Underwater organism voiceprint recognition and noise pollution monitoring integrated system

Abstract

The invention relates to the technical field of ship operation and marine ecology monitoring, in particular to an underwater biological voiceprint recognition and noise pollution monitoring integrated system, which comprises an underwater acoustic collaborative sensing array, a voiceprint recognition module, a voiceprint recognition module, a voiceprint recognition module, a voiceprint recognition module, a voiceprint recognition module, a voiceprint recognition module, a voiceprint recognition module, a voiceprint recognition module, a voiceprint recognition module, a voiceprint recognition module and a voiceprint recognition module, a ship working condition feature decoupling module; the noise pollution self-adaptive analysis engine is used for dynamically configuring band elimination filter parameters according to the anchoring state and the towing state, and generating a noise pollution level by calculating the noise energy proportion in a target biological frequency band; a biological voiceprint anti-interference recognition module; and a pollution biological effect evaluation module. The system can realize high-precision ship operation state recognition, noise pollution self-adaptive monitoring and biological voiceprint anti-interference recognition, provides a detailed pollution responsibility report at the same time, and greatly improves the accuracy and practicability of underwater noise pollution monitoring and biological acoustic behavior evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of ship operation and marine ecological monitoring, and in particular to an integrated system for underwater biological soundprint recognition and noise pollution monitoring. Background Art

[0002] In the field of marine environmental monitoring, the impact of noise pollution generated by ship operations such as anchoring and towing on the acoustic behavior of aquatic organisms has received increasing attention. Existing technologies have the following main limitations:

[0003] Currently deployed ship noise monitoring systems (such as the ISO / PAS17208-1 standard system) mainly record broadband sound pressure level data and are unable to separate biological soundprints from environmental noise components. Independently operated bioacoustic monitoring buoys, while able to collect biological sound signals, are completely disconnected from real-time ship operating conditions (such as windlass start and stop status and towing speed changes). This separation leads to two major drawbacks:

[0004] 1. It is impossible to determine whether the sudden change in noise spectrum is caused by the ship's operating state switching (such as the instantaneous locking of the anchor chain) or environmental interference;

[0005] 2. The biometric voiceprint database lacks background noise samples of standard ship operating conditions, which makes the recognition algorithm ineffective in real ship scenarios.

[0006] In addition, traditional noise suppression solutions use fixed-band band rejection filters (such as notch filters for propeller blade frequency), which cannot adapt to the dynamic spectrum migration caused by transient ship operating conditions. Typical problems include:

[0007] At anchoring: The main frequency of the impact noise generated by the collision between the anchor chain and the seabed shifts to a lower frequency as the length of the anchor chain increases (the migration amount can reach 30%-50% of the fundamental frequency), and the fixed stopband center frequency causes the leakage of key noise components;

[0008] Towing phase: The propeller blade frequency drifts continuously with the ship speed (the drift rate is positively correlated with the rate of change of ship speed), and the static filter bandwidth cannot cover the actual frequency shift range.

[0009] Secondly, existing methods only determine the degree of noise pollution by using the sound pressure level threshold, ignoring the specific interference of noise on biological acoustic behavior:

[0010] It is impossible to quantify the interruption of acoustic communication of aquatic organisms caused by ship pulse noise (such as anchor chain impact) (manifested as abnormal coefficient of variation of acoustic pulse interval), and lacks the ability to analyze the escape trajectory of organisms near the noise source in real time.

[0011] Therefore, there is an urgent need for an integrated system of underwater biological voiceprint recognition and noise pollution monitoring to solve the above problems. Summary of the Invention

[0012] Based on the above objectives, the present invention provides an integrated system for underwater bio-voiceprint recognition and noise pollution monitoring, which is applied to ships or marine platforms, including:

[0013] Underwater acoustic cooperative sensing array, including low-frequency pressure sensors, medium- and high-frequency hydrophones, and vibration accelerometers at the anchor chain guide holes, deployed at the hull waterline;

[0014] The ship operating condition feature decoupling module identifies the anchoring state based on the time domain envelope characteristics of the anchor chain vibration acceleration signal collected by the vibration accelerometer at the anchor chain guide hole, and identifies the towing state based on the coupling relationship between the propeller noise line spectrum components collected by the medium and high frequency hydrophones and the ship speed data;

[0015] Noise pollution adaptive analysis engine, which dynamically configures the band-stop filter parameters according to the mooring status and towing status, and generates the noise pollution level by calculating the noise energy proportion within the target biological frequency band;

[0016] The bio-voiceprint anti-interference recognition module decouples the ship operating condition feature from the ship operating condition noise feature, injects the bio-voiceprint training set into the dual-stream recognition model, and ultimately outputs the bio-voiceprint and bio-acoustic behavior distortion of the target organism.

[0017] The pollution biological effect assessment module takes noise pollution level, biological soundprint and bioacoustic behavior distortion as input, and generates a pollution responsibility report by analyzing the spatiotemporal correlation between bioacoustic behavior distortion and noise events.

[0018] Preferably, the layout of the underwater acoustic cooperative sensing array satisfies:

[0019] The spacing between any adjacent low-frequency pressure sensors and medium- and high-frequency hydrophones is evenly distributed according to the circumference of the hull waterline, and the number is determined by the ratio of the hull length to the reference spacing;

[0020] The low-frequency pressure sensor and the mid- and high-frequency hydrophones are integrated into an acoustic shield. The shield's waveguide structure is designed so that the incident sound wave enters the low-frequency pressure sensor after attenuating the hull's self-noise through the Helmholtz resonator. The attenuation frequency band covers the main frequency of the ship's mechanical vibration.

[0021] The installation direction of the vibration accelerometer is determined according to the stress analysis of the anchor chain guide hole, so that the angle between its sensitive axis and the normal vector of the anchor chain plane is less than the threshold angle.

[0022] Preferably, the process of identifying the anchoring state by the ship operating condition characteristic decoupling module includes:

[0023] Envelope extraction is performed on the anchor chain vibration acceleration signal, and an adaptive threshold is used to detect impact events.

[0024] Calculate the statistical distribution of the impact event intervals. When the coefficient of variation of the distribution peak is lower than the set threshold and the peak position and the anchor chain length meet the positive correlation model, it is determined to be a periodic impact.

[0025] The positive correlation model is calibrated through ship mooring physical experiments: the impact interval is measured at various anchor chain lengths, and a linear fitting equation of length-interval is established.

[0026] Preferably, the adaptive threshold detection specifically adopts a dynamic time window mechanism:

[0027] The initial width of the time window is set according to the rated impact duration of the anchor winch corresponding to the ship's tonnage;

[0028] The real-time adjustment coefficient of the window width is the ratio of the current anchor chain length to the reference length;

[0029] The local maximum of the signal envelope is calculated within the window, and when the maximum amplitude exceeds a multiple of the average amplitude of the previous N windows, it is marked as a valid impact, and the multiple is dynamically adjusted by the historical noise level.

[0030] Preferably, the dynamically configuring band-stop filter parameters includes:

[0031] Generation of stopband center frequency in anchored state: Extract the fundamental frequency and harmonic components of the anchor chain vibration signal, and select the component with energy proportion exceeding the total energy proportion as the stopband center;

[0032] Determination of stopband bandwidth in towing state: According to the rate of change of propeller blade frequency with ship speed, the stopband bandwidth expansion coefficient is calculated. This coefficient is positively correlated with the rate of change of ship speed.

[0033] Adjustment of stopband attenuation depth: Based on the percentile value of the current noise pollution level in historical data, the attenuation depth value is output according to the preset mapping relationship.

[0034] Preferably, the process of constructing the dual-stream recognition model includes:

[0035] Noise feature injection: Collect the original noise signal under standard working conditions, extract the noise fragments that overlap with the target biological voiceprint frequency band through bandpass filtering, and mix them with the pure biological voiceprint according to the preset signal-to-noise ratio gradient;

[0036] Two-stream network training: The first-stream network inputs the time-frequency graph of the mixed voiceprint, and the second-stream network inputs the noise type label and pollution level vector;

[0037] The feature fusion layer performs noise-aware spectral weighting: it generates a frequency domain mask based on the noise pollution level of the second stream output and performs channel-wise weighting on the convolutional feature maps of the first stream.

[0038] Preferably, the generating of the frequency domain mask specifically adopts a noise sensitivity mapping mechanism:

[0039] Establish a mapping table between noise type and biological hearing sensitivity, where the sensitivity value is generated by the inverse of the species hearing threshold curve, where the hearing threshold is the minimum sound pressure level that a specific species can perceive;

[0040] Multiply the noise pollution level by the sensitivity value to generate the mask weight coefficient for each frequency band;

[0041] The mask weight is applied to the feature map output by the first stream network according to the frequency channel.

[0042] Preferably, the analysis of bioacoustic behavior distortion comprises:

[0043] Calculation of acoustic communication interference index: Calculation of acoustic communication interference index: Extract the identified biological sound pulse sequence and calculate the real-time standard deviation σ of the interval between adjacent pulses 实时 , by querying the aquatic organism acoustic behavior database to obtain the species typical interval standard deviation σ 典型 , calculate the acoustic communication interference index value according to the following formula:

[0044] Acoustic communication interference index = σ 典型 / σ 实时 ;

[0045] Determination of the intensity of escape behavior: Calculate the biological motion vector through the sound source localization trajectory when the following conditions are met:

[0046] The angle between the vector direction and the noise source direction is continuously greater than 90°;

[0047] If the movement acceleration exceeds the species normal acceleration threshold, the escape behavior intensity value is output. The escape behavior intensity value is the ratio of the measured acceleration to the species normal acceleration threshold.

[0048] The acoustic communication interference index value and the escape behavior intensity value together constitute the bioacoustic behavior distortion value. Preferably, the generation of the pollution responsibility report includes:

[0049] Construct a spatiotemporal correlation matrix between noise events and biological responses. The matrix includes noise event timestamps, sound pressure levels, biological response delay times, acoustic communication interference index values, and escape behavior intensity values.

[0050] Verification of the validity of the biological response delay time: The delay time must be within the closed interval formed by the sound wave propagation delay and the maximum physiological response delay of the biological species. The maximum physiological response delay is determined by the multiple of the heart rate cycle of the biological species.

[0051] Pollution responsibility index calculation: The acoustic communication interference index and escape behavior intensity value are weighted and integrated according to the noise sensitivity of the species to obtain the pollution responsibility index. When the index exceeds the preset threshold, a responsibility report is generated.

[0052] Preferably, a closed-loop feedback control interface is also included, and its execution logic includes:

[0053] Anchoring condition optimization: The anchor chain release length adjustment is calculated based on the deviation rate of the anchor chain vibration main frequency relative to the ambient background frequency. The adjustment amount is linearly related to the frequency deviation rate.

[0054] Towing condition optimization: Based on the distribution of the target organism's auditory sensitive frequency band, the optimal speed solution set that causes the propeller blade frequency to deviate from the center of the sensitive frequency band is solved. This solution set is generated through the constraint relationship between the center frequency of the sensitive frequency band and the propeller speed.

[0055] Beneficial effects of the present invention:

[0056] 1. This invention integrates an underwater acoustic collaborative sensing array and a ship operating condition decoupling module to acquire real-time ship operating condition information and accurately decouple the noise spectrum based on this information. By analyzing the envelope characteristics of the anchor chain vibration acceleration signal and the propeller noise of the hydrophone, the system can effectively identify the source of the noise, whether it is caused by ship state switching or environmental noise interference, thus solving the problem of being unable to determine the source of sudden changes in the noise spectrum.

[0057] 2. This invention utilizes a ship operating condition feature decoupling module, combined with a bio-voiceprint anti-interference recognition module, to inject characteristic noise signals from ship operating conditions into a bio-voiceprint training set. This creates a dual-stream recognition model that accurately simulates and identifies potential noise interference encountered during actual ship operations. This system is able to accurately perform bio-voiceprint recognition under realistic ship operating conditions, significantly improving its accuracy and effectiveness in practical applications.

[0058] 3. This invention dynamically configures the band-stop filter parameters. When moored, the stopband center frequency is dynamically generated based on the fundamental frequency and harmonic components of the anchor chain vibration signal. When towing, the filter bandwidth is adjusted to account for actual spectrum drift based on the rate of change of the propeller frequency. This dynamic filtering strategy, adjusted in real time based on operating conditions, effectively avoids the drawback of fixed-band filters that cannot adapt to spectrum shifts, thereby improving the flexibility and accuracy of noise suppression.

[0059] 4. This invention incorporates a bio-voiceprint anti-interference recognition module to assess interference with aquatic organism acoustic communications. This includes analysis of acoustic pulse sequences, calculation of the coefficient of variation of pulse intervals, and real-time assessment of escape behavior. Through this detailed analysis of biological behavior, the system can quantify the impact of ship impulse noise on aquatic organisms, accurately determining the degree of interference with biological acoustic communications and the resulting escape responses, thereby providing a more comprehensive assessment of the impact of noise pollution.

[0060] 5. The present invention utilizes a pollution biological effects assessment module, sound source localization, and biological motion vector analysis to determine the escape behavior of aquatic organisms in real time. When the angle between the organism's motion trajectory and the noise source remains obtuse and the acceleration exceeds the species' normal acceleration threshold, the system flags the behavior as escape. This real-time monitoring capability helps assess the dynamic impact of noise on aquatic organisms and enhances the system's responsiveness to noise pollution sources.

[0061] 6. This invention constructs a spatiotemporal correlation matrix between noise events and biological responses, combined with a pollution responsibility index calculation. This method comprehensively considers the noise event's timestamp and sound pressure level, as well as the biological response delay, communication interference indicators, and the intensity of evasion behavior, thereby accurately assessing pollution responsibility. When the pollution responsibility index exceeds a set threshold, the system automatically generates a pollution responsibility report, identifying the source of pollution responsibility. This more accurate pollution responsibility assessment method provides strong data support for environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0063] Figure 1 It is a structural block diagram of the system of the present invention;

[0064] Figure 2 A flowchart of the steps of the process of constructing a dual-stream recognition model in the system of the present invention;

[0065] Figure 3 A flow chart of the steps for generating a pollution responsibility report in the system of the present invention. DETAILED DESCRIPTION

[0066] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0067] See Figure 1-Figure 3, an embodiment of the present invention provides an integrated system for underwater biological soundprint recognition and noise pollution monitoring. The underwater acoustic cooperative sensing array in this system is composed of a low-frequency pressure sensor, a medium- and high-frequency hydrophone, and a vibration accelerometer at the anchor chain guide hole, which are arranged at the waterline of the hull. The low-frequency pressure sensor is used to monitor the low-frequency noise of the ship and its operation process, and can capture the basic operating noise of the ship, such as the low-frequency vibration generated by the propeller and the hull; the medium- and high-frequency hydrophone can be used to capture high-frequency noise components, especially the separation of aquatic organisms and environmental noise, providing key acoustic data for biological soundprint recognition; and the vibration accelerometer is located at the anchor chain guide hole, specifically monitoring the anchor chain vibration signal, which is used to identify the ship's anchoring status. Through the collaborative work of these sensors, the system can accurately capture ship and biological sound signals in different frequency bands, providing accurate data support for subsequent noise monitoring and biological soundprint recognition.

[0068] This array enables comprehensive underwater acoustic sensing, capturing not only the ship's mechanical noise but also biological signals, significantly improving the accuracy of noise source location and bioacoustic behavior monitoring. In particular, the detection of anchor chain vibration by the vibration accelerometer accurately determines the ship's mooring status, providing precise operating condition information for subsequent noise identification and interference assessment.

[0069] The ship operating condition characteristic decoupling module uses time-domain analysis and spectrum analysis algorithms based on the time-domain envelope characteristics of the anchor chain vibration signal collected by the vibration accelerometer to identify the ship's anchoring status. This is achieved by real-time monitoring of the time-domain envelope signal of the impact noise generated by the anchor chain colliding with the seabed, capturing specific impact frequency characteristics and determining whether the ship is at anchor. The module also identifies the ship's towing status by analyzing the coupling relationship between the spectral components of the propeller noise and the ship's speed data. This process involves multiple signal analysis techniques, leveraging the dynamic relationship between the propeller noise spectrum captured by the hydrophone and the speed data to accurately determine the status of the towing operation.

[0070] The implementation of the ship operating condition decoupling module enables the system to switch noise analysis strategies in real time based on the vessel's operating conditions, accurately identifying whether the vessel is at anchor or under tow. This operating condition identification capability provides critical information for subsequent noise pollution analysis and bioacoustic behavior assessment, effectively resolving the existing problem of being unable to obtain real-time vessel operating conditions and decouple noise sources.

[0071] The noise pollution adaptive analysis engine dynamically adjusts the parameters of the band-stop filter according to the ship's operating conditions (anchored or towing). Specifically, during the anchoring phase, the center frequency of the filter is dynamically configured according to the spectral characteristics of the anchor chain vibration acceleration signal to effectively filter out the low-frequency noise generated by the anchor chain collision; while in the towing phase, the system adjusts the bandwidth of the filter based on the spectral variation of the propeller noise to ensure that it can cover the dynamic changes in the propeller frequency. Through this adaptive adjustment, the system can filter out noise components that are not related to the target biological frequency band in real time, calculate the noise energy ratio within the target biological frequency band, and generate the noise pollution level.

[0072] This engine effectively addresses spectrum drift caused by changes in ship status by dynamically adjusting filter parameters, ensuring accurate noise pollution assessment. Compared to traditional fixed-band filtering solutions, this engine is more adaptable to spectrum changes in transient ship operating conditions, thereby improving the accuracy and flexibility of noise pollution monitoring.

[0073] The bio-voiceprint anti-interference recognition module combines the ship operating noise characteristics identified by the ship operating condition feature decoupling module and injects these noise characteristics into the bio-voiceprint training set to construct a dual-stream recognition model. In this way, the system can simultaneously identify the target organism's bio-voiceprint and the distortion of its bioacoustic behavior. Specifically, it can identify whether the target organism is experiencing noise interference in noise-polluted environments. The dual-stream model ensures accurate bio-voiceprint recognition while improving anti-interference capabilities in noisy environments.

[0074] In this embodiment, analyzing the bioacoustic behavior distortion includes:

[0075] Calculation of acoustic communication interference index: Calculation of acoustic communication interference index: Extract the identified biological sound pulse sequence and calculate the real-time standard deviation σ of the interval between adjacent pulses 实时 , by querying the aquatic organism acoustic behavior database to obtain the species typical interval standard deviation σ 典型 , calculate the acoustic communication interference index value according to the following formula:

[0076] Acoustic communication interference index = σ 典型 / σ 实时 ;

[0077] Determination of the intensity of escape behavior: Calculate the biological motion vector through the sound source localization trajectory when the following conditions are met:

[0078] The angle between the vector direction and the noise source direction is continuously greater than 90°;

[0079] If the movement acceleration exceeds the species normal acceleration threshold, the escape behavior intensity value is output. The escape behavior intensity value is the ratio of the measured acceleration to the species normal acceleration threshold.

[0080] The acoustic communication interference index value and the escape behavior intensity value together constitute the bioacoustic behavior distortion value.

[0081] The pollution bioeffects assessment module receives noise pollution levels, bio-voice patterns, and bioacoustic behavior distortion as input. It uses spatiotemporal correlation analysis to analyze the relationship between bioacoustic behavior and noise events, generating a pollution responsibility report. By analyzing bioacoustic behavior variations and monitoring evasion behavior in real time, the system determines the pollution responsibility of noise sources and further refines the assessment of noise pollution impacts.

[0082] This module provides a comprehensive pollution responsibility assessment, comprehensively considering the impact of noise pollution on biological behavior, providing a reliable basis for environmental protection and management. By analyzing organisms' escape trajectories and communication interference in real time, this module can play a key role in underwater noise pollution monitoring and accurately assess the responsibility of pollution sources.

[0083] By integrating multiple advanced technologies, the system addresses existing technical challenges in ship noise pollution monitoring, including distinguishing between ship noise status and ambient noise, combining operating noise with bio-acoustic soundprint recognition, and dynamically monitoring and assessing noise pollution. The system enables highly accurate identification of ship operating status, adaptive noise pollution monitoring, and anti-interference bio-acoustic soundprint recognition, while also providing detailed pollution responsibility reports. This significantly improves the accuracy and practicality of underwater noise pollution monitoring and bio-acoustic behavior assessment.

[0084] In one possible implementation, to ensure the high efficiency of the underwater acoustic cooperative sensing array, the spacing between low-frequency pressure sensors and medium- and high-frequency hydrophones is evenly distributed based on the circumference of the ship's waterline. The specific number of sensors deployed is determined by the ratio of the hull length to the reference spacing. This ensures that each sensor covers a different area of ​​the water on the hull, reducing blind spots and improving the quality and spatial resolution of acoustic signal acquisition. This even distribution allows the sensors to more accurately capture mechanical noise, environmental noise, and bioacoustic signals from the ship, ensuring the system's acoustic perception.

[0085] This uniform layout strategy can effectively improve the system's all-round perception of ship noise, environmental noise and biological acoustic signals, thereby achieving high-precision voiceprint recognition and noise pollution monitoring in complex underwater environments, and avoiding local signal blind spots caused by uneven sensor layout.

[0086] The low-frequency pressure sensor and mid- and high-frequency hydrophones are integrated within a single acoustic shield. The shield utilizes a specific waveguide structure designed to effectively suppress interference from the ship's self-noise. Upon entering the shield, sound waves first pass through a Helmholtz resonator to attenuate the ship's self-noise, ensuring the sensor captures a clear signal from the outside world, particularly noise within the dominant mechanical vibration frequency range. The Helmholtz resonator design attenuates noise within a specific frequency band, prioritizing the elimination of the ship's inherent mechanical vibration noise.

[0087] The design of the shielding cover and Helmholtz resonance cavity significantly reduces the interference of the hull's self-noise on the sensor's collected signals, enabling the system to more accurately identify signals from organisms and the environment in a noise-polluted environment, thereby improving the accuracy of underwater voiceprint recognition.

[0088] The installation orientation of the vibration accelerometer is determined based on a force analysis of the anchor chain guide hole. First, through finite element analysis of the anchor chain forces, the relative angle between the anchor chain force direction and the ship's hull is calculated. To increase the vibration accelerometer's sensitivity to anchor chain vibration, the angle between its sensitive axis and the normal vector of the anchor chain plane must be less than a set threshold angle. The installation orientation of this sensitive axis is adjusted to maximize alignment with the anchor chain vibration direction, thereby accurately capturing the anchor chain vibration signal. This process is accomplished using a precise mechanical mounting or tilting device to ensure the sensor's precise orientation.

[0089] This design allows the vibration accelerometer to capture the maximum effective signal from the anchor chain vibration, improving the accuracy of monitoring the ship's mooring status. Compared to traditional installation methods that can be installed in any direction, this method significantly enhances the sensitivity and directionality of the anchor chain vibration, ensuring accurate determination of the ship's mooring status.

[0090] The angle between the sensitive axis of the vibration accelerometer and the normal vector of the anchor chain plane must be less than a threshold angle. The threshold angle is determined by analyzing the force on the anchor chain guide hole and the vibration characteristics of the hull. Normally, the best measurement effect will be obtained when the threshold angle is less than 30 degrees, especially less than 15 degrees. In this embodiment, the threshold angle is 14 degrees. By adjusting the angle of the accelerometer installation, it can be ensured that the sensitive axis and the anchor chain vibration direction are as consistent as possible, which can effectively reduce the signal loss and error caused by the deviation of the installation angle. The setting of the threshold angle is usually adjusted based on experimental data and the actual application effect of the vibration accelerometer.

[0091] By ensuring the sensitive axis is aligned with the anchor chain vibration direction, the vibration accelerometer's acquisition sensitivity can be maximized, providing an accurate anchor chain vibration signal. This feature is particularly important for identifying a ship's mooring status, helping to accurately assess the ship's operating condition and providing more reliable data support for noise pollution monitoring and biometric voiceprint recognition.

[0092] By optimizing sensor placement, shielding design, and the directionality of the vibration accelerometers, the system's accuracy in noise monitoring and bio-acoustic recognition has been improved in complex underwater environments. The even distribution of sensors and efficient acoustic shielding reduce interference from the ship's self-noise, while the optimized mounting angles of the vibration accelerometers enhance the accuracy of monitoring the ship's operating conditions. Overall, the system is capable of accurately monitoring noise pollution and analyzing its impact on organisms in dynamic water environments, providing strong technical support for underwater noise pollution prevention and control and environmental protection.

[0093] In one possible implementation, envelope extraction is first performed on the anchor chain vibration acceleration signal. Envelope extraction technology aims to extract the main vibration trends from complex vibration signals. Using a Hilbert transform or envelope detection algorithm, the signal is processed to remove high-frequency noise and unnecessary details, retaining only the low-frequency periodic variation characteristics. The envelope signal effectively reflects the amplitude and time domain characteristics of the impact event, facilitating subsequent analysis.

[0094] An adaptive threshold detection method is used to identify impact events in the extracted envelope signal. In practical applications, the amplitude of the impact signal may be affected by various external factors. Therefore, an adaptive method is needed to set the threshold so that the detection sensitivity can be adjusted according to the dynamic changes of the signal. The adaptive threshold method automatically determines an appropriate detection threshold based on the statistical characteristics of the signal, effectively identifying impact events in the anchor chain vibration signal.

[0095] This method can accurately identify impact events in anchor chain vibration, especially in complex water environments, reduce the possibility of false alarms and missed alarms, and ensure the accuracy of subsequent analysis.

[0096] Once the impact events are identified, the next step is to calculate the statistical distribution of the intervals between impact events. By calculating the time intervals between consecutive impact events, the temporal regularity of the events can be determined. If the interval distribution shows a clear periodic pattern, it indicates that the anchor chain may be experiencing periodic impact.

[0097] When calculating the intervals between impact events, the stability of the distribution is determined by calculating the coefficient of variation. The coefficient of variation is the ratio of the standard deviation to the mean and is used to measure the dispersion of the distribution. When the coefficient of variation of the interval between impact events is below a set threshold, the intervals between events can be considered relatively stable and exhibit periodic characteristics. At this point, if the peak position of the distribution conforms to a positive correlation model with the anchor chain length (i.e., the interval is proportional to the anchor chain length), it further confirms that the anchor chain is in a periodic impact state.

[0098] By combining the coefficient of variation with a positive correlation model, the system can accurately determine whether the anchor chain is in a periodic impact state. This is crucial for mooring monitoring, helping to avoid misjudgments and providing precise feedback on the ship's stability.

[0099] To establish a positive correlation model between anchor chain length and the interval between impact events, a series of ship mooring physical experiments were conducted. In these experiments, the relationship between anchor chain length and the interval between impact events was analyzed by measuring the interval between impact events in actual operations with anchor chains of different lengths.

[0100] The impact event interval data collected in the experiment was fitted with the anchor chain length to obtain a linear relationship between length and interval. The linear fitting equation can be expressed as:

[0101] T = aL + b, where T represents the impact event interval, L represents the anchor chain length, a is the fitting coefficient, and b is the constant term. This linear fitting equation can be used to predict the impact event interval for different anchor chain lengths, providing a basis for determining whether the anchor chain is in a periodic impact state.

[0102] By establishing a positive correlation model, we can accurately quantify the relationship between anchor chain length and the interval between impact events. This not only improves the accuracy of determining the periodicity of impact events, but also allows us to derive a universal anchor chain impact model from experimental data, providing a scientific basis for subsequent system applications.

[0103] By combining envelope extraction, adaptive threshold detection, statistical analysis, and a positive correlation model, the accuracy of anchor chain status identification is effectively improved. Envelope extraction removes high-frequency noise, making impact event detection clearer, while adaptive threshold detection dynamically adjusts sensitivity to avoid misjudgments. Furthermore, statistical analysis of impact event intervals and a positive correlation model provide a robust mathematical basis for identifying periodic impacts, enabling more accurate monitoring of a ship's anchoring status and providing data support for noise pollution monitoring.

[0104] In one possible implementation, the dynamic time window mechanism in this system first determines the initial time window width based on the vessel's tonnage. Specifically, the initial time window width considers the relationship between the vessel's tonnage and the rated impact duration of the anchor winch. Larger vessels typically have larger anchor chains and longer impact durations, resulting in a wider initial time window width, while smaller vessels may require a smaller time window width. This initial setting is derived based on experimental data and empirical formulas to ensure that it closely matches the vessel's actual operating conditions.

[0105] Over time, the window width adjusts in real time based on the ratio of the current anchor chain length to the reference length. Chain length varies with factors such as the vessel's position and tides, so dynamically adjusting the time window width based on the ratio of the current anchor chain length to the reference length better adapts to actual conditions. Longer anchor chains typically cause longer shock vibrations, increasing the window width, while shorter ones decrease. This dynamic adjustment mechanism enables the system to maintain consistently high detection accuracy across varying chain lengths.

[0106] By dynamically adjusting the time window width, the system can flexibly adapt to specific conditions, ensuring high detection sensitivity for different vessels and anchor chain lengths while also avoiding false or missed detections caused by overly long or short time windows. This mechanism improves the system's adaptability and accuracy under various operating conditions.

[0107] Within each time window, the system calculates the local maxima of the signal envelope. This signal envelope, derived through an envelope extraction algorithm, reflects the primary amplitude variations of the anchor chain vibration. By searching for local maxima within each time window, the system identifies shock peaks in the vibration signal. These local maxima typically represent significant shock signals generated by the interaction between the vessel and the underwater environment.

[0108] When the amplitude of a local maximum exceeds a multiple of the average amplitude over the previous N time windows, the system marks the maximum as a valid impact event. This multiple is dynamically adjusted based on the historical noise level to adapt to changes in the current noise environment. Specifically, when the noise level is high, the multiple is increased to mitigate the effects of noise interference; when the noise level is low, the multiple is decreased to increase sensitivity to minor impacts.

[0109] By calculating local maxima and dynamically adjusting the multiple, the system can more accurately identify valid impact events. This dynamic adjustment mechanism, based on historical noise levels, enables the system to operate stably and continuously in complex noise environments, unaffected by noise fluctuations, thereby effectively improving the accuracy of impact event detection. Furthermore, the use of local maxima and dynamic multiples effectively distinguishes true impact events from background noise, thus avoiding false alarms.

[0110] By introducing a dynamic time window mechanism and an impact marking method that dynamically adjusts based on historical noise levels, the integrated underwater bio-voiceprint recognition and noise pollution monitoring system has strong adaptability and accuracy in dynamic and complex environments. The mechanism of dynamically adjusting the time window width and marking effective impacts enables real-time optimization for different ships, different anchor chain lengths, and noise levels, making the system more accurate in monitoring the ship's anchoring status. Furthermore, the real-time parameter adjustment strategy enables the system to effectively cope with noise interference in a changing underwater environment, ensuring high accuracy of monitoring results and further improving the system's reliability and stability.

[0111] In one possible implementation, when anchored, a ship's anchor chain is affected by external factors, such as currents, wind, and the motion of the anchor winch, generating certain vibration signals. These signals contain a fundamental frequency and multiple harmonic components. The system extracts the fundamental frequency and harmonic components of the signal using Fourier transform or other frequency domain analysis methods.

[0112] After extracting the fundamental frequency and harmonic components, the system calculates the energy contribution of each frequency component. Typically, the fundamental frequency and its harmonic components of the anchor chain vibration signal are concentrated within a specific frequency band. The frequency component whose energy contribution exceeds the total energy contribution is selected as the center frequency of the stopband. This frequency component best represents the primary vibration characteristics of the ship at anchor and is therefore used as the center frequency of the band-stop filter for subsequent processing.

[0113] By extracting the fundamental frequency and harmonics of the anchor chain vibration signal, the characteristic frequencies of the ship at anchor can be accurately identified. The band-stop filter is precisely set near these frequencies, effectively suppressing noise interference caused by the anchor chain vibration and preventing it from interfering with underwater bio-voice recognition. Selecting the component with an energy proportion exceeding the total energy as the stopband center also reduces errors caused by environmental noise, further improving signal processing accuracy.

[0114] When towing, a ship's propeller generates blade-frequency noise of varying frequencies as speed changes. The system first calculates the propeller blade frequency at the current speed, taking into account the rate of change of propeller blade frequency with speed. Propeller blade frequency is positively correlated with speed, so as speed increases, so does blade frequency.

[0115] Based on the rate of change of the propeller blade frequency, the system generates a stopband bandwidth expansion factor. This factor is positively correlated with the rate of change of ship speed; that is, higher speeds increase the expansion factor, resulting in a wider stopband bandwidth. This factor calculation method can be established using experimental data and mathematical models to ensure that as ship speed increases, the filter can adapt to a wider range of frequency variations and accurately suppress multi-band noise.

[0116] This dynamic adjustment mechanism, based on speed changes, effectively addresses frequency variations during towing, allowing the system to adapt to varying noise frequency ranges at varying speeds. Accurately setting the stopband bandwidth optimizes the system's ability to suppress noise pollution during towing, ensuring the accuracy of underwater bio-voiceprint recognition and the effectiveness of noise pollution monitoring.

[0117] The system monitors underwater noise pollution levels in real time and outputs an attenuation depth value based on the current noise pollution level's percentile relative to historical data, using a pre-set mapping relationship. The noise pollution level is typically determined by comparing the total energy of the ambient noise with the background noise level. For example, when the noise pollution level is high, the attenuation depth should be increased to suppress excessive noise signals; when the noise pollution level is low, the attenuation depth can be appropriately reduced to avoid excessive suppression of valid signals.

[0118] By dynamically adjusting the attenuation depth, the band-stop filter maintains optimal performance under varying noise pollution levels. When noise pollution in the underwater environment is severe, increasing the attenuation depth effectively eliminates more noise components, ensuring that underwater bio-voice recognition is not affected by excessive noise. This adjustment mechanism also avoids unnecessary signal distortion, ensuring effective signal integrity and improving the stability and accuracy of the monitoring system.

[0119] Flexible bandstop filter parameter configuration effectively addresses diverse noise sources and frequency characteristics under both moored and towing conditions. By precisely controlling the stopband center frequency, bandwidth, and attenuation depth, the system dynamically adjusts parameters under varying vessel conditions, optimally suppressing unnecessary noise components and maximizing the retention of underwater bio-acoustic signatures. Furthermore, this dynamic adjustment strategy enables the system to maintain consistently high performance in a volatile marine environment, improving the accuracy and stability of underwater bio-acoustic identification and noise pollution monitoring, thereby providing stronger technical support for marine environmental protection and underwater acoustic research.

[0120] In one possible implementation, the system design first requires collecting raw noise signals under different operating conditions. For example, this may include noise signals from a ship sailing, anchoring, or towing. These raw signals can be collected using underwater microphones (such as hydrophones), which contain various environmental noises and underwater biological soundprints.

[0121] Based on the target underwater creature's soundprint characteristics, the system uses a bandpass filter to process the raw noise signal. This filter removes unnecessary frequency components while retaining noise fragments that overlap with the target creature's soundprint. The soundprint of underwater creatures typically has a specific frequency range, and the filtering process helps identify and extract noise components that could be confused with the creature's soundprint.

[0122] Next, the system mixes the extracted noise fragments with the pure bio-voiceprint signal according to a preset signal-to-noise ratio (SNR) gradient. In this way, the system can simulate bio-voiceprint signals under different noise contamination conditions and generate mixed data with different signal-to-noise ratios (SNRs). The core purpose of this step is to provide rich and diverse training data for subsequent dual-stream network training, covering situations in different noise environments.

[0123] The injection of noise signatures allows the system to better adapt to the complex noise conditions found in real-world ocean environments. By simulating mixed signals with varying signal-to-noise ratios, the trained dual-stream model demonstrates enhanced robustness and adaptability, enabling it to effectively handle underwater bio-voiceprint recognition tasks under diverse noise pollution conditions.

[0124] The first stream of the two-stream network receives a time-frequency graph of the mixed voiceprint data after noise injection. This graph is obtained by applying methods such as the Short-Time Fourier Transform (STFT) to the biometric voiceprint signal, and it simultaneously displays both the time and frequency characteristics of the signal. The first stream network extracts the spatial-temporal characteristics of the voiceprint from the time-frequency graph through convolution operations.

[0125] The second stream of the dual-stream network receives the noise type label and pollution level vector corresponding to the mixed voiceprint signal. These labels and vectors provide information about the noise type (such as propeller noise, hull vibration noise, etc.) and the intensity of the noise pollution. By inputting the noise classification information and pollution intensity vector into the network, the second stream network can learn the relationship between noise and biological voiceprints, helping to provide noise perception capabilities to the first stream network during the training process.

[0126] The dual-stream network design enables the model to not only learn the characteristics of biometric voiceprints but also effectively utilize information about noise type and pollution level. This multimodal input approach enables the network to accurately distinguish between biometric voiceprints and noise in complex noisy environments, improving recognition accuracy and robustness.

[0127] At the feature fusion layer, the system first generates a frequency domain mask based on the noise pollution level information output by the second-stream network. The frequency domain mask determines which frequency components should be emphasized and which should be suppressed by calculating the intensity of the noise pollution. For example, the mask will assign a lower weight to frequency bands with strong noise pollution, and conversely, a higher weight to frequency bands with light noise pollution.

[0128] The generated frequency-domain mask is then used to perform channel-weighted processing on the convolutional feature map of the first-stream network. Channel weighting adjusts the contribution of the feature map by weighting different frequency channels, highlighting those parts that are meaningful for identifying underwater creatures' voiceprints while suppressing those parts that are more affected by noise. In this way, the system can effectively improve the accuracy of voiceprint recognition, especially in environments with severe noise pollution.

[0129] The spectral weighting mechanism enables the network to better perceive the varying intensities of noise pollution and dynamically adjust different frequency components. This noise-aware feature fusion method can significantly improve the recognition of biometric voiceprints in noisy environments, especially in complex noise backgrounds, helping the network avoid noise interference and improving the system's robustness and accuracy.

[0130] By constructing a dual-stream recognition model and integrating a noise-aware mechanism into the network, the performance of the underwater bio-voiceprint recognition system in noise-polluted environments can be greatly improved. First, by injecting and mixing noise features, the system can cope with noise conditions found in a variety of real-world environments. Second, the dual-stream network design combines bio-voiceprint features with noise pollution information, enabling the system to more accurately identify underwater bio-voiceprints. Finally, the spectral weighting method of the feature fusion layer effectively reduces the interference of noise on the recognition results, thereby ensuring the accuracy of underwater bio-voiceprint recognition. These technical features combined enable the system to achieve efficient and accurate underwater bio-voiceprint recognition and noise pollution monitoring in a changing marine environment.

[0131] In one possible implementation, a typical noise type library is first established, including propeller noise, hull vibration, marine industrial machinery sound, bubble sound, earthquake background noise, etc. Each noise type has different spectral distribution characteristics.

[0132] Based on the hearing thresholds of underwater creatures (such as whales, dolphins, and fish) at different frequency bands, inverse frequency response curves are plotted. A lower hearing threshold indicates greater sensitivity to that frequency band; therefore, the inverse of this threshold is used as a "sensitivity value" to quantify hearing sensitivity in each frequency band.

[0133] A two-dimensional mapping matrix is ​​formed, with rows representing different noise types and columns representing frequency bands (e.g., 0–2kHz, 2–4kHz, etc.). The cell value is the biological auditory sensitivity weight corresponding to the noise type in a specific frequency band. This table serves as the basis for subsequent mask weight calculations.

[0134] The second-stream network predicts the contamination level of each noise type in the current signal (e.g., a normalized value of 0–1, or a graded value of 0 to 5).

[0135] The system multiplies the current noise pollution level by the corresponding frequency band sensitivity value to obtain a weighted masking coefficient for each frequency band. For example, if organisms are highly sensitive to a certain frequency band and the noise pollution level is high, this frequency band will receive a lower masking weight to suppress interference.

[0136] To avoid excessive mutations in the mask that may affect feature learning, all frequency band weight coefficients are normalized and smoothed using Gaussian filtering to maintain the continuity of the frequency domain mask and its consistency with biological auditory perception.

[0137] In the convolutional feature map output by the first-stream network, the band mask weights calculated in the previous step are applied channel by channel according to the distribution structure of the frequency channels.

[0138] Fusion of noise-aware weighted feature maps: This process can be viewed as a noise-aware attention mechanism that dynamically suppresses frequency band features susceptible to interference while retaining frequency band features that are useful for biometric voiceprint recognition.

[0139] Enhanced biological perception consistency: This mechanism simulates the real auditory sensitivity of underwater organisms, allowing the system to focus on frequency bands that are meaningful to organisms during the feature extraction stage, thereby achieving alignment with the natural perception mechanism of organisms.

[0140] By dynamically adjusting the frequency band response based on the noise pollution level, the system can adaptively optimize the feature map weights according to the noise intensity in the actual environment, thereby improving recognition stability and accuracy in complex noise backgrounds.

[0141] By combining structural knowledge (auditory sensitivity) with data-driven signals (noise pollution level), a rule-driven data fusion mechanism is introduced to effectively improve the generalization ability and interpretability of the model.

[0142] The system is no longer just a simple filtering and noise reduction, but uses biological perception logic to rationally "shield" specific frequency bands to improve the intelligence and ecological adaptability of the recognition system.

[0143] This method provides a spectrum mask generation method that is closer to the actual perception mechanism of underwater organisms. It can significantly improve the system's recognition accuracy, robustness and ecological perception consistency of underwater organism voiceprints in a noise-polluted environment. It is an important optimization and supplement to traditional masking technologies based on energy thresholds or convolutional attention mechanisms.

[0144] In one possible implementation, the system first captures acoustic signals from organisms using underwater acoustic sensors and performs signal processing, such as noise removal and feature extraction, to obtain a sequence of acoustic pulses emitted by the organisms. Each acoustic pulse represents a signal emitted by underwater organisms during activities such as communication, marking, and positioning.

[0145] The time intervals between adjacent pulses in the bioacoustic pulse sequence are analyzed, the time intervals between each pair of adjacent pulses are calculated, and their coefficient of variation is obtained. The formula for the coefficient of variation is:

[0146] Where σ is the standard deviation of the intervals between adjacent pulses, and μ is its mean. By comparing the coefficient of variation of these pulse intervals with the standard deviation of the typical interval for that species, we can determine whether an organism is experiencing noise interference. Noise pollution typically causes irregular fluctuations in the intervals between adjacent pulses, which in turn increases the coefficient of variation.

[0147] The system queries the aquatic organism acoustic behavior database to obtain the typical interval standard deviation of the species (such as the fluctuation range of the normal communication pulse interval of different species in their natural state) so as to compare it with the coefficient of variation calculated in real time to identify whether there is communication interference.

[0148] The system uses multiple underwater acoustic sensors to obtain the location and movement trajectory of biological sound sources and uses sound source localization techniques (such as multilateration and triangulation) to calculate the motion vector of underwater organisms, including speed and direction.

[0149] Based on the location information of the noise source, the angle between the underwater creature's movement direction and the noise source is calculated. If this angle remains in the obtuse range (i.e., an angle greater than 90 degrees, indicating that the creature is moving away from the noise source) and the creature's movement acceleration exceeds the species' normal threshold (i.e., the creature's normal movement acceleration), it is considered an escape behavior.

[0150] The normal acceleration threshold is provided by a database of aquatic organism acoustic behavior, which contains information on typical motion accelerations of different species in normal environments. The system queries the database to set a standard acceleration value for comparison with the organism motion data during actual monitoring.

[0151] This method can accurately analyze the changes in the acoustic behavior of underwater organisms in a noise pollution environment, thereby providing more detailed and accurate underwater biological behavior monitoring, and has significant ecological monitoring and protection value.

[0152] In one possible implementation, the system uses underwater acoustic sensors to detect noise events in the water, recording the event timestamps and sound pressure levels. Noise events include various man-made or natural noise sources, such as underwater sounds from ships and operational noise from subsea equipment.

[0153] The organism's response includes an acoustic communication interference index and an escape behavior intensity value. The system tracks the organism's behavioral response using acoustic sensors and other monitoring equipment (such as underwater motion sensors). For example, the acoustic communication interference index reflects the degree to which an organism's communication is affected by a noisy environment, while the escape behavior intensity value measures the intensity of the organism's escape response.

[0154] The noise event and the biological response data are associated according to the time and space coordinate points to form a spatiotemporal correlation matrix. Each data record in the matrix includes: the noise event timestamp, the sound pressure level of the noise event, the biological response delay time, the acoustic communication interference index value, and the escape behavior intensity value.

[0155] This matrix can be used to analyze the impact of noise events on different biological species and their response patterns.

[0156] Underwater, the speed of sound wave propagation is affected by factors such as water depth, temperature, and salinity. The system calculates the time delay of sound wave propagation based on the environmental parameters of the water area.

[0157] Underwater organisms respond to external stimuli (such as noise) with a certain physiological delay. This delay is related to the organism's physiological characteristics, such as heart rate and reaction mechanism. Each organism has a different response delay, and the maximum physiological response delay can be estimated by using the heart rate cycle multiple of the species.

[0158] The system compares the observed biological response delay with the calculated acoustic propagation delay and the maximum biological physiological response delay. If the delay falls within the closed interval between these two delays, it is considered valid; otherwise, it is considered invalid data.

[0159] The system uses the Acoustic Communication Interference Index and the Escape Behavior Intensity value obtained in the previous steps as the primary basis for assessing the intensity of the organism's response. The Acoustic Communication Interference Index reflects the impact of noise on biological communication, while the Escape Behavior Intensity value reflects the organism's avoidance response to noise.

[0160] Different species have varying sensitivities to noise. The system queries a database of aquatic organism noise sensitivities and weights the acoustic communication interference index and escape behavior intensity values ​​based on the species' noise sensitivity. This weighted value more accurately reflects the actual impact of noise on a specific species.

[0161] The weighted acoustic communication interference index and the evasion behavior intensity value are combined to produce a comprehensive pollution responsibility index. This index indicates the intensity of noise pollution's impact on underwater life. When this index exceeds a preset threshold, the system deems the pollution source to have a significant impact on the ecological environment, triggering the generation of a responsibility report.

[0162] By constructing a spatiotemporal correlation matrix between noise events and biological responses, we can deeply analyze the spatiotemporal relationship between noise pollution events and biological responses. This helps to more accurately determine which noise events have negatively impacted specific species, providing a basis for environmental protection decision-making.

[0163] Verifying the validity of biological response delay times can effectively eliminate unqualified data and ensure the accuracy and reliability of data in pollution liability reports. This can prevent inaccurate conclusions caused by environmental interference or data collection errors.

[0164] By weighting the responses of different organisms based on their noise sensitivity, we can accurately assess the responsibility of noise pollution sources. When the pollution responsibility index exceeds the threshold, the system automatically generates a responsibility report that clearly identifies the specific impact of the pollution source, providing data support for accountability.

[0165] In one possible implementation, the system uses an underwater vibration sensor to monitor the anchor chain's vibration signals in real time, focusing on the anchor chain's primary vibration frequency (i.e., the anchor chain's dominant vibration frequency). This data is compared with the ambient background noise frequency to determine the frequency offset between the anchor chain's vibration frequency and the ambient background frequency.

[0166] Frequency deviation refers to the difference between the anchor chain's primary vibration frequency and the ambient background noise frequency. This difference can cause the anchor chain's vibration frequency to overlap with the acoustic perception frequency band of underwater organisms, thereby affecting their communication and behavior. By calculating the frequency deviation rate, the system can determine whether the current anchoring conditions are generating unnecessary noise interference.

[0167] The vibration of the anchor chain is closely related to its release length. Based on the calculated frequency deviation rate, the system automatically adjusts the release length to change the chain's vibration frequency. The adjustment is linearly related to the frequency deviation rate: the greater the frequency deviation, the greater the release length adjustment. This adjustment allows the chain's vibration frequency to be adjusted to a range that does not disturb underwater life, reducing noise pollution.

[0168] By combining the biometric voiceprint recognition system with behavioral data from aquatic organisms, the system can determine the target organism's auditory sensitivity band. Different aquatic organisms have different sensitivity bands to sound; some species may be more sensitive to low-frequency sounds, while others are more sensitive to high-frequency sounds.

[0169] The system analyzes the relationship between the propeller's rotational speed and the frequency of the sound waves it generates. Propeller rotation generates a certain amount of sound waves. If the propeller's frequency approaches the sensitive frequency band of aquatic organisms, it may cause interference to the organisms. Therefore, the system needs to calculate a set of optimal speed solutions to shift the propeller's frequency away from the center frequency of the target organisms' auditory sensitivity band, thereby reducing interference.

[0170] Based on the constrained relationship between the center frequency of the target organism's sensitive frequency band and propeller speed, the system optimizes speed and generates a series of optimal speed scenarios. These scenarios ensure that the propeller noise frequency is as far away from the target organism's sensitive frequency band as possible, reducing acoustic interference to aquatic life. Speed ​​adjustments are made in real time through feedback control to ensure the system maintains optimal performance throughout the towing process.

[0171] The optimization of anchoring and towing conditions in the closed-loop feedback control interface can effectively reduce underwater noise pollution, optimize the operation mode of the ship, improve the intelligent control capability of the system, and provide more precise solutions for aquatic ecological protection.

[0172] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0173] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An integrated system for underwater bio-voiceprint recognition and noise pollution monitoring, applied to ships or marine platforms, characterized by: include: Underwater acoustic cooperative sensing array, including low-frequency pressure sensors, medium- and high-frequency hydrophones, and vibration accelerometers at the anchor chain guide holes, deployed at the hull waterline; The ship operating condition feature decoupling module identifies the anchoring state based on the time domain envelope characteristics of the anchor chain vibration acceleration signal collected by the vibration accelerometer at the anchor chain guide hole, and identifies the towing state based on the coupling relationship between the propeller noise line spectrum components collected by the medium and high frequency hydrophones and the ship speed data; Noise pollution adaptive analysis engine, which dynamically configures the band-stop filter parameters according to the mooring status and towing status, and generates the noise pollution level by calculating the noise energy proportion within the target biological frequency band; The bio-voiceprint anti-interference recognition module decouples the ship operating condition feature from the ship operating condition noise feature, injects the bio-voiceprint training set into the dual-stream recognition model, and ultimately outputs the bio-voiceprint and bio-acoustic behavior distortion of the target organism. The pollution biological effect assessment module takes noise pollution level, biological soundprint and bioacoustic behavior distortion as input, and generates a pollution responsibility report by analyzing the spatiotemporal correlation between bioacoustic behavior distortion and noise events.

2. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 1 is characterized in that: The layout of the underwater acoustic cooperative sensing array satisfies: The spacing between any adjacent low-frequency pressure sensors and medium- and high-frequency hydrophones is evenly distributed according to the circumference of the hull waterline, and the number is determined by the ratio of the hull length to the reference spacing; The low-frequency pressure sensor and the mid- and high-frequency hydrophones are integrated into an acoustic shield. The shield's waveguide structure is designed so that the incident sound wave enters the low-frequency pressure sensor after attenuating the hull's self-noise through the Helmholtz resonator. The attenuation frequency band covers the main frequency of the ship's mechanical vibration. The installation direction of the vibration accelerometer is determined according to the stress analysis of the anchor chain guide hole, so that the angle between its sensitive axis and the normal vector of the anchor chain plane is less than the threshold angle.

3. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 1 is characterized in that: The process of identifying the anchoring state by the ship operating condition characteristic decoupling module includes: Envelope extraction is performed on the anchor chain vibration acceleration signal, and an adaptive threshold is used to detect impact events. Calculate the statistical distribution of the impact event intervals. When the coefficient of variation of the distribution peak is lower than the set threshold and the peak position and the anchor chain length meet the positive correlation model, it is determined to be a periodic impact. The positive correlation model is calibrated through ship mooring physical experiments: the impact interval is measured at various anchor chain lengths, and a linear fitting equation of length-interval is established.

4. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 3 is characterized in that: The adaptive threshold detection specifically adopts a dynamic time window mechanism: The initial width of the time window is set according to the rated impact duration of the anchor winch corresponding to the ship's tonnage; The real-time adjustment coefficient of the window width is the ratio of the current anchor chain length to the reference length; The local maximum of the signal envelope is calculated within the window, and when the maximum amplitude exceeds a multiple of the average amplitude of the previous N windows, it is marked as a valid impact, and the multiple is dynamically adjusted by the historical noise level.

5. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 1 is characterized in that: The dynamic configuration of the band-stop filter parameters includes: Generation of stopband center frequency in anchored state: Extract the fundamental frequency and harmonic components of the anchor chain vibration signal, and select the component with energy proportion exceeding the total energy proportion as the stopband center; Determination of stopband bandwidth in towing state: According to the rate of change of propeller blade frequency with ship speed, the stopband bandwidth expansion coefficient is calculated. This coefficient is positively correlated with the rate of change of ship speed. Adjustment of stopband attenuation depth: Based on the percentile value of the current noise pollution level in historical data, the attenuation depth value is output according to the preset mapping relationship.

6. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 1 is characterized in that: The process of building the dual-stream recognition model includes: Noise feature injection: Collect the original noise signal under standard working conditions, extract the noise fragments that overlap with the target biological voiceprint frequency band through bandpass filtering, and mix them with the pure biological voiceprint according to the preset signal-to-noise ratio gradient; Two-stream network training: The first-stream network inputs the time-frequency graph of the mixed voiceprint, and the second-stream network inputs the noise type label and pollution level vector; The feature fusion layer performs noise-aware spectral weighting: it generates a frequency domain mask based on the noise pollution level of the second stream output and performs channel-wise weighting on the convolutional feature maps of the first stream.

7. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 6 is characterized in that: The frequency domain mask generation specifically adopts the noise sensitivity mapping mechanism: Establish a mapping table between noise type and biological hearing sensitivity, where the sensitivity value is generated by the inverse of the species hearing threshold curve, where the hearing threshold is the minimum sound pressure level that a specific species can perceive; Multiply the noise pollution level by the sensitivity value to generate the mask weight coefficient for each frequency band; The mask weight is applied to the feature map output by the first stream network according to the frequency channel.

8. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 1 is characterized in that: The analytical bioacoustic behavior distortion comprises: Calculation of acoustic communication interference index: Calculation of acoustic communication interference index: Extract the identified biological sound pulse sequence and calculate the real-time standard deviation σ of the interval between adjacent pulses 实时 , by querying the aquatic organism acoustic behavior database to obtain the species typical interval standard deviation σ 典型 , calculate the acoustic communication interference index value according to the following formula: Acoustic communication interference index = σ 典型 / σ 实时 ; Determination of the intensity of escape behavior: Calculate the biological motion vector through the sound source localization trajectory when the following conditions are met: The angle between the vector direction and the noise source direction is continuously greater than 90°; If the movement acceleration exceeds the species normal acceleration threshold, the escape behavior intensity value is output. The escape behavior intensity value is the ratio of the measured acceleration to the species normal acceleration threshold. The acoustic communication interference index value and the escape behavior intensity value together constitute the bioacoustic behavior distortion value.

9. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 8 is characterized in that: The generation of the pollution liability report includes: Construct a spatiotemporal correlation matrix between noise events and biological responses. The matrix includes noise event timestamps, sound pressure levels, biological response delay times, acoustic communication interference index values, and escape behavior intensity values. Verification of the validity of the biological response delay time: the delay time is within the closed interval formed by the sound wave propagation delay and the maximum physiological response delay of the biological species, and the maximum physiological response delay is determined according to the multiple of the heart rate cycle of the biological species; Pollution responsibility index calculation: The acoustic communication interference index and escape behavior intensity value are weighted and integrated according to the noise sensitivity of the species to obtain the pollution responsibility index. When the index exceeds the preset threshold, a responsibility report is generated.

10. The underwater biological voiceprint recognition and noise pollution monitoring integrated system according to claim 1 is characterized in that: It also includes a closed-loop feedback control interface, whose execution logic includes: Anchoring condition optimization: The anchor chain release length adjustment is calculated based on the deviation rate of the anchor chain vibration main frequency relative to the ambient background frequency. The adjustment amount is linearly related to the frequency deviation rate. Towing condition optimization: Based on the distribution of the target organism's auditory sensitive frequency band, the optimal speed solution set that causes the propeller blade frequency to deviate from the center of the sensitive frequency band is solved. This solution set is generated through the constraint relationship between the center frequency of the sensitive frequency band and the propeller speed.

Images