046] Preferably, in A4, the rotating sound source beamforming algorithm achieves acoustic imaging of the target sound source, and the specific process is as follows: Therefore, the imaging frequency is ω. i The moving source in an imaging time t i (From the perspective of the source) location The array beamforming at that location is as follows:

[0047]

[0048] b. The output represents the positions on the imaging plane. Imaging frequency ω and imaging time t iThe intensity or energy distribution of the processed sound source signal generates a time-frequency domain acoustic image, enabling acoustic imaging of a high-speed rotating sound source. This allows for more accurate reflection of the sound source's characteristics (such as frequency changes over time, spatial distribution of different frequency components, etc.) and location information in the time-frequency domain.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0050] Morse wavelet transform provides multi-resolution analysis capabilities, enabling the analysis of signals at different scales (frequency resolution) to better capture signal details, which is beneficial for identifying and locating complex sound sources.

[0051] Morse wavelet transform can adapt to the local characteristics of a signal, making it suitable for processing non-stationary signals. It can provide localized information in both the time and frequency domains, enabling beamforming methods to more accurately locate sound sources and distinguish between nearby sound sources. Attached Figure Description

[0052] Figure 1 The method flowchart for the invention;

[0053] Figure 2 This is a comparison diagram of the invented method and the original method. Detailed Implementation

[0054] The present invention will now be described in detail with reference to the accompanying drawings.

[0055] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: an improved wavelet transform method for rotating sound source beamforming, comprising the following steps:

[0056] A1: First, perform Morse wavelet transform on the data received by each microphone to convert the time-domain signal into a time-frequency domain signal;

[0057] A2: Next, calculate the parameters related to the Doppler effect and introduce the Doppler effect to correct the Green's function;

[0058] A3: Then, calculate the weight vector and cross-spectral matrix of the array beamformation;

[0059] A4: Finally, the plane containing the target signal is divided into grids, and the time-frequency array signal model in each wavelet domain is used to perform beamforming on each grid point to achieve acoustic imaging of the target sound source.

[0060] Preferably, in A1, the Morse wavelet transform of the received audio data is performed as follows:

[0061] Traditional beamforming algorithms mainly use Fourier transform to transform signals from the time domain to the frequency domain. Their time and frequency resolutions are fixed and lack adaptability. Using continuous wavelet transform to process transient signals in the time and frequency domains can improve the resolution in both domains. Continuous wavelet transform will be used to construct a time-frequency array signal model.

[0062] For the nth sensor, its distance to the sound source is in It is the position vector of the nth sensor. Is the sound source at imaging time t i The position vector at time. The speed of sound is c0, and the time (i.e., time delay) required for sound to travel from the sound source to the sensor is τ. n for:

[0063] The generalized Morse wavelet is used to perform continuous wavelet transform on the signal acquired by the sensor. The Morse wavelet is defined in the frequency domain as:

[0064]

[0065] in Used for wavelet normalization, U(ω) represents the unit step function. The parameters for the wavelet form are controlled by γ, with γ = 3 and γ = 4. For the nth sensor signal P n Its continuous wavelet transform is:

[0066]

[0067] In the formula This represents the sensor sample at time t+τ′.

[0068] Preferably, in A2, the calculation of the Doppler effect parameters involves introducing a Doppler effect-corrected Green's function; the specific process is as follows:

[0069] Point sound source At time 0 Starting from a velocity of x0, the governing equation for the sound pressure field p is:

[0070]

[0071] Where c0 is the speed of sound, and δ is the Dirac function. Using spatial coordinates, and utilizing the relationship between velocity potential and sound pressure, Substituting, we get:

[0072]

[0073] The velocity potential solution based on the free-field Green's function is:

[0074]

[0075] In the formula when hour, It is time t i The source location at time satisfies Using the Jacobian property of the Dirac delta function:

[0076]

[0077] The solution for the velocity potential is:

[0078]

[0079] In the formula Let n be the associated location of the nth sensor. This indicates the nth sensor (located in the array). ) and source location In the instantaneous time t i The distance between them. This is caused by the Doppler effect. The sound pressure solution is obtained by taking the time derivative of the velocity potential:

[0080]

[0081] In the formula

[0082] The above derivation (related to the Doppler effect) and (Related to near-field amplification effect) and other parameters.

[0083] Preferably, in A3, the weight vector formed by the computational array beam is... And the cross-spectral matrix A; the specific process is as follows: after obtaining α and β, the array rebroadcast model corresponding to the point sound source is:

[0084]

[0085] Calculate the weight vector using the array propagation vector.

[0086] In the formula, ||·|| represents the l2 norm.

[0087] After wavelet transform, in the equation A function to obtain time t and wavelet scaling parameter s:

[0088]

[0089] The relationship between s and angular frequency ω is: In the formula, f0 is the peak frequency, and T s This represents the sampling time interval. Rewritten as Thus, the frequency is obtained. and time Wavelet transform output Y at point n :

[0090]

[0091] In the formula This indicates that due to the Doppler effect, a given frequency ω i The rotating source, the imaging frequency sensed by the nth sensor. Finally, the cross matrix A is obtained.

[0092]

[0093] Preferably, in A4, the rotating sound source beamforming algorithm achieves acoustic imaging of the target sound source, and the specific process is as follows: Therefore, the imaging frequency is ω. i The moving source in an imaging time t i (From the perspective of the source) location The array beamforming at that location is as follows:

[0094]

[0095] b. The output represents the positions on the imaging plane. Imaging frequency ω and imaging time t i The intensity or energy distribution of the processed sound source signal generates a time-frequency domain acoustic image, enabling acoustic imaging of a high-speed rotating sound source. This allows for more accurate reflection of the sound source's characteristics (such as frequency changes over time, spatial distribution of different frequency components, etc.) and location information in the time-frequency domain.

[0096] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An improved wavelet transform based on a rotating acoustic source beamforming method, characterized by, The method comprises the following steps: A1: performing Morse wavelet transform on the data received by each microphone to convert the time domain signal into a time-frequency domain signal; A2: calculating a Doppler effect related parameter and introducing a Doppler effect correction Green function; A3: calculating a weight vector and a cross-spectrum matrix of array beam forming; A4: dividing a plane where the target signal is located into grids, using a time-frequency array signal model on each wavelet domain to perform beam forming on each grid point, and realizing acoustic imaging of the target sound source.

2. The improved wavelet transform-based rotational acoustic source beamforming method of claim 1, wherein, In A1, the Morse wavelet transform on the sound data received by each microphone is performed, and the specific process is as follows: For the nth sensor, its distance to the sound source is where is the position vector of the nth sensor, is the position vector of the sound source at the imaging time t i ; the sound speed is c0, and the time τ n required for sound to propagate from the sound source to the sensor is: The Morse wavelet is defined in the frequency domain as follows: where for normalization of the wavelet, U(ω) denotes the unit step function, and γ controls the parameters of the wavelet form, taken γ = 3 and for the nth sensor signal P n its continuous wavelet transform is: In the formula denotes the sensor sample at time t+τ'.

3. The improved wavelet transform-based rotational acoustic source beamforming method of claim 1, wherein, In A2, the specific process is as follows: The control equation of the sound pressure field p is as follows: where c0is the sound speed, δ is the Dirac function, for spatial coordinates, and gives: The velocity potential solution based on the free-field Green function is: where When , is the source position at time t i , satisfying Using the Jacobian property of the Dirac delta function: The solution of the velocity potential is as follows where is the associated position of the bth sensor, represents the distance between the nth sensor (located at array ) and the source position at the instant time t i ; and is due to the Doppler effect; the acoustic pressure solution is obtained by taking the time derivative of the velocity potential: In the formulae 4. The method of claim 1, wherein the wave field is a sound field. In A3, the weight vector of the array beamforming is calculated and the cross-spectrum matrix A; the specific process is as follows: after α and β are obtained, the array transfer vector corresponding to the point sound source is: Calculating weight vectors by array propagation of vectors where || · || denotes the l2 norm, After the wavelet transform, the equation becomes The function of time t and the scale parameter s of the wavelet is obtained The relationship of s to the angular frequency ω is: where f0is the peak frequency, T s is the sampling time interval; Rewritten as Further, the frequency is and the time at which the wavelet transform output Y n : wherein represents the imaging frequency perceived by the nth sensor due to the Doppler effect for a rotating source of given frequency ω i Finally, by proceeding to obtain the cross-matrix A,​ 5. The method of claim 1, wherein the wave field is a sound field. In A4, the rotating sound source beam forming algorithm is used to realize acoustic imaging of the target sound source.

6. The improved wavelet transform method for rotating sound source beamforming as described in claim 5, characterized in that, In A4, the detailed procedure is as follows: the imaging frequency is ω i of the moving source at an imaging time t i Array beamforming at the position​ b output representing the intensity or energy distribution of the sound source signal at each position on the imaging plane imaging frequency ω and imaging time t i The processed intensity or energy distribution of the sound source signal is then used to generate a sound image in the time-frequency domain, enabling acoustic imaging of high-speed rotating sound sources and more accurate reflection of the characteristics and position of the sound source in the time-frequency domain.

Images