Water supply pipeline leakage identification method and device

Abstract

The present disclosure relates to a water supply pipeline leakage identification method, a water supply pipeline leakage identification device, a storage medium and an electronic device. The water supply pipeline leakage identification method comprises obtaining a first acoustic feature generated when the water supply pipeline does not leak under a plurality of characteristic factors in a predetermined environment, and a second acoustic feature generated under a plurality of water supply leakage states; comparing the acoustic features under any two water supply leakage states in which only a unique different factor exists in the plurality of water supply leakage states, and establishing an acoustic feature library of the water supply pipeline in the predetermined environment; based on the acoustic feature library of the water supply pipeline in the predetermined environment, model training is performed to obtain an acoustic recognition model; inputting the MEL spectrum feature corresponding to the sound signal generated under the current state of the water supply pipeline in the predetermined environment to the acoustic recognition model to obtain the water supply leakage state of the current state of the water supply pipeline in the predetermined environment, which is conducive to the rapid identification and detection of the water supply leakage state of the water supply pipeline and reduces the false detection and missed detection.

Description

Technical Field

[0001] This disclosure relates to the field of pipeline inspection technology, and in particular to a method for identifying leaks in water supply pipelines, a device for identifying leaks in water supply pipelines, a storage medium, and electronic equipment. Background Technology

[0002] As a crucial component of urban infrastructure, water supply networks play a vital role in residents' lives and the development of industry and agriculture, driven by technological advancements and the ever-growing demands of rapid urban expansion. However, in actual engineering environments, water supply pipelines undergo various structural damages during construction and operation. These damages can occur due to changes in the underground environment, impacts from surrounding construction, drastic temperature fluctuations, and long-term corrosion and oxidation. These damages can lead to water loss and corresponding economic losses, and in severe cases, even ground subsidence, seriously threatening urban road traffic and public safety. Therefore, continuous pipeline inspection is necessary to promptly detect leaks. Traditional water supply network leakage monitoring primarily relies on manual inspections. This involves assigning personnel to periodically patrol the pipeline route, using equipment such as listening rods, pipe locators, and electronic amplification leak detectors to determine if leaks have occurred. Manual inspections are highly manpower-intensive, their effectiveness depends heavily on the pipeline environment, and they are prone to misdiagnosis of leaks. Summary of the Invention

[0003] In view of this, the present disclosure aims to provide a method for identifying leaks in water supply pipelines, a device for identifying leaks in water supply pipelines, a storage medium, and an electronic device.

[0004] The technical solution disclosed herein is implemented as follows:

[0005] Firstly, this disclosure provides a method for identifying leaks in water supply pipelines.

[0006] The water supply pipeline leakage identification method provided in this disclosure includes:

[0007] Acquire the first acoustic signature feature generated when there is no leakage in the water supply pipeline under various characteristic factors in a predetermined environment, and the second acoustic signature feature generated under various water supply leakage conditions;

[0008] The acoustic signature features of the water supply pipeline under the predetermined environment are compared between two water supply states with the same characteristic factors, namely, no leakage and multiple water supply leakage states.

[0009] Based on the voiceprint feature library of water supply pipelines under the predetermined environment, a model is trained to obtain a voiceprint recognition model.

[0010] Input the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage status of the water supply pipeline under the current state in the predetermined environment.

[0011] In some embodiments, the multiple characteristic factors in the predetermined environment include at least one of the following:

[0012] Characteristics of the pipeline itself, geographical environment of the pipeline, water supply flow rate, and water supply pressure;

[0013] The characteristics of the pipe itself include pipe diameter, pipe material, and pipe straightness / bending status;

[0014] The geographical environmental characteristics of the pipeline include the hardness of the soil layer near the pipeline and the moisture content of the soil layer near the pipeline.

[0015] The comparison of acoustic signature features under two water supply conditions—one with no leakage and the other with multiple leakage conditions—that share the same characteristic factors, to establish an acoustic signature feature library for the water supply pipeline under the predetermined environment includes:

[0016] The acoustic signature features of water supply pipelines under the predetermined environment are compared between any two water supply leakage states where only one of the following factors differs: pipe diameter, pipe material, pipe straightness, soil hardness near the pipe, soil moisture near the pipe, water supply flow rate, and water supply pressure.

[0017] In some embodiments, before inputting the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment, the method includes:

[0018] Collect sound signals generated by the water supply pipeline under the current state of a predetermined environment;

[0019] A short-time Fourier transform is performed on the sound signal generated by the water supply pipeline under the current state in the predetermined environment to obtain the MEL spectral characteristics corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

[0020] In some embodiments, performing a short-time Fourier transform on the sound signal generated by the water supply pipeline under the current state in the predetermined environment to obtain the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment includes:

[0021] If the sound signal generated by the water supply pipeline under the predetermined environment is in the low-frequency region below the predetermined frequency, then the sound signal generated by the water supply pipeline under the current state under the predetermined environment is subjected to frame-by-frame windowing processing to obtain a windowed signal.

[0022] The windowed signal is subjected to a short-time Fourier transform to obtain the MEL spectral characteristics of the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

[0023] In some embodiments, performing a short-time Fourier transform on the windowed signal to obtain the MEL spectral characteristics corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment includes:

[0024] Perform a short-time Fourier transform on the windowed signal, and then perform spectral correction on the obtained MEL spectral features;

[0025] The MEL spectral features after spectral correction are used as the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment; wherein...

[0026] ; To perform a short-time Fourier transform on the windowed signal, the signal frequency at the MEL scale is obtained. This refers to the signal frequency on the MEL scale after spectral correction.

[0027] In some embodiments, the voiceprint recognition model includes a correspondence between voiceprint features and the state of water supply leakage in the pipeline;

[0028] The process of inputting the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment includes:

[0029] Based on the voiceprint recognition model, the MEL spectrum features corresponding to the sound signals generated by the water supply pipeline under the current state in the predetermined environment are identified, and the MEL spectrum features are compared with the voiceprint features in the voiceprint feature library to obtain the comparison results.

[0030] Based on the comparison results and the correspondence between the voiceprint features and the water supply leakage status of the pipeline, the water supply leakage status of the current state of the water supply pipeline under the predetermined environment is determined.

[0031] In some embodiments, the step of training a model based on the voiceprint feature database of the water supply pipeline under the predetermined environment to obtain a voiceprint recognition model includes:

[0032] Based on the aforementioned voiceprint feature database, a GMM-UBM model is trained using the Expectation-Maximization (EM) algorithm to obtain a voiceprint recognition model; wherein, the probability density function of the GMM distribution is shown in the following equation:

[0033]

[0034] In the formula, For input MFCC voiceprint feature vector For the probability density function of the j-th GMM model, and Let be the expectation and covariance matrices of the j-th GMM model, respectively; where is the input voiceprint feature vector. The probability of belonging to the j-th cluster can be expressed as:

[0035] ;

[0036] In the formula, Let be the weight coefficients of the j-th GMM model; where are the parameters of the GMM model. , , The EM algorithm is used for initialization and iterative updates until convergence. The updated parameters are then combined with the UBM parameters to adjust the impact of the GMM model parameters on the final prediction model. The combination formula is shown below:

[0037]

[0038] In the formula, , , These are the parameters of the GMM model after combining them with the UBM parameters. , , They are respectively , , The weighting coefficients of the UBM parameters, where T is the number of UBM components, and n j , , These are the weights, expectations, and covariance matrices of the UBM components, respectively. For normalization coefficients, the GMM and UBM parameters are fused to establish a GMM-UBM-based voiceprint recognition model, as shown in the following equation:

[0039] ;

[0040] In the formula, The posterior probability that the input sound signal belongs to the GMM-UBM model.

[0041] Secondly, this disclosure provides a water supply pipeline leak detection device, comprising:

[0042] The voiceprint feature acquisition module is used to acquire the first voiceprint feature generated when there is no leakage in the water supply pipeline under various characteristic factors in a predetermined environment, and the second voiceprint feature generated under various water supply leakage conditions.

[0043] The feature library establishment module is used to compare the voiceprint features of the water supply pipeline under two water supply states with the same characteristic factors in the absence of leakage and multiple water supply leakage states, and to establish the voiceprint feature library of the water supply pipeline under the predetermined environment.

[0044] The model training module is used to train the model based on the voiceprint feature library of the water supply pipeline under the predetermined environment to obtain the voiceprint recognition model.

[0045] The leakage state prediction module is used to input the MEL spectrum features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment.

[0046] Thirdly, this disclosure provides a computer-readable storage medium storing a water supply pipeline leak identification program thereon, which, when executed by a processor, implements the water supply pipeline leak identification method described in the first aspect above.

[0047] Fourthly, this disclosure provides an electronic device, including a memory, a processor, and a water supply pipeline leak identification program stored in the memory and executable on the processor. When the processor executes the water supply pipeline leak identification program, it implements the water supply pipeline leak identification method described in the first aspect above.

[0048] The water supply pipeline leakage identification method according to embodiments of this disclosure includes acquiring a first acoustic signature feature generated when there is no leakage in the water supply pipeline under multiple characteristic factors in a predetermined environment, and a second acoustic signature feature generated under multiple water supply leakage states; comparing the acoustic signature features of any two water supply leakage states with only one different factor to establish an acoustic signature feature library of the water supply pipeline under the predetermined environment; training a model based on the acoustic signature feature library of the water supply pipeline under the predetermined environment to obtain an acoustic signature recognition model; and inputting the MEL spectrum feature corresponding to the sound signal generated under the current state of the water supply pipeline under the predetermined environment to the acoustic signature recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment. This application can obtain an acoustic signature recognition model by establishing an acoustic signature feature library of multiple different water supply leakage states of the water supply pipeline and training a model. Then, by inputting the MEL spectrum feature corresponding to the sound signal generated under the current state of the water supply pipeline under the predetermined environment to the acoustic signature recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment, it is beneficial to quickly identify and detect water supply leakage states of the water supply pipeline and reduce false detections and missed detections.

[0049] Additional aspects and advantages of this disclosure will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this disclosure. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating a water supply pipeline leak identification method according to an exemplary embodiment;

[0051] Figure 2 This is a flowchart illustrating the extraction of acoustic signature features of a water supply pipeline leak, according to an exemplary embodiment.

[0052] Figure 3 This is a flowchart illustrating acoustic signature identification of a water supply pipeline leak according to an exemplary embodiment.

[0053] Figure 4 This is a schematic diagram of a water supply pipeline leak detection device according to an exemplary embodiment. Detailed Implementation

[0054] Embodiments of this disclosure are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting this disclosure.

[0055] As a crucial component of urban infrastructure, water supply networks play a vital role in residents' lives and the development of industry and agriculture, driven by technological advancements and the ever-growing demands of rapid urban expansion. However, in actual engineering environments, water supply pipelines undergo various structural damages during construction and operation. These damages can occur due to changes in the underground environment, impacts from surrounding construction, drastic temperature fluctuations, and long-term corrosion and oxidation. These damages can lead to water resource loss and corresponding economic losses, and in severe cases, even ground subsidence, seriously threatening urban road traffic and public safety. Therefore, continuous pipeline inspection is necessary to promptly detect leaks. Traditional water supply network leakage monitoring primarily relies on manual inspections. This involves assigning personnel to periodically patrol the pipeline route, using equipment such as listening rods, pipe locators, and electronic amplification leak detectors to determine if leaks have occurred. Manual inspection methods are highly demanding in terms of manpower, and their effectiveness depends heavily on the pipeline environment, potentially leading to false or missed detections of leaks.

[0056] In response to the above situation, this disclosure provides a method for identifying leaks in water supply pipelines. Figure 1 This is a flowchart illustrating a water supply pipeline leak identification method according to an exemplary embodiment. Figure 1 As shown, the water supply pipeline leak detection method includes:

[0057] Step 10: Obtain the first acoustic signature generated when there is no leakage in the water supply pipeline under various characteristic factors in the predetermined environment, and the second acoustic signature generated under various water supply leakage conditions;

[0058] Step 11: Compare the acoustic signature features of the water supply pipeline under two water supply conditions with the same characteristic factors in the absence of leakage and multiple water supply leakage conditions, and establish an acoustic signature feature library of the water supply pipeline under the predetermined environment.

[0059] Step 12: Based on the voiceprint feature library of the water supply pipeline under the predetermined environment, perform model training to obtain a voiceprint recognition model;

[0060] Step 13: Input the MEL spectrum features corresponding to the sound signal generated by the water supply pipeline under the current state of the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment.

[0061] In this exemplary embodiment, the water supply pipeline in the predetermined environment can be a water supply pipeline in a known specific environment, such as a water supply pipeline under a street in a city.

[0062] In this exemplary embodiment, the voiceprint features of water supply pipes under two different water supply conditions—one with no leakage and the other with multiple leakage conditions—are compared to establish a voiceprint feature database for water supply pipes in the predetermined environment. For example, water supply pipe A under a street in a city has pipe characteristics (diameter 100mm), pipe geographical environment characteristics, etc. When there is no leakage, it corresponds to the first voiceprint feature; and when water supply pipe A has the same pipe characteristics (diameter 100mm) and pipe geographical environment characteristics, and there is a leakage, it corresponds to the second voiceprint feature. These features are compared and stored in the voiceprint feature database.

[0063] For example, water supply pipe B under a street in a certain city has the characteristics of the pipe itself (diameter 80mm) and the geographical environment of the pipe. When there is no leakage, it corresponds to the first voiceprint feature. Water supply pipe A has the same characteristics of the pipe itself (diameter 80mm) and the geographical environment of the pipe. When there is leakage, it corresponds to the second voiceprint feature. The features are compared and stored in the voiceprint feature database.

[0064] The water supply pipeline leakage identification method according to embodiments of this disclosure includes acquiring a first acoustic signature feature generated when there is no leakage in the water supply pipeline under multiple characteristic factors in a predetermined environment, and a second acoustic signature feature generated under multiple water supply leakage states; comparing the acoustic signature features of any two water supply leakage states with only one different factor to establish an acoustic signature feature library of the water supply pipeline under the predetermined environment; training a model based on the acoustic signature feature library of the water supply pipeline under the predetermined environment to obtain an acoustic signature recognition model; and inputting the MEL spectrum feature corresponding to the sound signal generated under the current state of the water supply pipeline under the predetermined environment to the acoustic signature recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment. This application can obtain an acoustic signature recognition model by establishing an acoustic signature feature library of multiple different water supply leakage states of the water supply pipeline and training a model. Then, by inputting the MEL spectrum feature corresponding to the sound signal generated under the current state of the water supply pipeline under the predetermined environment to the acoustic signature recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment, it is beneficial to quickly identify and detect water supply leakage states of the water supply pipeline and reduce false detections and missed detections.

[0065] In some embodiments, the multiple characteristic factors in the predetermined environment include at least one of the following:

[0066] Characteristics of the pipeline itself, geographical environment of the pipeline, water supply flow rate, and water supply pressure;

[0067] The characteristics of the pipe itself include pipe diameter, pipe material, and pipe straightness / bending status;

[0068] The geographical environmental characteristics of the pipeline include the hardness of the soil layer near the pipeline and the moisture content of the soil layer near the pipeline.

[0069] The comparison of acoustic signature features under two water supply conditions—one with no leakage and the other with multiple leakage conditions—that share the same characteristic factors, to establish an acoustic signature feature library for the water supply pipeline under the predetermined environment includes:

[0070] The acoustic signature features of water supply pipelines under the predetermined environment are compared between any two water supply leakage states where only one of the following factors differs: pipe diameter, pipe material, pipe straightness, soil hardness near the pipe, soil moisture near the pipe, water supply flow rate, and water supply pressure.

[0071] In this exemplary embodiment, there are multiple characteristic factors in the predetermined environment, including pipe characteristics, pipe geographical environment characteristics, water supply flow rate, water supply pressure, etc. The voiceprint features of the water supply pipes under two different water supply states—one without leakage and the other with multiple leakage conditions—are compared to establish a voiceprint feature database for the water supply pipes in the predetermined environment. For example, a water supply pipe C under a street in a city has pipe characteristics (50mm diameter) and pipe geographical environment characteristics. When there is no leakage, it corresponds to the first voiceprint feature. Conversely, when water supply pipe C has the same pipe characteristics (50mm diameter) and pipe geographical environment characteristics, but a leakage occurs, it corresponds to the second voiceprint feature. These features are compared and stored in the voiceprint feature database.

[0072] In some embodiments, before inputting the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment, the method includes:

[0073] Collect sound signals generated by the water supply pipeline under the current state of a predetermined environment;

[0074] A short-time Fourier transform is performed on the sound signal generated by the water supply pipeline under the current state in the predetermined environment to obtain the MEL spectral characteristics corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

[0075] In this exemplary embodiment, performing a short-time Fourier transform on the sound signal generated by the water supply pipeline under the current state in the predetermined environment to obtain the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment includes:

[0076] If the sound signal generated by the water supply pipeline under the predetermined environment is in the low-frequency region below the predetermined frequency, then the sound signal generated by the water supply pipeline under the current state under the predetermined environment is subjected to frame-by-frame windowing processing to obtain a windowed signal.

[0077] The windowed signal is subjected to a short-time Fourier transform to obtain the MEL spectrum characteristics of the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

[0078] In this exemplary embodiment, a water supply pipeline leakage simulation test platform is set up with water supply pipelines made of various materials commonly found in actual engineering sites, including ductile iron, galvanized steel, stainless steel, PVC, and PE. Each material should include multiple pipelines with diameters such as DN100, DN150, DN200, DN250, DN300, and DN400. Pipeline characteristics, such as diameter, material, water pressure, flow rate, underground environment, wetness / dryness, and straightness / bends, affect various aspects, including the acoustic frequency band and intensity (frequency band, intensity, spectrum).

[0079] Raw data of underground ambient noise and sound sensor system noise were obtained through a water supply pipeline leakage simulation experiment under non-operational conditions with no leakage in the pipeline.

[0080] Inverters, drainage pipes, semi-open pump stations, and valves are installed near the sensor sound acquisition points to collect sound data of external interference noise under various environmental interference conditions, which are then used to collect leakage signals under various pipe materials and diameters.

[0081] Multiple simulated leak points were arranged at one end of the water supply pipeline simulation test platform. The types of simulated leak points included valve-controllable nozzles in circular, square, triangular, and wedge shapes, as well as irregularly shaped cracked welds and corrosion perforations formed by direct cutting or drilling on the pipeline surface. The leakage flow rates for each type of simulated leak point were designed to be 2 L / min, 4 L / min, 8 L / min, and 16 L / min, respectively. Electromagnetic pumps were used to set the water flow velocities in the pipeline to common actual engineering site conditions of 0.4 m / s, 0.8 m / s, 1.2 m / s, and 1.6 m / s, respectively, to collect raw leakage sound data for each type of simulated leak point shape, flow rate, and water flow velocity.

[0082] Near the leak point in the simulated water supply pipeline experimental platform, straight pipes with right-angle bends, U-bends, and no bends were respectively arranged. The aim was to collect sound signals propagating through the straight pipes, right-angle bends, and U-bends. Sensor sound data acquisition points were placed at distances of 20m, 40m, 60m, 80m, and 100m from the simulated leak point.

[0083] Based on the pipe material and diameter, pipe flow velocity, simulated leak point shape and flow rate, whether there are bends in the pipe and the type of bend, and the distance between the data sampling point and the simulated leak point set in the above steps, the original sound data under each combination of test factors are collected.

[0084] After collecting sufficient noise and leakage data on the water supply pipeline simulation platform, it is necessary to extract the MFCC (Melch Cepstral Coefficients) acoustic signature features from this data. The specific feature extraction process is as follows:

[0085] The acquired raw audio data is mainly distributed in the low-frequency region below 100Hz, so a high-pass filter can be used to process the acquired raw sound signal. Here, 100Hz is the filter cutoff frequency.

[0086] The pre-emphasized audio signal exhibits acoustic signature characteristics within the time domain. Therefore, framing and windowing processing are necessary before Fourier transform. Since leaky audio signals in actual engineering environments are relatively stable within the time domain, a larger frame length can be chosen; here, we select a specific frame length.

[0087] Take the frame overlap length Meanwhile, to reduce the impact of spectral leakage and ripple caused by signal truncation due to framing, the truncated signal needs to be multiplied by the original signal using a weighting function within the signal's time domain. This involves introducing a window function, and the common Hanning window is chosen as the window function in the windowing process. The formula for the Hanning window function is shown below:

[0088] ;

[0089] In the formula, Let N be the Hanning window function, and N be the window function length. The framed, windowed audio signal needs further STFT short-time Fourier transform to be converted to the time-frequency domain for further analysis and processing, as shown in the following equation:

[0090]

[0091] In the formula, This is the time-frequency domain representation of the sound signal after short-time Fourier transform. This refers to the time-domain representation of sound signals. To analyze the window function, a set of triangular filters based on the Mel frequency is constructed to eliminate fine structures in the audio data, extract envelope information, and reduce the amount of data required for computation. The center frequencies of multiple triangular filters within the Mel scale range need to be equally spaced pairwise, represented by the envelope formed by multiple overlapping isosceles triangles of equal size on the Mel scale, which are then mapped to the STFT linear scale. Since the traditional Mel scale is based on the human ear's ability to distinguish sound signals, it is more sensitive to changes in the low-frequency components of the signal. However, based on the spectral characteristics of data collected from the test environment and actual engineering sites, noise signals are mostly approximate white noise and low-frequency noise below 150Hz distributed across the entire frequency domain. Therefore, the original Mel scale needs to be corrected to a certain extent. The conversion formula between the corrected Mel scale and the linear scale frequency band is shown below:

[0092] The MEL spectral features after spectral correction are used as the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment; wherein...

[0093] ; To perform a short-time Fourier transform on the windowed signal, the signal frequency at the MEL scale is obtained. This represents the signal frequency under the MEL scale after spectral correction. Because this formula involves conversion between linear and exponential / logarithmic scales, the converted filter triangle windows are asymmetrical. Simultaneously, the spacing between the individual triangle filters gradually increases within 150Hz, remains constant in the 150-300Hz range, and gradually increases above 300Hz. This trend can highlight the acoustic signature characteristics near the high-frequency band of leakage signals. The formula for the triangle filter bank under the STFT linear scale is shown below: very large vector, multi-dimensional, frame energy, second derivative dynamic characteristics, static characteristics. Among them,

[0094]

[0095] In the formula, Here are the filter coefficients of the triangular window filter corresponding to frequency f on the STFT linear scale, where n is the filter sequence and f is the frequency. n-1 f n f n+1 These represent the frequencies of the left endpoint, maximum value, and right endpoint of the nth triangular window filter, respectively. The maximum filter coefficient is given. Subsequently, a triangular window filter bank is used to filter the calculated STFT power spectrum, and logarithmic operations are performed simultaneously on both ends of the filtered power spectrum to convert the multiplication operation of the short-time Fourier transform signal into an addition operation. This yields the Mel spectrum. A further discrete cosine transform (DCT) is performed on the calculated Mel spectrum to separate the signal envelope from the fine structure. The Mel cepstral coefficients (MFCCs) can then be calculated from the DCT-transformed Mel spectrum. The second form of the DCT used in this method is shown in the following equation:

[0096] ;

[0097] ;

[0098] In the formula, For Mel spectrum input, The MFCC obtained after discrete cosine transform. N The input spectrum length is... These are the compensation coefficients used for weighted calculations. On the other hand, performing first- and second-order difference calculations on MFCC yields a result that can describe the dynamic characteristics of sound signal leakage. The difference calculation formula for MFCC is shown below:

[0099]

[0100]

[0101] In the formula, and These are the first and second differences of MFCC, respectively. This refers to the time difference in the differential calculation. N denoted by the order of the Mel-frequency cepstral coefficients. Frame energy, representing the volume per unit length after the audio signal is framed, can also reflect the speaker characteristics of the audio signal. The formula for calculating frame energy is shown below:

[0102] ;

[0103] in, For the input sound signal The frame energy of the frame, n is the length of each frame after the input signal is divided into frames, and x[] is the sound signal in the time domain. At this time, the MFCC features, the first-order and second-order difference features of MFCC, and the frame energy features obtained above can be merged to form an N-dimensional MFCC voiceprint feature vector.

[0104] In this exemplary embodiment, performing a short-time Fourier transform on the windowed signal to obtain the MEL spectral characteristics corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment includes:

[0105] Perform a short-time Fourier transform on the windowed signal, and then perform spectral correction on the obtained MEL spectral features;

[0106] The MEL spectral features after spectral correction are used as the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

[0107] In some embodiments, the voiceprint recognition model includes a correspondence between voiceprint features and the state of water supply leakage in the pipeline;

[0108] The process of inputting the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment includes:

[0109] Based on the voiceprint recognition model, the MEL spectrum features corresponding to the sound signals generated by the water supply pipeline under the current state in the predetermined environment are identified, and the MEL spectrum features are compared with the voiceprint features in the voiceprint feature library to obtain the comparison results.

[0110] Based on the comparison results and the correspondence between the voiceprint features and the water supply leakage status of the pipeline, the water supply leakage status of the current state of the water supply pipeline under the predetermined environment is determined.

[0111] In this exemplary embodiment, the water supply leakage state includes both the presence of a leak and the absence of a leak. The correspondence between acoustic signature features and the water supply leakage state includes the acoustic signature features corresponding to the presence of a leak point in a water supply pipeline with multiple characteristic factors under a predetermined environment, and the acoustic signature features corresponding to the absence of a leak point in a water supply pipeline with the same characteristic factors.

[0112] The determination of the water supply leakage status of the water supply pipeline under the predetermined environment based on the comparison results and the correspondence between the voiceprint features and the water supply leakage status includes:

[0113] If the MEL spectral features are matched with the voiceprint features corresponding to the presence of a leak point in the voiceprint feature library, then the water supply leakage state of the water supply pipeline under the predetermined environment is determined to be the presence of a leak point.

[0114] If the MEL spectral features are matched with the voiceprint features corresponding to the absence of a leak point in the voiceprint feature library, then the water supply leakage state of the water supply pipeline under the predetermined environment is determined to be the absence of a leak point.

[0115] In some embodiments, the step of training a model based on the voiceprint feature database of the water supply pipeline under the predetermined environment to obtain a voiceprint recognition model includes:

[0116] Based on the aforementioned voiceprint feature database, a GMM-UBM model is trained using the Expectation-Maximization (EM) algorithm to obtain a voiceprint recognition model; wherein, the probability density function of the GMM distribution is shown in the following equation:

[0117] ;

[0118] In the formula, For input MFCC voiceprint feature vector For the probability density function of the j-th GMM model, and Let be the expectation and covariance matrices of the j-th GMM model, respectively; where is the input voiceprint feature vector. The probability of belonging to the j-th cluster can be expressed as:

[0119] ;

[0120] In the formula, Let be the weight coefficients of the j-th GMM model; where are the parameters of the GMM model. , , The EM algorithm is used for initialization and iterative updates until convergence. The updated parameters are then combined with the UBM parameters to adjust the impact of the GMM model parameters on the final prediction model. The combination formula is shown below:

[0121]

[0122] In the formula, , , These represent the weight coefficients, expected value, and variance of the GMM-UBM model after combining with UBM parameters. , , They are respectively , , The parameters correspond to the weighting coefficients, T is the number of UBM components, and n j , , These are the weights, expectations, and covariance matrices of the UBM components, respectively. For normalization coefficients, the GMM and UBM parameters are fused to establish a GMM-UBM-based voiceprint recognition model, as shown in the following equation:

[0123] ;

[0124] In the formula, This represents the posterior probability that the input sound signal belongs to the GMM-UBM model. equal The weight coefficients corresponding to the GMM-UBM model, equal The expected value corresponding to the GMM-UBM model, equal This represents the variance corresponding to the GMM-UBM model.

[0125] In this exemplary embodiment, sound signals in actual engineering sites are subject to interference from various external factors, such as the material and diameter of water supply pipes, their dimensions, the number and type of bends and valves, the soil quality and burial depth, the pressure and flow velocity within the pipes, the shape and size of leak points, and various types and intensities of external noise. These factors are difficult to fully represent in a simulated leak test platform due to space and cost limitations. Furthermore, the MFCC acoustic signature vector is only one type of sound signal characteristic and cannot comprehensively represent all the characteristics of noise and leak signals. Therefore, it is still necessary to update and iterate the database model parameters and classification cluster data using the collected sound signals and the verified pipe leakage results during the data collection and analysis process at the actual engineering site, in order to gradually improve the detection accuracy of the leak identification system.

[0126] Figure 2 This is a flowchart illustrating the extraction of acoustic signature features from a water supply pipeline leak, according to an exemplary embodiment. Figure 2 As shown, the process for extracting acoustic signatures of water supply pipeline leaks includes:

[0127] Step 20: Obtain the raw audio data;

[0128] Step 21: Pre-weighting treatment;

[0129] Step 22: Frame segmentation, including obtaining frame length. Take the frame overlap length ;

[0130] Step 23, Add window treatment, select Hanning windows, for example,

[0131] ;

[0132] Step 24: Short-time Fourier transform to the time-frequency domain, such as... ;

[0133] Step 25: Convert linearly to Mel scale and create a set of triangular filters, for example, ;

[0134] Step 26: Calculate the frame energy, for example... ;

[0135] Step 27: Obtain MFCC using Discrete Cosine Transform, for example... ;

[0136] Step 28: Calculate the first-order difference of MFCC;

[0137] Step 29: Merge MFCCs, first-order and second-order differences, and frame energy to obtain an N-dimensional MFCC voiceprint feature vector.

[0138] Figure 3 This is a flowchart illustrating the acoustic signature identification of a water supply pipeline leak, according to an exemplary embodiment. Figure 3 As shown, the process for identifying the acoustic signature of a water supply pipeline leak includes:

[0139] Step 30: Train the sound data;

[0140] Step 31: Extract the MFCC voiceprint feature vector;

[0141] Step 32: Obtain actual sound data;

[0142] Step 33: Extract the MFCC voiceprint feature vector;

[0143] Step 34: Use the EM algorithm to adjust the parameters of the GMM model. , , Continue training until convergence;

[0144] Step 35: Calculate the parameters of the GMM-UBM model, for example,

[0145] ;

[0146] Step 36: Determine the probability density function of GMM-UBM, for example,

[0147] ;

[0148] Step 37: Determine the GMM-UBM classification probability, for example,

[0149] ;

[0150] Step 38: Construct a water supply pipeline leakage model using the combined GMM-UBM model parameters, for example,

[0151] ;

[0152] Step 39: Determine the pipeline leakage status;

[0153] Step 40: Conduct on-site verification based on the predicted leakage status. (In this application) Figure 2 , Figure 3 The parameter definitions involved are explained in the above embodiments.

[0154] This disclosure provides a water supply pipeline leak detection device. Figure 4 This is a schematic diagram illustrating the structure of a water supply pipeline leak detection device according to an exemplary embodiment. Figure 4 As shown, the water supply pipeline leak detection device includes:

[0155] The voiceprint feature acquisition module 40 is used to acquire the first voiceprint feature generated when there is no leakage in the water supply pipeline under various feature factors in a predetermined environment, and the second voiceprint feature generated under various water supply leakage conditions.

[0156] The feature library establishment module 41 is used to compare the voiceprint features of the water supply pipeline under two water supply states with the same characteristic factors in the absence of leakage and multiple water supply leakage states, and to establish the voiceprint feature library of the water supply pipeline under the predetermined environment.

[0157] Model training module 42 is used to train a model based on the voiceprint feature library of the water supply pipeline under the predetermined environment to obtain a voiceprint recognition model.

[0158] The leakage state prediction module 43 is used to input the MEL spectrum features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the current state in the predetermined environment.

[0159] In this exemplary embodiment, the water supply pipeline in the predetermined environment can be a water supply pipeline in a known specific environment, such as a water supply pipeline under a street in a city.

[0160] In this exemplary embodiment, the voiceprint features of water supply pipes under two different water supply conditions—one with no leakage and the other with multiple leakage conditions—are compared to establish a voiceprint feature database for water supply pipes in the predetermined environment. For example, water supply pipe A under a street in a city has pipe characteristics (diameter 100mm), pipe geographical environment characteristics, etc. When there is no leakage, it corresponds to the first voiceprint feature; and when water supply pipe A has the same pipe characteristics (diameter 100mm) and pipe geographical environment characteristics, and there is a leakage, it corresponds to the second voiceprint feature. These features are compared and stored in the voiceprint feature database.

[0161] For example, water supply pipe B under a street in a certain city has the characteristics of the pipe itself (diameter 80mm) and the geographical environment of the pipe. When there is no leakage, it corresponds to the first voiceprint feature. Water supply pipe A has the same characteristics of the pipe itself (diameter 80mm) and the geographical environment of the pipe. When there is leakage, it corresponds to the second voiceprint feature. The features are compared and stored in the voiceprint feature database.

[0162] The water supply pipeline leakage identification device according to embodiments of this disclosure acquires a first acoustic signature feature generated when there is no leakage in the water supply pipeline under various characteristic factors in a predetermined environment, and a second acoustic signature feature generated under various water supply leakage states. It compares the acoustic signature features of any two water supply leakage states where only one factor differs, thus establishing an acoustic signature feature library for the water supply pipeline under the predetermined environment. Based on this library, it trains a model to obtain an acoustic signature recognition model. It then inputs the MEL spectrum features corresponding to the sound signal generated under the current state of the water supply pipeline in the predetermined environment to the acoustic signature recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment. This application can establish an acoustic signature feature library for various water supply leakage states and train a model to obtain an acoustic signature recognition model. Then, by inputting the MEL spectrum features corresponding to the sound signal generated under the current state of the water supply pipeline in the predetermined environment to the acoustic signature recognition model, it obtains the water supply leakage state of the water supply pipeline under the predetermined environment. This facilitates rapid identification and detection of water supply pipeline leakage states and reduces false positives and false negatives.

[0163] In some embodiments, the multiple characteristic factors in the predetermined environment include at least one of the following:

[0164] Characteristics of the pipeline itself, geographical environment of the pipeline, water supply flow rate, and water supply pressure;

[0165] The characteristics of the pipe itself include pipe diameter, pipe material, and pipe straightness / bending status;

[0166] The geographical environmental characteristics of the pipeline include the hardness of the soil layer near the pipeline and the moisture content of the soil layer near the pipeline.

[0167] The feature library establishment module 41 is used for

[0168] The acoustic signature features of water supply pipelines under the predetermined environment are compared between any two water supply leakage states where only one of the following factors differs: pipe diameter, pipe material, pipe straightness, soil hardness near the pipe, soil moisture near the pipe, water supply flow rate, and water supply pressure.

[0169] In this exemplary embodiment, there are multiple characteristic factors in the predetermined environment, including pipe characteristics, pipe geographical environment characteristics, water supply flow rate, water supply pressure, etc. The voiceprint features of the water supply pipes under two different water supply states—one without leakage and the other with multiple leakage conditions—are compared to establish a voiceprint feature database for the water supply pipes in the predetermined environment. For example, a water supply pipe C under a street in a city has pipe characteristics (50mm diameter) and pipe geographical environment characteristics. When there is no leakage, it corresponds to the first voiceprint feature. Conversely, when water supply pipe C has the same pipe characteristics (50mm diameter) and pipe geographical environment characteristics, but a leakage occurs, it corresponds to the second voiceprint feature. These features are compared and stored in the voiceprint feature database.

[0170] In some embodiments, before inputting the MEL spectral features corresponding to the sound signal generated under the current state of the water supply pipeline in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment, the leakage state prediction module 43 is used for...

[0171] Collect sound signals generated by the water supply pipeline under the current state of a predetermined environment;

[0172] A short-time Fourier transform is performed on the sound signal generated by the water supply pipeline under the current state in the predetermined environment to obtain the MEL spectral characteristics corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

[0173] In this exemplary embodiment, performing a short-time Fourier transform on the sound signal generated by the water supply pipeline under the current state in the predetermined environment to obtain the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment includes:

[0174] If the sound signal generated by the water supply pipeline under the predetermined environment is in the low-frequency region below the predetermined frequency, then the sound signal generated by the water supply pipeline under the current state under the predetermined environment is subjected to frame-by-frame windowing processing to obtain a windowed signal.

[0175] The windowed signal is subjected to a short-time Fourier transform to obtain the MEL spectral characteristics of the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

[0176] In this exemplary embodiment, a set of triangular filters based on the Mel frequency is constructed to eliminate fine structures in audio data, extract envelope information, and reduce the amount of data required for computation. The center frequencies of multiple triangular filters within the Mel scale range need to be equally spaced pairwise, resulting in an envelope formed by multiple overlapping isosceles triangles of equal size on the Mel scale, which is then mapped to the STFT linear scale. Since the traditional Mel scale is based on the human ear's ability to distinguish sound signals, it is more sensitive to changes in the low-frequency components of the signal. However, based on the spectral characteristics of data collected from the test environment and actual engineering sites, noise signals are mostly approximate white noise and low-frequency noise below 150Hz distributed across the entire frequency domain. Therefore, the original Mel scale needs to be corrected to a certain extent. The conversion formula between the corrected Mel scale and the linear scale frequency band is shown below:

[0177] The MEL spectral features after spectral correction are used as the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment; wherein...

[0178] ; To perform a short-time Fourier transform on the windowed signal, the signal frequency at the MEL scale is obtained. This represents the signal frequency under the MEL scale after spectral correction. Because this formula involves conversion between linear and exponential / logarithmic scales, the converted filter triangle windows are asymmetrical. Simultaneously, the spacing between the individual triangle filters gradually increases within 150Hz, remains constant in the 150-300Hz range, and gradually increases above 300Hz. This trend can highlight the acoustic signature characteristics near the high-frequency band of leakage signals. The formula for the triangle filter bank under the STFT linear scale is shown below: very large vector, multi-dimensional, frame energy, second derivative dynamic characteristics, static characteristics. Among them,

[0179]

[0180] In the formula, Here are the filter coefficients of the triangular window filter corresponding to frequency f on the STFT linear scale, where n is the filter sequence and f is the frequency. n-1 f n f n+1 These represent the frequencies of the left endpoint, maximum value, and right endpoint of the nth triangular window filter, respectively. The maximum filter coefficient is given. Subsequently, a triangular window filter bank is used to filter the calculated STFT power spectrum, and logarithmic operations are performed simultaneously on both ends of the filtered power spectrum to convert the multiplication operation of the short-time Fourier transform signal into an addition operation. This yields the Mel spectrum. A further discrete cosine transform (DCT) is performed on the calculated Mel spectrum to separate the signal envelope from the fine structure. The Mel cepstral coefficients (MFCCs) can then be calculated from the DCT-transformed Mel spectrum. The second form of the DCT used in this method is shown in the following equation:

[0181] ;

[0182] ;

[0183] In the formula, For Mel spectrum input, The MFCC obtained after discrete cosine transform. N The input spectrum length is... These are the compensation coefficients used for weighted calculations. On the other hand, performing first- and second-order difference calculations on MFCC yields a result that can describe the dynamic characteristics of sound signal leakage. The difference calculation formula for MFCC is shown below:

[0184]

[0185]

[0186] In the formula, and These are the first and second differences of MFCC, respectively. This refers to the time difference in the differential calculation. N denoted by the order of the Mel-frequency cepstral coefficients. Frame energy, representing the volume per unit length after the audio signal is framed, can also reflect the speaker characteristics of the audio signal. The formula for calculating frame energy is shown below:

[0187] ;

[0188] in, For the input sound signal The frame energy of the frame, n is the length of each frame after the input signal is divided into frames, and x[] is the sound signal in the time domain. At this time, the MFCC features, the first-order and second-order difference features of MFCC, and the frame energy features obtained above can be merged to form an N-dimensional MFCC voiceprint feature vector.

[0189] In this exemplary embodiment, performing a short-time Fourier transform on the windowed signal to obtain the MEL spectral characteristics corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment includes:

[0190] Perform a short-time Fourier transform on the windowed signal, and then perform spectral correction on the obtained MEL spectral features;

[0191] The MEL spectral features after spectral correction are used as the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

[0192] In some embodiments, the voiceprint recognition model includes a correspondence between voiceprint features and the state of water supply leakage in the pipeline;

[0193] The process of inputting the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment includes:

[0194] Based on the voiceprint recognition model, the MEL spectrum features corresponding to the sound signals generated by the water supply pipeline under the current state in the predetermined environment are identified, and the MEL spectrum features are compared with the voiceprint features in the voiceprint feature library to obtain the comparison results.

[0195] Based on the comparison results and the correspondence between the voiceprint features and the water supply leakage status of the pipeline, the water supply leakage status of the current state of the water supply pipeline under the predetermined environment is determined.

[0196] In this exemplary embodiment, the water supply leakage state includes both the presence of a leak and the absence of a leak. The correspondence between acoustic signature features and the water supply leakage state includes the acoustic signature features corresponding to the presence of a leak point in a water supply pipeline with multiple characteristic factors under a predetermined environment, and the acoustic signature features corresponding to the absence of a leak point in a water supply pipeline with the same characteristic factors.

[0197] The determination of the water supply leakage status of the water supply pipeline under the predetermined environment based on the comparison results and the correspondence between the voiceprint features and the water supply leakage status includes:

[0198] If the MEL spectral features are matched with the voiceprint features corresponding to the presence of a leak point in the voiceprint feature library, then the water supply leakage state of the water supply pipeline under the predetermined environment is determined to be the presence of a leak point.

[0199] If the MEL spectral features are matched with the voiceprint features corresponding to the absence of a leak point in the voiceprint feature library, then the water supply leakage state of the water supply pipeline under the predetermined environment is determined to be the absence of a leak point.

[0200] In some embodiments, the step of training a model based on the voiceprint feature database of the water supply pipeline under the predetermined environment to obtain a voiceprint recognition model includes:

[0201] Based on the aforementioned voiceprint feature database, a GMM-UBM model is trained using the Expectation-Maximization (EM) algorithm to obtain a voiceprint recognition model; wherein, the probability density function of the GMM distribution is shown in the following equation:

[0202] ;

[0203] In the formula, For input MFCC voiceprint feature vector For the probability density function of the j-th GMM model, and Let be the expectation and covariance matrices of the j-th GMM model, respectively; where is the input voiceprint feature vector. The probability of belonging to the j-th cluster can be expressed as:

[0204] ;

[0205] In the formula, Let be the weight coefficients of the j-th GMM model; where are the parameters of the GMM model. , , The EM algorithm is used for initialization and iterative updates until convergence. The updated parameters are then combined with the UBM parameters to adjust the impact of the GMM model parameters on the final prediction model. The combination formula is shown below:

[0206] ;

[0207] In the formula, , , These are the parameters of the GMM model after combining them with the UBM parameters. , , They are respectively , , The weighting coefficients of the UBM parameters, where T is the number of UBM components, and n j , , These are the weights, expectations, and covariance matrices of the UBM components, respectively. For normalization coefficients, the GMM and UBM parameters are fused to establish a GMM-UBM-based voiceprint recognition model, as shown in the following equation:

[0208] ;

[0209] In the formula, The posterior probability that the input sound signal belongs to the GMM-UBM model.

[0210] This disclosure provides a computer-readable storage medium storing a water supply pipeline leak identification program thereon. When the water supply pipeline leak identification program is executed by a processor, it implements the water supply pipeline leak identification method described in the above embodiments.

[0211] This disclosure provides an electronic device, including a memory, a processor, and a water supply pipeline leak identification program stored in the memory and executable on the processor. When the processor executes the water supply pipeline leak identification program, it implements the water supply pipeline leak identification method described in the above embodiments.

[0212] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0213] It should be understood that various parts of this disclosure can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0214] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0215] In the description of this disclosure, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this disclosure and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this disclosure.

[0216] Furthermore, the terms "first," "second," etc., used in the embodiments of this disclosure are for descriptive purposes only and should not be construed as indicating or implying relative importance, or implicitly specifying the number of technical features indicated in this embodiment. Therefore, features defined with terms such as "first" and "second" in the embodiments of this disclosure can explicitly or implicitly indicate that the embodiment includes at least one of those features. In the description of this disclosure, the word "multiple" means at least two or more, such as two, three, four, etc., unless otherwise explicitly specified in the embodiments.

[0217] In this disclosure, unless otherwise explicitly specified or limited in the embodiments, the terms "installation," "connection," "joining," and "fixing," etc., appearing in the embodiments should be interpreted broadly. For example, a connection can be a fixed connection, a detachable connection, or an integral part; it can also be a mechanical connection, an electrical connection, etc. Of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication between two components, or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this disclosure based on the specific implementation.

[0218] In this disclosure, unless otherwise expressly specified and limited, "above" or "below" the second feature can mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Furthermore, "above," "on top of," and "over" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0219] Although embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present disclosure.

Claims

1. A method for identifying leaks in water supply pipelines, characterized in that, include: Acquire the first acoustic signature feature generated when there is no leakage in the water supply pipeline under various characteristic factors in a predetermined environment, and the second acoustic signature feature generated under various water supply leakage conditions; The acoustic signature features of the water supply pipeline under the predetermined environment are compared between two water supply states with the same characteristic factors, namely, no leakage and multiple water supply leakage states. Based on the voiceprint feature library of water supply pipelines under the predetermined environment, a model is trained to obtain a voiceprint recognition model. Input the MEL spectrum features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the current state in the predetermined environment. The step of performing a short-time Fourier transform on the sound signal generated by the water supply pipeline under the current state in the predetermined environment to obtain the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment includes: If the sound signal generated by the water supply pipeline under the predetermined environment is in the low-frequency region below the predetermined frequency, then the sound signal generated by the water supply pipeline under the current state under the predetermined environment is subjected to frame-by-frame windowing processing to obtain a windowed signal. Perform a short-time Fourier transform on the windowed signal to obtain the MEL spectral characteristics of the sound signal generated by the water supply pipeline under the current state in the predetermined environment; The step of performing a short-time Fourier transform on the windowed signal to obtain the MEL spectral characteristics corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment includes: Perform a short-time Fourier transform on the windowed signal, and then perform spectral correction on the obtained MEL spectral features; The MEL spectral features after spectral correction are used as the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment; wherein... ; To perform a short-time Fourier transform on the windowed signal, the signal frequency at the MEL scale is obtained. This refers to the signal frequency on the MEL scale after spectral correction.

2. The water supply pipeline leakage identification method according to claim 1, characterized in that, The predetermined environment includes at least one of the following characteristics: Characteristics of the pipeline itself, geographical environment of the pipeline, water supply flow rate, and water supply pressure; The characteristics of the pipe itself include pipe diameter, pipe material, and pipe straightness / bending status; The geographical environmental characteristics of the pipeline include the hardness of the soil layer near the pipeline and the moisture content of the soil layer near the pipeline. The comparison of acoustic signature features under two water supply conditions—one with no leakage and the other with multiple leakage conditions—that share the same characteristic factors, to establish an acoustic signature feature library for the water supply pipeline under the predetermined environment includes: The acoustic signature features of water supply pipelines under the predetermined environment are compared between any two water supply leakage states where only one of the following factors differs: pipe diameter, pipe material, pipe straightness, soil hardness near the pipe, soil moisture near the pipe, water supply flow rate, and water supply pressure.

3. The water supply pipeline leakage identification method according to claim 1, characterized in that, Before inputting the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment, the method includes: Collect sound signals generated by the water supply pipeline under the current state of a predetermined environment; A short-time Fourier transform is performed on the sound signal generated by the water supply pipeline under the current state in the predetermined environment to obtain the MEL spectral characteristics corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment.

4. The water supply pipeline leakage identification method according to claim 1, characterized in that, The voiceprint recognition model includes a correspondence between voiceprint features and the state of water supply leakage in the pipeline. The process of inputting the MEL spectral features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment includes: Based on the voiceprint recognition model, the MEL spectrum features corresponding to the sound signals generated by the water supply pipeline under the current state in the predetermined environment are identified, and the MEL spectrum features are compared with the voiceprint features in the voiceprint feature library to obtain the comparison results. Based on the comparison results and the correspondence between the voiceprint features and the water supply leakage status of the pipeline, the water supply leakage status of the current state of the water supply pipeline under the predetermined environment is determined.

5. The water supply pipeline leakage identification method according to claim 1, characterized in that, The process of training a model based on the voiceprint feature database of the water supply pipeline under the predetermined environment to obtain a voiceprint recognition model includes: Based on the aforementioned voiceprint feature database, a GMM-UBM model is trained using the expectation-maximization algorithm to obtain a voiceprint recognition model; wherein, the probability density function of the GMM distribution is shown in the following equation: ； In the formula, For input MFCC voiceprint feature vector For the probability density function of the j-th GMM model, and Let be the expectation and covariance matrices of the j-th GMM model, respectively; where is the input voiceprint feature vector. The probability of belonging to the j-th cluster can be expressed as: ； In the formula, Let be the weight coefficients of the j-th GMM model; where are the parameters of the GMM model. , , The EM algorithm is used for initialization and iterative updates until convergence. The updated parameters are then combined with the UBM parameters to adjust the impact of the GMM model parameters on the final prediction model. The combination formula is shown below: In the formula, , , These are the parameters of the GMM model after combining them with the UBM parameters. , , They are respectively , , The weighting coefficients of the UBM parameters, where T is the number of UBM components, and n j , , These are the weights, expectations, and covariance matrices of the UBM components, respectively. For normalization coefficients, the GMM and UBM parameters are fused to establish a GMM-UBM-based voiceprint recognition model, as shown in the following equation: ； In the formula, This represents the posterior probability that the input sound signal belongs to the GMM-UBM model. equal The weight coefficients corresponding to the GMM-UBM model, equal The expected value corresponding to the GMM-UBM model, equal This represents the variance corresponding to the GMM-UBM model.

6. A water supply pipeline leak detection device, suitable for control using the water supply pipeline leak detection method according to any one of claims 1-5, characterized in that, include: The voiceprint feature acquisition module is used to acquire the first voiceprint feature generated when there is no leakage in the water supply pipeline under various characteristic factors in a predetermined environment, and the second voiceprint feature generated under various water supply leakage conditions. The feature library establishment module is used to compare the voiceprint features of the water supply pipeline under two water supply states with the same characteristic factors in the absence of leakage and multiple water supply leakage states, and to establish the voiceprint feature library of the water supply pipeline under the predetermined environment. The model training module is used to train the model based on the voiceprint feature library of the water supply pipeline under the predetermined environment to obtain the voiceprint recognition model. The leakage state prediction module is used to input the MEL spectrum features corresponding to the sound signal generated by the water supply pipeline under the current state in the predetermined environment into the voiceprint recognition model to obtain the water supply leakage state of the water supply pipeline under the predetermined environment.

7. A computer-readable storage medium, characterized in that, It stores a water supply pipeline leak detection program, which, when executed by the processor, implements the water supply pipeline leak detection method according to any one of claims 1-5.

8. An electronic device, characterized in that, The system includes a memory, a processor, and a water supply pipeline leak identification program stored in the memory and executable on the processor. When the processor executes the water supply pipeline leak identification program, it implements the water supply pipeline leak identification method according to any one of claims 1-5.

Images