A noise reduction method for helicopter voice processing systems

Abstract

This invention provides a noise reduction method for a helicopter speech processing system, comprising the following steps: First, an audio signal is input and placed into a preset buffer in a window noise estimation. The window size should be set between 10ms and 30ms, during which longer speech is considered constant. When the input signal is less than a threshold θ, the input signal is considered noise; when the input signal is greater than the threshold θ, the input signal is considered real-time speech. The useful signal is obtained by subtracting the noise signal spectrum from the real-time speech signal. Following this logic, the prior clean noise λ is obtained. x (k), and the useful signal λ d (k), and calculate the signal-to-noise ratio at this time. The present invention provides a noise reduction method for helicopter voice processing system, which reduces the noise intensity in the collected voice by running the Wiener filtering algorithm internally, suppresses the noise of rotor, body vibration, engine and other noise, and ensures that the voice heard in the headphones is clear and free of noise.

Description

Technical Field

[0001] This invention relates to the field of speech processing, and more particularly to a noise reduction method for a helicopter speech processing system. Background Technology

[0002] Helicopters generate a lot of noise during operation and flight, including noise from engine operation, noise from airframe vibration, and specific noise from rotor rotation cutting through the air. The intensity of these noises can be as high as 110dB or more, and they will enter the voice processing system through the helmet microphone.

[0003] Currently, traditional speech denoising methods have certain limitations in helicopter speech processing systems. For example, simple filtering methods will lose some speech signals while removing noise, resulting in speech distortion. On the other hand, some denoising methods based on statistical models often have difficulty accurately estimating noise characteristics when dealing with complex helicopter noise environments, thus affecting the denoising effect.

[0004] In addition, the flight environment of helicopters is complex and changeable, and the intensity and frequency characteristics of noise will also change with the flight status.

[0005] Therefore, it is necessary to provide a noise reduction method for helicopter voice processing systems to solve the above-mentioned technical problems. Summary of the Invention

[0006] This invention provides a noise reduction method for a helicopter voice processing system, which solves the problem that different noises are generated in the cabin of a helicopter during flight, and that the noise enters the voice processing system from the helmet microphone and eventually reaches the headset, affecting the call quality.

[0007] To address the aforementioned technical problems, this invention provides a noise reduction method for a helicopter voice processing system, comprising the following steps:

[0008] S1. First, input the audio signal and input this audio signal into the preset buffer in the window noise estimation. The window size should be set between 10ms and 30ms. During this time, long voices are considered constant.

[0009] S2. When the input signal is less than the threshold θ, the input signal is considered to be a noise signal. When the input signal is greater than the threshold θ, the input signal is considered to be a real-time voice signal. The useful signal can be obtained by subtracting the noise signal spectrum from the real-time voice signal.

[0010] S3. Following this logic, we obtain the prior pure noise λ. x (k), and the useful signal λ d (k), and calculate the signal-to-noise ratio at this time.

[0011] S4, λ in S3 x (k), λ d (k), ξ k The initial parameters are stored in the memory chip for subsequent use by the MMSE-STSA algorithm and the Wiener filter algorithm;

[0012] S5. At this point, based on the default Wiener filter function in S3, the Wiener filter formula is... The ξ input to the window noise estimation function is called up. k For input signal X k Perform filtering, and define the filtered data as... Send it to the MMSE-STSA algorithm;

[0013] S6, Comparison If these two values ​​are the same, then data convergence is complete; if they are not the same, then... Substitute into the formula to calculate ξ k and ξ k The data is fed into the Wiener filter to update the filtering algorithm H(ω). k );

[0014] S7. Finally, by analogy, the Wiener filter is continuously updated until convergence is achieved.

[0015] Preferably, step S4 includes the following steps:

[0016] S41. Run the MMSE-STSA algorithm and obtain the corresponding formula using the principle of least mean square estimation:

[0017]

[0018] S42, where the prior noise signal-to-noise ratio is ξ stored in the chip in the early stage. k Y k For the voice signal acquired at this time, the posterior signal-to-noise ratio Given, according to the formula v can be calculated k I0 and I1 are Bessel functions, from which the following can be calculated: The value of .

[0019] Preferably, the value in S42 is the value of the useful signal estimated based on the least mean square estimation principle, defined as...

[0020] Preferably, the window noise estimation in S1 includes the following methods:

[0021] S11, Noise estimation algorithm for statistical models;

[0022] S12, Noise estimation algorithm for signal subspace;

[0023] S13, Noise estimation algorithm in deep learning;

[0024] S14. Hybrid noise estimation algorithm combining multiple algorithms.

[0025] Preferably, the noise estimation algorithm for the statistical model in S11 includes the following steps:

[0026] S111, Minimum Statistic Tracking Algorithm;

[0027] S112, Recursive Averaging Algorithm.

[0028] Preferably, the noise estimation algorithm for the signal subspace in S12 includes the following steps:

[0029] S121, Eigenvalue decomposition algorithm;

[0030] S122, Singular Value Decomposition Algorithm.

[0031] Preferably, the deep learning noise estimation algorithm in S13 includes the following steps:

[0032] S131, Deep Neural Network;

[0033] S132, Convolutional Neural Network.

[0034] Preferably, the hybrid noise estimation algorithm combining multiple algorithms in S14 includes the following steps:

[0035] S141, Combining statistical models with deep learning;

[0036] S142, Multi-microphone noise estimation.

[0037] Preferably, the combination of statistical model and deep learning in S141 can combine statistical model-based algorithms and deep learning algorithms to give full play to the advantages of both.

[0038] Preferably, the multi-microphone noise estimation in S142 can utilize multiple microphones on the helicopter to collect signals, and fuse the information from multiple microphones through signal processing algorithms to improve the accuracy of noise estimation.

[0039] Compared with related technologies, the noise reduction method for a helicopter voice processing system provided by this invention has the following beneficial effects:

[0040] This invention provides a noise reduction method for a helicopter voice processing system. By running an internal Wiener filtering algorithm, the noise intensity in the acquired speech is reduced, and the noise from the rotor, airframe vibration, and engine is suppressed, ensuring that the speech heard in the headphones is clear and free of noise.

[0041] The short-time spectral amplitude estimate of the noise is calculated by running the MMSE-STSA algorithm, and this parameter is used as the basic parameter of the Wiener algorithm to improve the convergence speed of the Wiener filtering algorithm, ensure that the Wiener filtering algorithm can converge quickly, and improve the algorithm's response speed. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the structure of a first embodiment of a noise reduction method for a helicopter voice processing system provided by the present invention;

[0043] Figure 2 This is a schematic diagram of the hardware principle;

[0044] Figure 3 This is a schematic diagram of the hardware layout;

[0045] Figure 4 A schematic diagram of the hardware principle with CS320AIC3106 as the core device;

[0046] Figure 5 A schematic diagram of the hardware principle using JTLV320AC23 as the core device;

[0047] Figure 6 This is a schematic diagram of the software processing procedure;

[0048] Figure 7 This is a schematic diagram of the hardware circuit.

[0049] Figure 8 This is a schematic diagram of the second embodiment of a noise reduction method for a helicopter voice processing system provided by the present invention. Detailed Implementation

[0050] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0051] First Embodiment

[0052] Please refer to the following: Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 and Figure 7 ,in, Figure 1 This is a schematic diagram of the structure of a first embodiment of a noise reduction method for a helicopter voice processing system provided by the present invention; Figure 2 This is a schematic diagram of the hardware principle; Figure 3 This is a schematic diagram of the hardware layout; Figure 4 A schematic diagram of the hardware principle with CS320AIC3106 as the core device; Figure 5 A schematic diagram of the hardware principle using JTLV320AC23 as the core device; Figure 6 This is a schematic diagram of the software processing procedure; Figure 7 This is a schematic diagram of the hardware circuit. A noise reduction method for a helicopter voice processing system includes the following steps:

[0053] S1. First, input the audio signal and input this audio signal into the preset buffer in the window noise estimation. The window size should be set between 10ms and 30ms. During this time, long voices are considered constant.

[0054] S2. When the input signal is less than the threshold θ, the input signal is considered to be a noise signal. When the input signal is greater than the threshold θ, the input signal is considered to be a real-time voice signal. The useful signal can be obtained by subtracting the noise signal spectrum from the real-time voice signal.

[0055] S3. Following this logic, we obtain the prior pure noise λ. x (k), and the useful signal λ d (k), and calculate the signal-to-noise ratio at this time.

[0056] S4, λ in S3 x (k), λ d (k), ξ k The initial parameters are stored in the memory chip for subsequent use by the MMSE-STSA algorithm and the Wiener filter algorithm;

[0057] S5. At this point, based on the default Wiener filter function in S3, the Wiener filter formula is... The ξ input to the window noise estimation function is called up. k For input signal X k Perform filtering, and define the filtered data as... Send it to the MMSE-STSA algorithm;

[0058] S6, Comparison If these two values ​​are the same, then data convergence is complete; if they are not the same, then... Substitute into the formula to calculate ξ k and ξ k The data is fed into the Wiener filter to update the filtering algorithm H(ω). k );

[0059] S7. Finally, by analogy, the Wiener filter is continuously updated until convergence is achieved.

[0060] S4 includes the following steps:

[0061] S41. Run the MMSE-STSA algorithm and obtain the corresponding formula using the principle of least mean square estimation:

[0062]

[0063] S42, where the prior noise signal-to-noise ratio is ξ stored in the chip in the early stage. k Y k For the voice signal acquired at this time, the posterior signal-to-noise ratio Given, according to the formula v can be calculated k I0 and I1 are Bessel functions, from which the following can be calculated: The value of .

[0064] The value in S42 is the value of the useful signal estimated based on the least mean square estimation principle, defined as follows:

[0065] The digital voice automatic gain control method eliminates the reliance on imported automatic gain control circuits, overcomes the problem of insufficient range of domestically produced analog automatic gain circuits, simplifies the overall circuitry, saves hardware costs, and can be flexibly configured into digital circuits, improving configuration efficiency. Its main advantages include the following:

[0066] reliability:

[0067] Digital voice automatic gain control reduces the overall hardware circuitry and can be flexibly configured into various ports, improving the overall reliability of the device.

[0068] Analog automatic gain control circuit: Increases the number of hardware circuits in the whole machine and reduces the overall reliability.

[0069] Economic benefits:

[0070] The digital voice automatic gain control method does not add any components, thus reducing the overall hardware cost;

[0071] Analog automatic gain control circuit: Increases the number of hardware circuits in the whole machine, and increases the cost of components.

[0072] Please refer to the following: Figure 6 It is known that, considering the bandwidth of 300–3400 Hz, the center frequency f of the speech signal is confirmed. c =1850Hz. Since the purpose of this method is to achieve automatic gain control of the speech signal amplitude, it only needs to detect the peak value of the signal. There is no need to specifically judge the voiced and unvoiced sounds in the speech, and the detected envelope signal does not need to be very precise. Therefore, the Hilbert transform method can be used, and the formula is: The original signal is x(t), and the Hilbert transform of the original signal is: Once the analytical signal z(t) is obtained, the envelope signal of the speech signal can be obtained by taking the modulus of the analytical signal. The baseband frequency parameter is set between 0.1 and 0.5 Hz and adjusted according to the required accuracy.

[0073] After obtaining the peak value of the speech signal, the peak value is compared with the corresponding peak value in the gain table. For example, when using the first circuit, the suppression parameter in the gain table is set to 3000 for speech input and 5000 for receiving digital signals. When using the second circuit, the suppression parameter in the gain table is set to 6000 and 10000 for receiving digital signals.

[0074] The volume of the signal is proportionally reduced according to the coefficient in the gain table, and this volume coefficient is assigned to the audio signal in the buffer. The audio signal multiplied by this coefficient is the output signal after automatic gain adjustment.

[0075] Please refer to the following: Figure 4 3D1 is a domestically produced ARM chip HWD32F767, 3D24 is a domestically produced 422 bus chip HWD488, 3D2 is a domestically produced voice acquisition chip CS320AIC3106, 3C123, 3C124, 3C66, and 3C67 are coupling capacitors, 3C305 is a bypass filter capacitor, 3R228 is a pull-up resistor, and 3R51, 3R52, and 3R210 are matching resistors.

[0076] Please refer to the following: Figure 5 The 3D30 chip is a domestically produced ARM chip, FMQL20S484. D42 and D6 are domestically produced 422 bus transceiver chips, XZJ26C32 and XZJ26C31, respectively. 3D2 is a domestically produced voice acquisition chip, CS320AIC3106. C471, C474, C477, and C478 are coupling capacitors. C334, C335, C336, C337, C287, C96, C72, and C69 are also present. C75 and C70 are bypass filter capacitors; R223 and R227 are pull-up resistors; R204, R206, R225, R229, R415, and R417 are pull-down resistors; R21, R30, R31, R34, R164, R165, R516, R166, and R167 are matching resistors; and V7, V8, V11, V13, V71, V74, V75, and V78 are anti-static diodes.

[0077] The first method uses the CS320AIC3106 as the data acquisition chip, the HWD488 as the 422 bus communication interface, and the HWD32F767 as the processor circuit. This circuit requires an input signal of 100-120mV. The second method uses the JTLV320AC23 as the data acquisition chip, the XZJ26C32 and XZJ26C31 as the 422 bus communication interface, and the FMQL20S484 as the processor circuit. This circuit requires an input signal of 200-240mV. The hardware circuit is shown in the figure. Communication between the chips uses the I2S interface.

[0078] Compared with related technologies, the noise reduction method for a helicopter voice processing system provided by this invention has the following beneficial effects:

[0079] This invention provides a noise reduction method for a helicopter voice processing system. By running the Wiener filtering algorithm internally, the noise intensity in the collected speech is reduced, and noise from rotor, airframe vibration, engine, etc. is suppressed, ensuring that the speech heard in the headphones is clear and free of noise.

[0080] The short-time spectral amplitude estimate of the noise is calculated by running the MMSE-STSA algorithm, and this parameter is used as the basic parameter of the Wiener algorithm to improve the convergence speed of the Wiener filtering algorithm, ensure that the Wiener filtering algorithm can converge quickly, and improve the algorithm's response speed.

[0081] Second Embodiment

[0082] Please refer to the following: Figure 8 Based on the noise reduction method for a helicopter voice processing system provided in the first embodiment of this application, the second embodiment of this application proposes another noise reduction method for a helicopter voice processing system. The second embodiment is merely a preferred embodiment of the first embodiment, and the implementation of the second embodiment will not affect the separate implementation of the first embodiment.

[0083] Specifically, the second embodiment of this application provides a noise reduction method for a helicopter speech processing system, which differs in that the window noise estimation in S1 includes the following methods:

[0084] S11, Noise estimation algorithm for statistical models;

[0085] S12, Noise estimation algorithm for signal subspace;

[0086] S13, Noise estimation algorithm in deep learning;

[0087] S14. Hybrid noise estimation algorithm combining multiple algorithms.

[0088] The noise estimation algorithm for the statistical model in S11 includes the following steps:

[0089] S111, Minimum Statistic Tracking Algorithm;

[0090] S112, Recursive Averaging Algorithm.

[0091] Minimum Statistical Tracking Algorithm: This algorithm estimates noise power by tracking the minimum value of the signal power. In helicopter environments, due to the non-stationarity of noise, the minimum statistical tracking algorithm can quickly adapt to changes in noise and provide a more accurate noise estimate.

[0092] Recursive averaging algorithm: The input signal is recursively averaged to estimate the average power of the noise. This algorithm is simple to calculate and suitable for real-time processing, but it may not be accurate enough for rapidly changing noise.

[0093] The noise estimation algorithm for the signal subspace in S12 includes the following steps:

[0094] S121, Eigenvalue decomposition algorithm;

[0095] S122, Singular Value Decomposition Algorithm.

[0096] Eigenvalue decomposition algorithm: The covariance matrix of the signal is decomposed into eigenvalues, dividing the signal into a signal subspace and a noise subspace. By estimating the eigenvalues ​​of the noise subspace, the noise power can be estimated. In helicopter speech processing, this algorithm can effectively remove noise that is not related to the speech signal.

[0097] Singular Value Decomposition (SVD): Similar to eigenvalue decomposition, SVD decomposes the signal matrix to obtain estimates of the signal and noise. This algorithm has good performance when processing high-dimensional data.

[0098] The deep learning noise estimation algorithm in S13 includes the following steps:

[0099] S131, Deep Neural Network;

[0100] S132, Convolutional Neural Network.

[0101] Deep Neural Networks (DNNs): By training DNNs to learn the characteristics of noise, noise estimation can be achieved. They can be trained using a large amount of noisy speech data and corresponding clean speech data to improve the accuracy and robustness of the algorithm. In helicopter environments, DNNs can learn different types of noise features and adapt to complex noise environments.

[0102] Convolutional Neural Networks (CNNs): CNNs process the time-frequency features of speech signals and estimate noise power. CNNs have powerful feature extraction capabilities and can effectively capture the local and global features of noise.

[0103] The hybrid noise estimation algorithm combining multiple algorithms in S14 includes the following steps:

[0104] S141, Combining statistical models with deep learning;

[0105] S142, Multi-microphone noise estimation.

[0106] The combination of statistical model and deep learning in S141 can combine statistical model-based algorithms and deep learning algorithms to give full play to the advantages of both.

[0107] A preliminary noise estimation can be performed using statistical model algorithms first, and then the estimation results can be used as input for deep learning algorithms to further improve the accuracy of noise estimation.

[0108] The multi-microphone noise estimation in S142 can utilize multiple microphones on the helicopter to collect signals, and then fuse the information from multiple microphones through signal processing algorithms to improve the accuracy of noise estimation.

[0109] Beamforming technology can be used to weight and sum the signals from multiple microphones, thereby enhancing the voice signal and suppressing noise.

[0110] Compared with related technologies, the noise reduction method for a helicopter voice processing system provided by this invention has the following beneficial effects:

[0111] This invention provides a noise reduction method for helicopter speech processing systems. The noise estimation algorithm using statistical models, signal subspaces, deep learning, and hybrid algorithms can enhance overall adaptability, simplify computation, improve stability, significantly reduce noise, and enhance the ability to handle complex noise with high robustness.

[0112] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for noise reduction for a helicopter speech processing system, the method comprising: It comprises the following steps: S1, first input the audio signal, input the audio signal into the window noise estimation for preset buffering, the window size should be set between 10ms-30ms, the long voice is regarded as constant; S2, when the input signal is less than the threshold θ value, it is considered that the signal input at this time is a noise signal, when the input signal is greater than the threshold θ value, it is considered that the input signal is a real-time voice signal, the real-time voice signal is subtracted from the noise signal spectrum to obtain the useful signal; S3, according to this logic, the a priori pure noise λ x (k), and the useful signal λ d (k), and the signal-to-noise ratio at this time is calculated S4, λ in S3 x (k), λ d (k), ξ k As initial parameters stored in the memory chip for subsequent MMSE-STSA algorithm and Wiener filter algorithm call; The S4 comprises the following steps: S41, run MMSE-STSA algorithm, get corresponding formula through the principle of least square estimation: S42, the prior noise signal-to-noise ratio is the ξ stored in the chip in advance k , Y k is the voice signal collected at this time, the posterior signal-to-noise ratio It is known that the value of v can be calculated according to the formula k , I0, I1 are Bessel functions, and the value of can be calculated accordingly; The value of S42 The value of S42 S5, at this time, according to the Wiener filter default function in S3, the Wiener filter formula The window noise estimation function is called in ξ k The input signal X k is filtered, and the filtered data is defined as is sent to the MMSE-STSA algorithm; S6、Convergence If the two values are consistent, the data convergence is completed, if not, the Put into the formula, calculate ξ k and send ξ k to the Wiener filter, update the filter algorithm H(ω k ); S7, finally through the method of analogy, constantly update the Wiener filter until the convergence is completed.

2. The noise reduction method for a helicopter speech processing system according to claim 1, characterized in that, The window noise estimation in the S1 comprises the following methods: S11, statistical model noise estimation algorithm; S12, signal subspace noise estimation algorithm; S13, deep learning noise estimation algorithm; S14, combined with a variety of algorithm hybrid noise estimation algorithm.

3. The noise reduction method for a helicopter speech processing system according to claim 2, characterized in that, The statistical model noise estimation algorithm in the S11 comprises the following steps: S111, minimum statistical quantity tracking algorithm; S112, recursive average algorithm.

4. The noise reduction method for a helicopter speech processing system according to claim 2, characterized in that, The signal subspace noise estimation algorithm in the S12 comprises the following steps: S121, eigenvalue decomposition algorithm; S122, singular value decomposition algorithm.

5. The noise reduction method for a helicopter speech processing system according to claim 2, characterized in that, The deep learning noise estimation algorithm in the S13 comprises the following steps: S131, deep neural network; S132, convolutional neural network.

6. The noise reduction method for a helicopter speech processing system according to claim 2, characterized in that, The combined with a variety of algorithm hybrid noise estimation algorithm in the S14 comprises the following steps: S141, statistical model and deep learning combination; S142, multi-microphone noise estimation.

7. The noise reduction method for a helicopter speech processing system according to claim 6, characterized in that, The statistical model and deep learning combination in the S141 refers to the combination of statistical model based algorithm and deep learning algorithm.

8. The noise reduction method for a helicopter speech processing system according to claim 6, characterized in that, The multi-microphone noise estimation in the S142 refers to the use of multiple microphones on the helicopter to collect signals, and fuse the information of multiple microphones through signal processing algorithm.

Images