Fish culture net cage noise tracking method based on optimal smoothness and minimum statistic

Abstract

The invention discloses a fish culture net cage noise tracking method based on optimal smoothness and minimum statistics, and relates to the field of aquaculture. According to the method, the smoothing parameter is automatically adjusted according to the frequency energy distribution characteristic of the input signal, the power spectrum is smoothed, the smoothing result is estimated by combining the minimum value deviation compensation, accurate tracking of the background noise signal is realized, the noise power spectrum component can be effectively extracted in a complex noise environment, and extra detection does not need to be carried out on a target signal. According to the method, a time-frequency adaptive smoothing parameter mechanism is introduced, the background noise power spectrum is tracked and estimated by fusing the minimum statistic, and the noise condition of the sampling signal can be effectively determined under the complex noise condition without training data; the problems that an existing noise separation method is poor in performance and needs a large amount of training data when separating non-stationary noise are solved.

Description

Technical Field

[0001] The present invention relates to the field of aquaculture, and in particular to a fish farming cage noise tracking method based on optimal smoothing and minimum statistics. Background Art

[0002] With the increasing application of passive acoustic technology in aquaculture, removing the impact of complex environmental noise has become a significant challenge. Accurately and efficiently tracking noise is crucial. Under natural conditions, collected acoustic signals are often contaminated by both environmental and human-induced noise. In underwater noise monitoring of deep-sea aquaculture platforms, surface waves, platform equipment operating noise, ship noise, and even distant aircraft passing by can all contribute to interference. These numerous factors combine to make cage monitoring signals non-stationary. Tracking and removing this non-stationary noise has always been a challenge in acoustic signal processing.

[0003] In recent years, various deep learning-based extensions and non-negative matrix factorization (NMF) methods have demonstrated strong performance in noise separation. Deep learning methods train neural network models to learn mapping relationships and output target signals; NMF (non-negative matrix factorization) extracts potential signal and noise components by decomposing mixed signals into a basis matrix and a coding matrix. However, both methods suffer from a significant drawback: they require a large amount of training data to achieve optimal results. Specifically, when using these two methods or their extensions, the raw data of the research object must be collected in advance to separate the pure target signal as training data for the model. In fish farming environments, cage noise interference is ubiquitous. Collecting raw data and then separating the pure signal source is costly, time-consuming, and difficult, making this premise difficult to achieve under real-world conditions. Summary of the Invention

[0004] In view of the above-mentioned deficiencies in the prior art, the fish farming cage noise tracking method based on optimal smoothing and minimum statistics provided by the present invention solves the problem that the prior art is difficult to accurately track the fish farming cage noise.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] A method for tracking noise in fish farming cages based on optimal smoothing and minimum statistics is provided, which comprises the following steps:

[0007] S1. Acquire the sound signal of the fish breeding area, divide it into frames, and convert it into the frequency domain to obtain a frame-level frequency domain signal;

[0008] S2. Calculate the energy of each frequency of each frame signal in the frame-level frequency domain signal to obtain the acoustic energy of each frequency point, and then obtain the power spectrum corresponding to each frame signal in the frame-level frequency domain signal;

[0009] S3. In the same frame-level frequency domain signal, obtain a power spectrum value corresponding to the power spectrum of the previous frame signal after smoothing, and record it as the power spectrum value after smoothing of the previous frame signal;

[0010] S4. Calculating the posterior signal-to-noise ratio of the current frame signal based on the power spectrum value after smoothing the previous frame signal, and obtaining an initial value of the smoothing factor of the power spectrum corresponding to the current frame signal during the smoothing process based on the posterior signal-to-noise ratio of the current frame signal;

[0011] S5, calculating the frequency average ratio of the current frame signal according to the sum of the power spectrum values ​​of the previous frame signal after smoothing and the sum of the power spectrum values ​​corresponding to the current frame signal;

[0012] S6. updating an initial value of a smoothing factor of the current frame signal during the smoothing process based on a frequency average value ratio of the current frame signal to obtain a final value of the smoothing factor of the current frame signal during the smoothing process;

[0013] S7, smoothing the power spectrum corresponding to the current frame signal based on the final value of the smoothing factor of the current frame signal during the smoothing process to obtain a smoothed power spectrum value of the current frame signal;

[0014] S8. Obtaining the minimum value of the power spectrum of the previous L frames of signal after smoothing based on the minimum statistics method, and using the obtained minimum value as the noise estimation result of the current frame signal; where L is the length of the sliding window;

[0015] S9. Calculate the normalized variance of the second-order statistics of the power spectrum value after smoothing the current frame signal, and combine it with the length of the sliding window to generate a deviation compensation factor for the noise estimation result of the current frame signal, thereby obtaining the noise estimation result after compensation of the current frame signal and completing the fish farming cage noise tracking.

[0016] Furthermore, the specific method of converting the sound signal of the fish breeding area into the frequency domain after framing in step S1 includes:

[0017] The sound signal from the fish farming area is segmented into frames with a frame length of 20ms and a frame shift of 10ms. After adding a Hamming window to each frame, a fast Fourier transform is performed to obtain a frame-level frequency domain signal.

[0018] Furthermore, in step S4, the expression for calculating the posterior signal-to-noise ratio of the current frame signal based on the smoothed power spectrum value of the previous frame signal is:

[0019]

[0020] in Indicates the Frame signal The posterior signal-to-noise ratio of each frequency point; Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Indicates the Frame signal The noise power spectrum at each frequency point.

[0021] Furthermore, in step S4, the expression for obtaining the initial value of the smoothing factor of the power spectrum corresponding to the current frame signal during the smoothing process according to the posterior signal-to-noise ratio of the current frame signal is:

[0022]

[0023] in Indicates the Frame signal The initial value of the smoothing factor of the power spectrum corresponding to the frequency point during the smoothing process.

[0024] Furthermore, in step S5, the expression for calculating the frequency average ratio of the current frame signal based on the sum of the smoothed power spectrum values ​​of the previous frame signal and the sum of the power spectrum values ​​corresponding to the current frame signal is:

[0025]

[0026] in Indicates the Frequency average ratio of frame signal; Indicates the The sum of the power spectrum values ​​of the frame signal after smoothing; Indicates the The sum of the power spectrum values ​​corresponding to the frame signal.

[0027] Furthermore, in step S6, the expression for updating the initial value of the smoothing factor of the current frame signal during the smoothing process based on the frequency average ratio of the current frame signal is:

[0028]

[0029]

[0030] in Indicates the Frame signal The final value of the smoothing factor of each frequency point during the smoothing process; For the Correction factor of frame signal; is the weight parameter; For the Correction factor for the frame signal.

[0031] Furthermore, in step S7, the expression for smoothing the power spectrum corresponding to the current frame signal based on the final value of the smoothing factor of the current frame signal during the smoothing process is:

[0032]

[0033] in Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Indicates the Frame signal The final value of the smoothing factor of each frequency point during the smoothing process; Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Shidi Frame signal The power spectrum value corresponding to the frequency point.

[0034] Furthermore, in step S8, the expression for obtaining the minimum value of the power spectrum value after smoothing the previous L frames of signal based on the minimum statistics method is:

[0035]

[0036] in Indicates the Frame signal Noise estimation results of frequency points; Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Indicates taking the minimum value.

[0037] Furthermore, the expression for calculating the normalized variance of the second-order statistics of the power spectrum value after smoothing the current frame signal in step S9 is:

[0038]

[0039]

[0040]

[0041] in Indicates the Frame signal The normalized variance of the second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The first-order statistics of the power spectrum value after smoothing the frequency points; is the weight coefficient; Indicates the Frame signal The second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The square of the power spectrum value after smoothing of the frequency points; Indicates the Frame signal The first-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The power spectrum value after smoothing of the frequency points.

[0042] Furthermore, in step S9, a deviation compensation factor of the noise estimation result of the current frame signal is generated, and the expression of the noise estimation result after compensation of the current frame signal is obtained as follows:

[0043]

[0044]

[0045] in Indicates the Frame signal Noise estimation results after frequency compensation; Indicates the Frame signal Deviation compensation factor of the noise estimation result of each frequency point; It represents the logarithm with the natural constant e as the base; Represents the Digamma function.

[0046] The beneficial effects of the present invention are:

[0047] 1. The present invention automatically adjusts the smoothing parameters according to the frequency energy distribution characteristics of the input signal, smoothes the power spectrum, and combines the minimum deviation compensation to estimate the smoothing result, thereby achieving accurate tracking of background noise signals and effectively extracting noise power spectrum components in complex noise environments without the need for additional detection of the target signal.

[0048] 2. The present invention introduces a time-frequency adaptive smoothing parameter mechanism and integrates the minimum statistic tracking to estimate the background noise power spectrum. Without the need for training data, it can effectively determine the noise situation of the sampled signal under complex noise conditions, solving the problem that existing noise separation methods have poor performance in separating non-stationary noise and require a large amount of training data.

[0049] 3. This invention seamlessly integrates with existing gain function-based denoising methods, significantly improving denoising performance and efficiency. This provides the technical prerequisite for cage acoustic monitoring and creates the potential for rapid, efficient, and accurate analysis of fish vocalizations in cages. Furthermore, the present invention incorporates a preprocessing step, requiring time-domain input data. Raw data obtained by mainstream monitoring methods can be directly fed into the algorithm (if necessary, filtering and other operations can be performed based on actual needs before input), eliminating the need for tedious preprocessing. This method offers low computational complexity and high speed. When combined with various gain functions, it enables on-site sampling and analysis, enabling real-time tracking of cage noise conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Schematic diagram of the process of this method. DETAILED DESCRIPTION

[0051] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0052] like Figure 1 As shown in FIG, the fish farming cage noise tracking method based on optimal smoothing and minimum statistics includes the following steps:

[0053] S1. Acquire the sound signal of the fish breeding area and convert it into the frequency domain after dividing it into frames to obtain a frame-level frequency domain signal;

[0054] S2. Calculate the energy of each frequency of each frame signal in the frame-level frequency domain signal to obtain the acoustic energy of each frequency point, and then obtain the power spectrum corresponding to each frame signal in the frame-level frequency domain signal;

[0055] S3. In the same frame-level frequency domain signal, obtain a power spectrum value corresponding to the power spectrum of the previous frame signal after smoothing, and record it as the power spectrum value after smoothing of the previous frame signal;

[0056] S4. Calculating the posterior signal-to-noise ratio of the current frame signal based on the power spectrum value after smoothing the previous frame signal, and obtaining an initial value of the smoothing factor of the power spectrum corresponding to the current frame signal during the smoothing process based on the posterior signal-to-noise ratio of the current frame signal;

[0057] S5, calculating the frequency average ratio of the current frame signal according to the sum of the power spectrum values ​​of the previous frame signal after smoothing and the sum of the power spectrum values ​​corresponding to the current frame signal;

[0058] S6. updating an initial value of a smoothing factor of the current frame signal during the smoothing process based on a frequency average value ratio of the current frame signal to obtain a final value of the smoothing factor of the current frame signal during the smoothing process;

[0059] S7, smoothing the power spectrum corresponding to the current frame signal based on the final value of the smoothing factor of the current frame signal during the smoothing process to obtain a smoothed power spectrum value of the current frame signal;

[0060] S8. Obtaining the minimum value of the power spectrum of the previous L frames of signal after smoothing based on the minimum statistics method, and using the obtained minimum value as the noise estimation result of the current frame signal; where L is the length of the sliding window;

[0061] S9. Calculate the normalized variance of the second-order statistics of the power spectrum value after smoothing the current frame signal, and combine it with the length of the sliding window to generate a deviation compensation factor for the noise estimation result of the current frame signal, thereby obtaining the noise estimation result after compensation of the current frame signal and completing the fish farming cage noise tracking.

[0062] The specific method of converting the sound signal of the fish breeding area into the frequency domain after framing in step S1 includes:

[0063] The sound signal from the fish farming area is segmented into 20ms frames with a 10ms frame shift. Each frame is Hamming-windowed and then subjected to a Fast Fourier Transform (FFT) to generate a frame-level frequency domain signal. This step divides the continuous input signal into segments, and spectral analysis is performed on each segment to determine the energy distribution at each frequency and identify the frequency components.

[0064] The purpose of step S2 is to calculate the energy of each frequency of each frame signal and obtain the sound energy of each frequency point, so as to prepare for the power spectrum smoothing and minimum value tracking in the subsequent steps. In this embodiment, the power spectrum corresponding to each frame signal is directly calculated by taking the square of the Fourier transform coefficient:

[0065]

[0066] The results showed that Frame No. The probability density function of the sound energy at each frequency point is:

[0067]

[0068] in , Represents the power spectral density of the signal emitted by the potential sound source (such as the sound signal of a school of fish); represents the noise power spectral density; Indicates the value of the power spectrum; is a step function.

[0069] In step S4, the expression for calculating the posterior signal-to-noise ratio of the current frame signal based on the smoothed power spectrum value of the previous frame signal is:

[0070]

[0071] in Indicates the Frame signal The posterior signal-to-noise ratio of the frequency point, The larger it is, the stronger the target signal is; the smaller it is, the signal is dominated by noise; Indicates the Frame signal The power spectrum value after smoothing of each frequency point (including target signal and background noise); Indicates the Frame signal The noise power spectrum at each frequency point.

[0072] In step S4, the expression for obtaining the initial value of the smoothing factor of the power spectrum corresponding to the current frame signal during the smoothing process according to the posterior signal-to-noise ratio of the current frame signal is:

[0073]

[0074] in Indicates the Frame signal The initial value of the smoothing factor of the power spectrum corresponding to the frequency point in the smoothing process. When the target signal dominates, Approaching 1, a more stable effect is achieved; when When the background noise is dominant, The above processing provides the basis for estimating the minimum value of the noise power spectrum density in the subsequent steps.

[0075] In step S5, the expression for calculating the frequency average ratio of the current frame signal based on the sum of the power spectrum values ​​after smoothing of the previous frame signal and the sum of the power spectrum values ​​corresponding to the current frame signal is:

[0076]

[0077] in Indicates the Frequency average ratio of frame signal; Indicates the The sum of the power spectrum values ​​of the frame signal after smoothing; Indicates the The sum of the power spectrum values ​​corresponding to the frame signal. If the difference between the value and the actual power value is too large, it means that there is a problem with the current smoothing factor;

[0078] In step S6, the expression for updating the initial value of the smoothing factor of the current frame signal during the smoothing process based on the frequency average value ratio of the current frame signal is:

[0079]

[0080]

[0081] in Indicates the Frame signal The final value of the smoothing factor of each frequency point during the smoothing process; For the Correction factor of frame signal; is the weight parameter, usually 0.7-0.9; For the Correction factor of the frame signal. Even though the tracking is within the normal range, corrections should still be made to prevent mistracking. This step dynamically fine-tunes the smoothing factor when errors occur. The smoothing factor changes dynamically, taking a smaller value to better track the target signal when the target signal is more pronounced, and a value close to 1 when the target signal is less pronounced to achieve a smoother tracking effect and prevent stuttering or severe offsets.

[0082] In step S7, the expression for smoothing the power spectrum corresponding to the current frame signal based on the final value of the smoothing factor of the current frame signal in the smoothing process is:

[0083]

[0084] in Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Indicates the Frame signal The final value of the smoothing factor of each frequency point during the smoothing process; Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Shidi Frame signal The power spectrum value corresponding to the frequency point.

[0085] Smoothing the power spectrum corresponding to the current frame signal can prevent drastic fluctuations between frames. Since biological sounds and background noise do not jump randomly, smoothing helps to stabilize the observation of signal energy trends and identify changing trends. For the first frame signal, the power spectrum value of the first frame signal itself is directly used as its smoothed power spectrum value, that is, As the smoothing factor changes dynamically in subsequent frames, the initial value will be recursively updated step by step. Even if there are defects in the first frame, the subsequent frames will be gradually adjusted.

[0086] Since the target signal is not continuous, the background noise will be fully visible in some frames. Using this feature, the expression for obtaining the minimum value of the power spectrum value after smoothing the previous L frames of the signal based on the minimum statistics method in step S8 is:

[0087]

[0088] in Indicates the Frame signal Noise estimation results of frequency points; Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Indicates taking the minimum value.

[0089] Since the minimum value is often smaller than the true value, an upward correction is required to obtain a more realistic noise estimate to avoid underestimation of the noise, which may lead to misjudgment in subsequent processing and thus weaken the target signal. Therefore, the expression for calculating the normalized variance of the second-order statistics of the power spectrum value after smoothing the current frame signal in step S9 is:

[0090]

[0091]

[0092]

[0093] in Indicates the Frame signal The normalized variance of the second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The first-order statistics of the power spectrum value after smoothing the frequency points; is the weight coefficient; Indicates the Frame signal The second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The square of the power spectrum value after smoothing of the frequency points; Indicates the Frame signal The first-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The power spectrum value after smoothing the frequency points. The normalized variance measures the volatility of the smoothed signal and is the core indicator for determining the deviation compensation factor.

[0094] In step S9, the deviation compensation factor of the noise estimation result of the current frame signal is generated, and the expression of the noise estimation result after compensation of the current frame signal is obtained as follows:

[0095]

[0096]

[0097] in Indicates the Frame signal Noise estimation results after frequency compensation; Indicates the Frame signal Deviation compensation factor of the noise estimation result of each frequency point; It represents the logarithm with the natural constant e as the base; Represents the Digamma function.

[0098] In the specific implementation process, in order to improve the response speed to non-stationary noise, the block acceleration minimum value (tree structure) update strategy is used:

[0099] 1. Divide the sliding window length L into Q sub-windows;

[0100] 2. Calculate the local minimum value for each sub-window;

[0101] 3. The global minimum is synthesized from these local minima;

[0102] 4. Introduce a local minimum detection mechanism: If the minimum value in a sub-window does not appear in the first and last frames, it is likely to reflect stationary noise rather than the target signal gap.

[0103] The tree structure maintains the minimum values ​​of multiple windows and only updates them in sub-windows, thus ensuring low complexity and fast update.

[0104] In one embodiment of the present invention, many traditional noise estimation methods perform poorly when faced with dynamically changing non-stationary noise (such as ship noise and wind and wave noise in a cage environment). In particular, when the noise changes suddenly, problems such as over-suppression or residual noise often occur. This method introduces optimal smoothing and adaptive time-frequency smoothing parameters, enabling the noise power spectrum estimation to automatically adjust to the actual situation. Even with large noise fluctuations, the estimation results can be smoothly transitioned. Furthermore, an accelerated minimum tracking process is implemented, improving adaptability to non-stationary noise.

[0105] In the application of machine learning algorithms and non-negative matrix decomposition algorithms with strong noise suppression capabilities, if the model is to be highly accurate, it is crucial to control the amount of training to prevent overfitting, underfitting, and the like. Therefore, the time cost of obtaining a large amount of training data and model training has become an issue that cannot be ignored. The present invention directly analyzes the input signal through methods such as minimum statistics to achieve noise tracking. There is no need to prepare supervised training data, and it can achieve unsupervised tracking of noise throughout the process. It can be combined with various gain-type denoising methods to solve the problem of difficulty in preparing sufficient training data under realistic conditions. It is more efficient in processing long-term recorded cage monitoring data, and its accuracy is not inferior to the performance of the above-mentioned method under ideal training conditions. In addition, the present invention only needs to input time domain data to process it by itself, and the original time domain data obtained by monitoring can be directly analyzed. The data preparation stage is simple and fast, and it can achieve near real-time monitoring and rapid tracking of cages.

[0106] Many noise estimation or tracking methods rely on manually set thresholds (such as the signal-to-noise ratio threshold in noise activity detectors). These thresholds need to be adjusted based on the noise environment. Improper adjustment can lead to misclassification of the target signal as noise, resulting in over-suppression. This method, using the minimum statistic approach, completely avoids threshold setting. Instead, it dynamically estimates noise by tracking the minimum value of the signal, eliminating the need for manual intervention and resulting in greater adaptability.

[0107] Existing traditional methods work well under specific noise conditions, but fail when exposed to multiple noise types or require parameter readjustment. This method, which does not rely on a fixed noise model but dynamically adapts to different noise environments through minimum value evaluation and smoothing adjustments, can operate stably in environments with multiple noise types.

[0108] Many noise estimation methods (such as deep learning methods and gain function methods) require extensive computational resources and require training models or multiple complex modules (such as noise activity detectors and gain calculations). This high computational complexity can affect performance. This method uses recursive smoothing and local minimum estimation, and tracks the minimum using a simple sliding window. This method has low computational complexity and is suitable for real-time system implementation.

[0109] In summary, the present invention calculates the minimum value by smoothing the power spectrum, dynamically adjusts the smoothing coefficient, and performs deviation compensation in a timely manner to accurately track background noise, thereby achieving targeted noise extraction and solving the problem of difficulty in separating background noise in non-stationary environmental noise. At the same time, its tree-like update structure also provides a more efficient and accurate noise component tracking solution for cage aquaculture, which requires high timeliness of data analysis.

Claims

1. A fish farming cage noise tracking method based on optimal smoothing and minimum statistics, characterized in that: The following steps are involved: S1. Acquire the sound signal of the fish breeding area, divide it into frames, and convert it into the frequency domain to obtain a frame-level frequency domain signal; S2. Calculate the energy of each frequency of each frame signal in the frame-level frequency domain signal to obtain the acoustic energy of each frequency point, and then obtain the power spectrum corresponding to each frame signal in the frame-level frequency domain signal; S3. In the same frame-level frequency domain signal, obtain a power spectrum value corresponding to the power spectrum of the previous frame signal after smoothing, and record it as the power spectrum value after smoothing of the previous frame signal; S4. Calculating the posterior signal-to-noise ratio of the current frame signal based on the power spectrum value after smoothing the previous frame signal, and obtaining an initial value of the smoothing factor of the power spectrum corresponding to the current frame signal during the smoothing process based on the posterior signal-to-noise ratio of the current frame signal; S5, calculating the frequency average ratio of the current frame signal according to the sum of the power spectrum values ​​of the previous frame signal after smoothing and the sum of the power spectrum values ​​corresponding to the current frame signal; S6. updating an initial value of a smoothing factor of the current frame signal during the smoothing process based on a frequency average value ratio of the current frame signal to obtain a final value of the smoothing factor of the current frame signal during the smoothing process; S7, smoothing the power spectrum corresponding to the current frame signal based on the final value of the smoothing factor of the current frame signal during the smoothing process to obtain a smoothed power spectrum value of the current frame signal; S8. Obtaining the minimum value of the power spectrum of the previous L frames of signal after smoothing based on the minimum statistics method, and using the obtained minimum value as the noise estimation result of the current frame signal; where L is the length of the sliding window; S9. Calculate the normalized variance of the second-order statistics of the power spectrum value after smoothing the current frame signal, and combine it with the length of the sliding window to generate a deviation compensation factor for the noise estimation result of the current frame signal, thereby obtaining the noise estimation result after compensation of the current frame signal and completing the fish farming cage noise tracking.

2. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 1 is characterized in that: The specific method of converting the sound signal of the fish breeding area into the frequency domain after framing in step S1 includes: The sound signal from the fish farming area is segmented into frames with a frame length of 20ms and a frame shift of 10ms. After adding a Hamming window to each frame, a fast Fourier transform is performed to obtain a frame-level frequency domain signal.

3. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 1, characterized in that: In step S4, the expression for calculating the posterior signal-to-noise ratio of the current frame signal based on the power spectrum value after smoothing the previous frame signal is: ； in Indicates the Frame signal The posterior signal-to-noise ratio of each frequency point; Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Indicates the Frame signal The noise power spectrum at each frequency point.

4. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 3 is characterized in that: In step S4, the expression for obtaining the initial value of the smoothing factor of the power spectrum corresponding to the current frame signal during the smoothing process according to the posterior signal-to-noise ratio of the current frame signal is: ； in Indicates the Frame signal The initial value of the smoothing factor of the power spectrum corresponding to the frequency point during the smoothing process.

5. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 4, characterized in that: In step S5, the expression for calculating the frequency average ratio of the current frame signal based on the sum of the power spectrum values ​​after smoothing of the previous frame signal and the sum of the power spectrum values ​​corresponding to the current frame signal is: ； in Indicates the Frequency average ratio of frame signal; Indicates the The sum of the power spectrum values ​​of the frame signal after smoothing; Indicates the The sum of the power spectrum values ​​corresponding to the frame signal.

6. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 5, characterized in that: In step S6, the expression for updating the initial value of the smoothing factor of the current frame signal during the smoothing process based on the frequency average value ratio of the current frame signal is: ； ； in Indicates the Frame signal The final value of the smoothing factor of each frequency point during the smoothing process; For the Correction factor of frame signal; is the weight parameter; For the Correction factor for the frame signal.

7. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 1, characterized in that: In step S7, the expression for smoothing the power spectrum corresponding to the current frame signal based on the final value of the smoothing factor of the current frame signal in the smoothing process is: ； in Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Indicates the Frame signal The final value of the smoothing factor of each frequency point during the smoothing process; Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Shidi Frame signal The power spectrum value corresponding to the frequency point.

8. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 1, characterized in that: In step S8, the expression for obtaining the minimum value of the power spectrum value after smoothing the previous L frames of signal based on the minimum statistics method is: ； in Indicates the Frame signal Noise estimation results of frequency points; Indicates the Frame signal The power spectrum value after smoothing of the frequency points; Indicates taking the minimum value.

9. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 8, characterized in that: The expression for calculating the normalized variance of the second-order statistics of the power spectrum value after smoothing the current frame signal in step S9 is: ； ； ； in Indicates the Frame signal The normalized variance of the second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The first-order statistics of the power spectrum value after smoothing the frequency points; is the weight coefficient; Indicates the Frame signal The second-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The square of the power spectrum value after smoothing of the frequency points; Indicates the Frame signal The first-order statistics of the power spectrum value after smoothing the frequency points; Indicates the Frame signal The power spectrum value after smoothing of the frequency points.

10. The fish farming cage noise tracking method based on optimal smoothing and minimum statistics according to claim 9, characterized in that: In step S9, the deviation compensation factor of the noise estimation result of the current frame signal is generated, and the expression of the noise estimation result after compensation of the current frame signal is obtained as follows: ； ； in Indicates the Frame signal Noise estimation results after frequency compensation; Indicates the Frame signal Deviation compensation factor of the noise estimation result of each frequency point; It represents the logarithm with the natural constant e as the base; Represents the Digamma function.

Images