Audio segment extraction method, device, equipment and storage medium

Abstract

The present disclosure provides an audio segment extraction method, device and equipment and a storage medium, and belongs to the technical field of acoustic detection. The technical scheme provided by the present disclosure converts an audio signal into an audio feature sequence, slides a first window on the sequence, extracts a feature point with the maximum energy in each window as a first feature point to lock the potential wind-swept sound position of each window, then screens out points located in the same audio segment as second feature points through the first feature points of adjacent windows, and finally extracts a wind-swept sound audio segment based on the second feature points. Since there is a stable difference in the time sequence characteristics between the wind-swept sound signal and the noise signal, the second feature points are determined by limiting the condition that the first feature points of adjacent first windows must be located in the same audio segment, the stability characteristics of the wind-swept sound signal in the time sequence are utilized to filter out the pseudo-peak values of random noise signals, and thus the audio segment of the blade wind-swept sound is effectively extracted.

Description

Technical Field

[0001] This disclosure relates to the field of acoustic detection technology, and in particular to a method, apparatus, device, and storage medium for extracting audio segments. Background Technology

[0002] In the field of wind power generation, using non-contact acoustic detection technology to detect the sweeping sound of wind turbine blades is an important means of assessing blade condition and ensuring the safe operation of wind turbines. Therefore, extracting the audio segments of the sweeping sound is crucial. In actual wind turbine operating environments, interference signals such as wind noise and ambient noise last for a long time, and these interference signals account for a high proportion of the sweeping sound audio signal. They are prone to generating spurious peaks in the sweeping sound audio signal, which disrupt the sweeping sound audio signal and make it difficult to extract the sweeping sound audio segments. Summary of the Invention

[0003] This disclosure provides a method, apparatus, device, and storage medium for extracting audio segments, so as to effectively extract audio segments of blade sweeping sounds from audio signals. The technical solution of this disclosure is as follows: According to one aspect of the embodiments of this disclosure, a method for extracting audio segments is provided, comprising: Based on the audio signals collected during the operation of the wind turbine, an audio feature sequence of the audio signals is obtained, and the audio feature sequence is used to characterize the energy of the target frequency band in each audio segment of the audio signal; Slide a first window on the audio feature sequence. For each first window, determine a first feature point of the first window. The first feature point is the feature point with the highest energy in the audio feature subsequence within the first window. Among multiple first feature points in adjacent first windows, multiple second feature points are determined, wherein the second feature points are first feature points located in the same audio segment; Based on multiple second feature points, audio segments of blade sweeping sound are extracted from the audio signal.

[0004] According to another aspect of the present disclosure, an audio segment extraction apparatus is provided, comprising: The feature sequence determination module is configured to acquire an audio feature sequence of an audio signal based on an audio signal collected during the operation of the wind turbine. The audio feature sequence is used to characterize the energy of the target frequency band in each audio segment of the audio signal. The first feature point determination module is configured to slide a first window on the audio feature sequence, and for each first window, determine a first feature point of the first window, wherein the first feature point is the feature point with the highest energy in the audio feature subsequence within the first window; The second feature point determination module is configured to determine multiple second feature points among multiple first feature points in adjacent first windows, wherein the second feature points are first feature points located in the same audio segment; The audio segment extraction module is configured to extract audio segments of blade sweeping sound from the audio signal based on multiple second feature points.

[0005] In some embodiments, the device further includes: The window parameter setting module is configured to: determine the window length of the first window based on a first time, where the first time is the time required for two adjacent blades to pass the same position; The sliding step size of the first window is determined based on the second time, which is the time when the sweeping sound is collected once when the blade rotates.

[0006] In some embodiments, the device further includes: The audio segmentation module is configured to divide the audio signal into multiple audio segments based on a time granularity, the time granularity being set based on the rotation parameters of the blade.

[0007] In some embodiments, the device further includes: The first filtering module is configured to slide a second window on the audio feature sequence, wherein the properties, window length and sliding step size of the second window are different from those of the first window. For each second window, an energy threshold for the second window is determined based on the second feature points in the audio feature subsequence within the second window. The energy threshold is used to characterize the energy limit of the sweeping sound in the audio signal. For each second window, filter out second feature points whose energy is greater than the energy threshold within the second window; The audio segment extraction module is configured to extract audio segments of blade sweeping sounds from the audio signal based on multiple filtered second feature points.

[0008] In some embodiments, the first filtering module is configured to: for each second window, determine a reference energy value in the second window based on a second feature point in the audio feature subsequence within the second window, wherein the reference energy value is used to characterize the energy reference value of the sweeping sound in the second window; For each second window, an energy threshold is determined based on the reference energy value and energy ratio coefficient of the second window, wherein the energy ratio coefficient is used to characterize the degree of energy deviation of the sweeping sound relative to the energy reference value.

[0009] In some embodiments, the first filtering module is configured to: for each second window, if the second feature point in the audio feature subsequence within the second window is not empty, determine the reference energy value in the second window based on the second feature point in the second window; For each second window, if the second feature point in the audio feature subsequence within the second window is empty, the baseline energy value in the second window is determined based on the feature points in the second window.

[0010] In some embodiments, the first filtering module is configured to: for each second window, mark feature points in the audio feature subsequence within the second window whose energy is greater than the energy threshold as a first state, and mark feature points whose energy is less than or equal to the energy threshold as a second state; Based on the labeled states of feature points in multiple second windows, a state labeling matrix for the audio feature sequence is constructed. Based on the state label matrix, the second feature points marked as the first state are filtered out from each second window.

[0011] In some embodiments, the device further includes: The second filtering module is configured to determine multiple third feature points based on the audio segments to which the multiple second feature points belong after filtering. The third feature points are second feature points that satisfy a preset index condition with the audio segments to which the adjacent second feature points are located. The preset index condition is related to the time required for two adjacent leaves to pass through the same position. The audio segment extraction module is configured to extract audio segments of blade sweeping sound from the audio signal based on multiple third feature points.

[0012] In some embodiments, the audio segment extraction module is configured to extract segments within a preset time period around the second feature point in the audio signal, centered on the second feature point, to obtain multiple audio segments of the blade sweeping sound.

[0013] According to another aspect of the embodiments of the present disclosure, a computer device is provided, the computer device comprising: One or more processors; Memory used to store the executable program code of the processor; The processor is configured to execute the program code to implement the above-mentioned method for extracting audio segments.

[0014] According to another aspect of the present disclosure, a computer-readable storage medium is provided, which, when the program code in the computer-readable storage medium is executed by the processor of a computer device, enables the computer device to perform the above-described audio segment extraction method.

[0015] According to another aspect of the present disclosure, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the above-described audio segment extraction method.

[0016] The technical solution provided in this disclosure converts an audio signal into an audio feature sequence, slides a first window on the sequence, and extracts the feature point with the highest energy in each window as the first feature point to lock the potential sweeping sound location in each window. Then, using the first feature points of adjacent windows, points located in the same audio segment are selected as second feature points. Finally, the sweeping sound audio segment is extracted from the audio signal based on the second feature points. Since the sweeping sound signal and the noise signal have different stability in temporal characteristics, by limiting the first feature points of adjacent first windows to be located in the same audio segment to determine the second feature point, the stability characteristics of the sweeping sound signal in temporal sequence can be utilized to filter out the pseudo-peaks of random noise signals, thereby effectively extracting the audio segment of blade sweeping sound.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0019] Figure 1 This is a schematic diagram illustrating an audio segment extraction system according to an exemplary embodiment.

[0020] Figure 2 This is a flowchart illustrating an audio segment extraction method according to an exemplary embodiment.

[0021] Figure 3 This is an audio signal and its time-frequency waveform in a wind turbine environment, as illustrated in an exemplary embodiment.

[0022] Figure 4 This is a schematic diagram illustrating a sweeping sound audio clip according to an exemplary embodiment.

[0023] Figure 5 This is a flowchart illustrating another method for extracting audio segments according to an exemplary embodiment.

[0024] Figure 6 This is a flowchart illustrating the overall implementation of an audio segment extraction method according to an exemplary embodiment.

[0025] Figure 7This is an original audio signal and its voiceprint time-frequency diagram shown in a wind turbine environment according to an exemplary embodiment.

[0026] Figure 8 This is a processed audio signal and its voiceprint time-frequency diagram illustrated according to an exemplary embodiment.

[0027] Figure 9 This is a block diagram illustrating an audio segment extraction device according to an exemplary embodiment.

[0028] Figure 10 This is a block diagram illustrating a terminal according to an exemplary embodiment.

[0029] Figure 11 This is a schematic diagram of the structure of a server according to an exemplary embodiment. Detailed Implementation

[0030] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0031] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0032] It should be noted that the information (including but not limited to wind turbine equipment information), data (including but not limited to data used for analysis, data stored, data displayed), and signals (including but not limited to audio signals) involved in this disclosure are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0033] Figure 1 This is a schematic diagram illustrating an audio segment extraction system according to an exemplary embodiment. See also: Figure 1The audio segment extraction system includes: a wind turbine 101, a voiceprint acquisition terminal 102, a server 103, and a network 104. The wind turbine 101, as the detection object of the entire system, is used to rotate its blades to generate electricity under wind power. During operation, the rotation of its blades and friction with the air generate sweeping noise, accompanied by wind noise, mechanical vibration, and ambient noise. These mixed audio signals are the raw input to be processed in this embodiment.

[0034] The voiceprint acquisition terminal 102 is deployed at a fixed location on or near the tower base of the wind turbine 101 (as shown in the figure) to collect audio signals from the wind turbine's operating environment in real time in a non-contact manner. For example, the voiceprint acquisition terminal 102 can be a high-sensitivity microphone array.

[0035] Server 103, as the core processing unit of this system, is responsible for executing the complete audio segment extraction process. The audio segments extracted by server 103 can be directly used for subsequent identification of blade icing status, or the extraction results can be sent back to the acoustic signature acquisition terminal 102 or the wind farm monitoring center for further processing.

[0036] Network 104 serves as a medium for providing a communication link between the voiceprint acquisition terminal 102 and the server 103. Network 104 may include various types of wired fiber optic or wireless communication links, such as Bluetooth, Wi-Fi, or microwave communication links.

[0037] As a core pillar of my country's energy transition, the safe operation of wind power is of paramount importance. However, in cold or high-humidity environments, icing on wind turbine blades significantly increases blade weight, disrupts aerodynamic shape, and leads to mechanical load imbalance, thereby exacerbating the risk of bearing overload and blade breakage, and even causing safety accidents such as tower collapse. Detecting blade sweeping sound is an effective means of determining icing status. Currently, contact detection technologies such as thermocouples and capacitive sensors are mostly used. This method requires installing sensors on the blade surface, which not only damages the blade structure and increases aerodynamic drag, but also limits the detection range to a single point, making it difficult to cover the entire area. Furthermore, it is susceptible to electromagnetic and vibration interference, resulting in poor data accuracy. Although non-contact acoustic detection technology has the advantages of wide-area coverage and anti-interference, in actual wind farms, due to the presence of high-intensity wind noise and environmental background noise, blade sweeping sound is still difficult to effectively extract from the complex audio signals, seriously affecting the accuracy of subsequent icing status identification.

[0038] Based on this, in order to effectively extract sweeping sound audio segments in complex environments, this disclosure proposes an audio segment extraction method. The basic idea is as follows: First, an audio feature sequence is constructed based on the original audio signal. Then, the first feature point with the highest energy in each window is extracted through a first window to lock the potential sweeping sound location in each window. The stability of the first feature point selected above is verified by the positional consistency between adjacent windows, and second feature points are further selected to suppress false peaks caused by local noise. Finally, based on these second feature points, the pure sweeping sound audio segments are accurately located and extracted from the audio signal, providing high-quality data support for the accurate identification of the subsequent blade icing state.

[0039] The technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Figure 2 This is a flowchart illustrating an audio segment extraction method according to an exemplary embodiment, see [link to flowchart]. Figure 2 The method is executed by a server and includes the following steps: In step 201, based on the audio signals collected during the operation of the wind turbine, an audio feature sequence of the audio signal is obtained. The audio feature sequence is used to characterize the energy of the target frequency band in each audio segment of the audio signal.

[0040] In actual wind farm environments, the audio signals collected by the acoustic signature acquisition terminal are mixed signals from multiple sound sources, including blade sweeping noise, wind noise, ambient noise, and mechanical vibration. These sound sources are superimposed on each other in the time domain. If the original audio signal is processed directly, it is difficult to effectively distinguish between sweeping noise and noise signals. Since sweeping noise and noise signals have different energy distribution characteristics in the frequency domain, the energy of sweeping noise signals is usually more concentrated in the high-frequency band, while the energy of noise signals is mainly distributed in other lower frequency bands. Based on this, the embodiments of this disclosure first generate an audio feature sequence based on the original audio signal. This audio feature sequence is used to characterize the energy distribution of each audio segment in the audio signal in the target frequency band, thereby converting the audio signal from a time-domain waveform representation into a one-dimensional feature representation with energy as the core.

[0041] In some embodiments, the temporal granularity of the audio signal needs to be determined before generating an audio feature sequence based on the original audio signal. Since the sweeping sound is a periodic event directly related to blade rotation, its occurrence time interval is closely related to rotational parameters such as blade rotation speed. If the temporal granularity of the audio segments is unrelated to the blade rotational parameters, the duration of each audio segment may be too short to characterize the energy features of the sweeping sound signal, or too long, resulting in excessive irrelevant noise. Therefore, in this embodiment, the temporal granularity of the audio segments is set based on the blade rotational parameters to ensure that the duration of each audio segment appropriately covers the key features of the sweeping sound signal, providing a reasonable temporal resolution for subsequent feature point selection.

[0042] Specifically, such as Figure 1 As shown, in this relative installation position of the acoustic signature acquisition terminal and the wind turbine, the acoustic signature acquisition terminal can only effectively capture the sweeping sound generated by the friction between the blades and the air when the wind turbine blades rotate to an angle less than a certain critical angle θ with the vertical direction where the acoustic signature acquisition terminal is located. When the angle between the blades and the vertical direction is greater than the critical angle θ, the signal collected by the terminal is mainly noise from the wind farm environment. This means that the sweeping sound will only be collected when the blades move to a specific position, and the time interval between the occurrence of the sweeping sound is directly related to the rotation parameters such as the rotation speed of the blades. Therefore, the time granularity of the audio segment needs to be correlated with the blade rotation parameters.

[0043] Furthermore, Figure 3 This is an audio signal and its time-frequency waveform diagram under a wind turbine environment, illustrated according to an exemplary embodiment. Figure 3 As shown, the upper graph is the waveform of the original audio signal, with time on the horizontal axis and decibels on the vertical axis, reflecting the change trend of the audio signal amplitude over time. The lower graph is the time-frequency graph of the sound pattern, with time on the horizontal axis and frequency on the vertical axis, and the color intensity indicating the energy of the signal at the corresponding frequency. It can be clearly seen from the lower graph that when the blade is within the angular range of 2θ from which the sweeping sound can be collected, vertical bright bars, as shown in the rectangular area, appear in the time-frequency graph. The time period corresponding to these bright bars is the period when the blade sweeping sound occurs. At other times, such vertical bright bars do not appear in the time-frequency graph, indicating that the signal collected at this time is environmental noise. Therefore, the sweeping sound has a clear energy concentration characteristic in the time-frequency graph, further proving that the time interval of its occurrence is determined by the rotation parameters of the blade.

[0044] In step 202, a first window is slid across the audio feature sequence. For each first window, a first feature point is determined. The first feature point is the feature point with the highest energy in the audio feature subsequence within the first window.

[0045] In this embodiment of the disclosure, Figure 4This is a schematic diagram illustrating a sweeping sound audio clip according to an exemplary embodiment. Figure 4 As shown in the figure, this diagram illustrates the distribution of energy values ​​over time in each audio segment after the aforementioned frequency domain analysis of the audio signal. The figure reveals multiple maxima in the audio feature sequence. These maxima include both true energy peaks generated by the blade sweeping sound signal and pseudo-energy peaks generated by noise signals. Since the sweeping sound signal often exhibits strong energy, the point with the highest energy value within each local region is most likely the sweeping sound signal in that region. By analyzing the distribution and energy characteristics of these maxima on the time axis, the sweeping sound signal and noise signal can be distinguished.

[0046] Based on this, after constructing the audio feature sequence, in order to initially locate the potential sweeping sound signal position from the sequence, the embodiments of this disclosure employ a feature point screening mechanism based on a sliding window. Specifically, a first window is used to slide successively on the audio feature sequence. For each area covered by the first window, it is regarded as an independent analysis area, and the energy values ​​of all feature points in the area are compared. Feature points whose energy values ​​reach a local maximum are identified, and the feature point with the largest value is marked as the first feature point (i.e., the window peak point) corresponding to that window.

[0047] In step 203, among the multiple first feature points in adjacent first windows, multiple second feature points are determined, where the second feature points are the first feature points located in the same audio segment.

[0048] In this embodiment, after obtaining the first feature points corresponding to each window by sliding the first window, these first feature points inevitably include spurious peaks caused by noise signals. If the extraction of sweeping sound segments is directly based on these first feature points, the accuracy of segment extraction will be affected. Considering that the sweeping sound signal is a periodic event synchronized with the blade rotation parameters, its position within adjacent first windows should remain stable, while the noise signal is random, and the spurious peaks it generates often have large deviations in position within adjacent first windows. Based on this, this embodiment utilizes the essential difference between the stability of the sweeping sound signal and the randomness of the noise signal, and performs cross-window stability verification on the first feature points based on the audio segments to which the first feature points belong, thereby filtering out spurious peaks and retaining the first feature points that truly correspond to the sweeping sound signal.

[0049] In step 204, based on multiple second feature points, an audio segment of the blade sweeping sound is extracted from the audio signal.

[0050] In this embodiment, to ensure that the extracted audio segments completely cover the blade sweeping sound signal while avoiding the introduction of excessive background noise, it is necessary to perform precise segment extraction from the original audio signal based on the verified feature points of the sweeping sound signal. Since the sweeping sound signal has a certain duration on the time axis, and its energy peaks are usually concentrated in the middle region of the signal, the prior feature of the concentrated and symmetrical distribution of the energy peaks of the sweeping sound signal can be utilized. Taking the aforementioned confirmed feature points of the sweeping sound signal as the center, a preset duration is extended to both the front and back sides to construct a symmetrical time window. The audio segments of the blade sweeping sound are then extracted based on this time window. This can completely encompass the main energy distribution of the sweeping sound signal and ensure the integrity of the extracted segments.

[0051] The technical solution provided in this disclosure converts an audio signal into an audio feature sequence, slides a first window on the sequence, and extracts the feature point with the highest energy in each window as the first feature point to lock the potential sweeping sound location in each window. Then, using the first feature points of adjacent windows, points located in the same audio segment are selected as second feature points. Finally, the sweeping sound audio segment is extracted from the audio signal based on the second feature points. Since the sweeping sound signal and the noise signal have different stability in temporal characteristics, by limiting the first feature points of adjacent first windows to be located in the same audio segment to determine the second feature point, the stability characteristics of the sweeping sound signal in temporal sequence can be utilized to filter out the pseudo-peaks of random noise signals, thereby effectively extracting the audio segment of blade sweeping sound.

[0052] The technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Figure 5 This is a flowchart illustrating another method for extracting audio segments according to an exemplary embodiment; see also... Figure 5 The method is executed by a server and includes the following steps: In step 501, the audio signals collected during the operation of the wind turbine are divided into multiple audio segments based on the time granularity, which is set based on the rotation parameters of the blades.

[0053] In the embodiments of this application, Figure 6 This is a flowchart illustrating the overall implementation of an audio segment extraction method according to an exemplary embodiment, such as... Figure 6 As shown in step 1, based on parameters such as blade rotation speed recorded by the Supervisory Control and Data Acquisition (SCADA) system, the path of a blade is determined. Figure 1 The time required for the mid-angle range of 2θ is determined, and the time granularity is then set to be slightly smaller than this duration to ensure that the segmented audio segments can accurately cover the key processes that generate the sweeping sound.

[0054] For example, if the time required for a single blade to pass through the angle range 2θ below the acoustic signature acquisition terminal is determined to be approximately 0.3 seconds based on the rotation parameters of the blade, then the time granularity can be set to 0.1 seconds, that is, the audio signal is divided into several consecutive audio segments at equal intervals of 0.1 seconds each.

[0055] In step 502, based on the audio signal, an audio feature sequence of the audio signal is obtained. The audio feature sequence is used to characterize the energy of the target frequency band in each audio segment of the audio signal.

[0056] In this embodiment of the application, after dividing the audio signal into multiple audio segments according to time granularity, it is necessary to further extract the values ​​of effective energy features from each audio segment to construct an audio feature sequence. Specifically, as follows... Figure 3 As shown in the figure below, in the frequency range below 2000Hz, it is difficult to effectively distinguish between the sweeping sound signal and the noise signal; while in the frequency range above 2000Hz, the noise signal is significantly weakened, and the sweeping sound signal is more prominent. This indicates that the energy of the sweeping sound is mainly concentrated in the high-frequency band, while the energy of the noise signal is distributed in the lower frequency band.

[0057] Based on this, frequency domain analysis (e.g., power spectral density calculation) is performed on each audio segment to extract its energy value in the target frequency band (e.g., ≥2000Hz), thus achieving feature conversion from time-domain waveform to frequency-domain energy. Arranging the energy values ​​corresponding to all audio segments in chronological order yields the audio feature sequence. Where N is the total number of audio segments, Let be the energy value of the i-th audio segment. This audio feature sequence is a one-dimensional sequence used to characterize the energy distribution of each audio segment in the target frequency band.

[0058] In step 503, the window length of the first window is determined based on a first time, which is the time required for two adjacent blades to pass the same position; the sliding step size of the first window is determined based on a second time, which is the time when the sweeping sound is collected once when the blade rotates.

[0059] In this embodiment, since the occurrence time of the sweeping sound is closely related to the rotation parameters of the blade, if the window length of the first window does not match the blade rotation period, the first window may not be able to completely cover one sweeping sound signal, resulting in missed detection; if the sliding step size of the first window is not set properly, the sampling interval between adjacent windows may be too large, resulting in missed sweeping sound signals.

[0060] Specifically, such as Figure 1As shown, the time required for two adjacent blades below to pass the same position of the acoustic signature acquisition terminal constitutes a complete cycle of the sweeping sound signal. Determining the window length based on this cycle ensures that the coverage of the first window is consistent with the periodic characteristics of the sweeping sound. The time it takes for the sweeping sound to be acquired when the blade rotates (i.e., the time it takes for the blade to pass through the lower angle range 2θ) reflects the duration of a single sweeping sound signal. Determining the sliding step size based on this duration ensures that the first window will not miss any time period that may contain the sweeping sound signal during the sliding process.

[0061] In some embodiments, such as Figure 6 As shown in step 2, the time required for two adjacent blades to pass the same position is first obtained, and this time is used as the window length of the first window; simultaneously, the time it takes for the sweeping sound to be collected during the blade rotation is obtained, and this time is used as the sliding step size of the first window. Through this parameter configuration, the first window can slide regularly across the audio feature sequence, thereby dividing multiple continuous and partially overlapping analysis regions within each cycle.

[0062] Taking a specific implementation scenario as an example, assume that the time required to determine the same position of two adjacent blades passing the acoustic signature acquisition terminal is 1.2 seconds, and the time for acquiring a sweeping sound when the blade rotates is 0.3 seconds; and the audio signal is divided into time granularities of 0.1 seconds. Based on this, the window length of the first window is set to 1.2 seconds, and the sliding step size is set to 0.3 seconds. Under this parameter configuration, the first window performs an analysis every 0.3 seconds (i.e., 3 audio segments) on the audio feature sequence, and each analysis covers a time range of 1.2 seconds (i.e., 12 audio segments). Thus, within a complete 1.2-second cycle, four different first windows are formed, and the first feature point with the highest energy is selected for each of these four windows. This method of analyzing the audio feature sequence successively through sliding windows ensures that every time period that may generate a sweeping sound signal is fully covered, and the accuracy of the sweeping sound signal localization is enhanced through overlap analysis.

[0063] In step 504, a first window is slid across the audio feature sequence. For each first window, a first feature point is determined. The first feature point is the feature point with the highest energy in the audio feature subsequence within the first window.

[0064] In this embodiment, after setting the parameters of the first window, the first window is used to slide sequentially on the audio feature sequence. For each area covered by the first window, it is regarded as an independent analysis area, and the energy values ​​of all feature points in the area are compared. Feature points whose energy values ​​reach local maxima are identified, and the feature point with the largest value is marked as the first feature point corresponding to the window.

[0065] In step 505, among the multiple first feature points in adjacent first windows, multiple second feature points are determined, where the second feature points are the first feature points located in the same audio segment.

[0066] In this embodiment, after obtaining the first feature points corresponding to each of the first windows, a stability verification stage is entered. Specifically, the first feature points of adjacent first windows are compared one by one to see if they fall within the same audio segment. If the first feature points of two (or more) adjacent first windows are all located in the same audio segment, it means that the position indices of these first feature points are the same. Therefore, it is determined that the first feature point at that audio segment is generated by the sweeping sound signal, and this first feature point is then identified as a second feature point. Finally, all the second feature points that pass the stability verification are aggregated to form a second feature point set M = [m1, m2, ..., m...]. k ], where m k This is the index of the audio segment corresponding to the kth second feature point.

[0067] In step 506, a second window is slid across the audio feature sequence. The properties, window length, and sliding step size of the second window are different from those of the first window.

[0068] In this embodiment, the stability verification described above can filter out most of the spurious peaks caused by noise signals. However, considering that some noise signals may fall within the same audio segment in adjacent first windows, relying solely on signal position for stability verification still carries the risk of misjudgment. Since the energy value of the sweeping sound signal should remain relatively stable, while the energy value of the noise signal may change abruptly, if the energy value of a certain second feature point deviates significantly from the energy range of the sweeping sound signal, it may still be a spurious peak. Furthermore, if a globally fixed threshold is used to characterize the energy range of the sweeping sound signal, it is difficult to adapt to the dynamic changes in the energy level of the sweeping sound signal at different times. And since the second feature point is a high-confidence feature point retained after stability verification, its energy value can more accurately reflect the overall energy level of the sweeping sound signal. Therefore, the energy range of the sweeping sound signal can be adaptively determined based on the energy value of the second feature point at different times.

[0069] Based on this, such as Figure 6As shown in step 3, this embodiment further introduces a second window that slides across the audio feature sequence to perform secondary screening of the second feature points from an energy dimension. Specifically, for each second window, an energy threshold is determined based on the energy values ​​of the second feature points contained within that second window. This energy threshold characterizes the energy limit of the sweeping sound in the audio signal; that is, the energy of the sweeping sound signal should not exceed this threshold. Based on this, the energy of the second feature points within the second window is compared, and second feature points with energy greater than the energy threshold are identified as pseudo-peaks and filtered out.

[0070] Taking the specific implementation scenario mentioned above as an example, assume that the time required to determine the same position of two adjacent blades passing the acoustic signature acquisition terminal is 1.2 seconds, and the time to collect a sweeping sound when the blades rotate is 0.3 seconds; the audio signal is divided into time granularities of 0.1 seconds; based on this, the window length of the first window is determined to be 1.2 seconds, and the sliding step size is 0.3 seconds. Based on this, the window length of the second window can be set to 30 audio segments, i.e., 3.0 seconds, and the sliding step size can be set to 10 audio segments, i.e., 1.0 seconds. Sliding along the audio feature sequence with a window length of 3.0 seconds and a sliding step size of 1.0 seconds, each second window covers 30 consecutive audio segments, and adjacent second windows are spaced 10 audio segments apart.

[0071] It should be noted that the window length and sliding step size of the second window can be flexibly adjusted according to the actual wind field environment and the data in the audio feature sequence. For example, when the wind speed is high or the energy values ​​of each audio segment in the audio feature sequence differ significantly, the sliding step size and window length can be appropriately reduced to improve the temporal resolution.

[0072] In step 507, for each second window, a reference energy value is determined based on the second feature point in the audio feature subsequence within the second window. The reference energy value is used to characterize the energy reference value of the sweeping sound in the second window.

[0073] In this embodiment of the application, since the second feature point is a high-confidence feature point retained after stability verification, its energy value can better reflect the overall energy level of the sweeping sound signal in the current environment. Therefore, a reference benchmark for measuring the sweeping sound energy level in the current second window can be determined based on the second feature point contained in each second window. That is, a benchmark energy value is determined based on the second feature point contained in the second window.

[0074] In some embodiments, since not every second window can exactly cover the valid second feature points, skipping the energy threshold calculation for second windows that do not contain second feature points may affect the feature point selection within adjacent second windows. Therefore, as Figure 6As shown in step 3, to ensure the continuity and completeness of the overall feature point selection, for each second window, if the second feature point in the audio feature subsequence within the second window is not empty, the baseline energy value in the second window is determined based on the second feature point in the second window; for each second window, if the second feature point in the audio feature subsequence within the second window is empty, the baseline energy value in the second window is determined based on the feature point in the second window. This allows for the adaptive calculation of the energy threshold by prioritizing the use of these high-confidence second feature points when they exist in the second window, and by reflecting the energy level of the current window based on the initial feature points within the window when no second feature point exists, thus achieving full-time, adaptive energy threshold construction.

[0075] In step 508, for each second window, the energy threshold of the second window is determined based on the reference energy value and energy ratio coefficient of the second window. The energy ratio coefficient is used to characterize the degree of energy deviation of the sweeping sound relative to the energy reference value.

[0076] In this embodiment of the application, based on the aforementioned reference energy value, considering that although the real sweeping sound signal may have energy fluctuations, its deviation from the reference reference is limited, while the pseudo peaks caused by noise signals often show drastic energy changes, the maximum allowable energy deviation of the sweeping sound signal relative to the reference reference can be preset, thereby determining the energy threshold in each second window.

[0077] Specifically, the baseline energy value is adjusted by combining a preset energy ratio coefficient (e.g., within the standard deviation range or a preset multiple range) to determine the energy threshold of the second window. In this way, the energy threshold of each second window is adaptively determined based on the energy of the second feature points it contains, rather than using a globally uniform fixed threshold, thus improving the flexibility of feature point selection.

[0078] In step 509, for each second window, feature points with energy greater than the energy threshold in the audio feature subsequence within the second window are marked as first states, and feature points with energy less than or equal to the energy threshold are marked as second states.

[0079] In this embodiment, to transform the complex energy comparison process into efficient logical state operations and improve the operational efficiency of the feature point selection process, a feature point selection mechanism based on a state label matrix is ​​introduced. Specifically, for each second window, all feature points within that second window whose energy is greater than the energy threshold are marked as a first state (e.g., marked as "abnormal"), meaning that the energy value of the feature point exceeds the energy limit of the sweeping sound signal; feature points whose energy is less than or equal to the energy threshold are marked as a second state (e.g., marked as "normal"), meaning that the energy value of the feature point is within the reasonable range of the sweeping sound signal.

[0080] In step 510, a state labeling matrix for the audio feature sequence is constructed based on the labeling states of feature points in multiple second windows; based on the state labeling matrix, second feature points labeled as first states are filtered out from each second window.

[0081] In the embodiments of this application, such as Figure 6 As shown in step 3, after completing the state labeling of all feature points in the second window, the labeling results generated by each second window are integrated according to the position of the feature points in the audio feature sequence to form a state labeling matrix. Each element in the state label matrix corresponds to the label state of an audio segment in the audio feature sequence. , The value of is greater than or equal to 1 and less than or equal to N. This indicates that the i-th audio segment is abnormal. This indicates that the i-th audio segment is normal.

[0082] Furthermore, such as Figure 6 As shown in step 4, based on the state label matrix, all second feature points marked as the first state are filtered out, and only the second feature points marked as the second state are retained as the filtered second feature points, thereby completing the secondary filtering of the second feature points from the energy dimension.

[0083] In step 511, based on the audio segments to which the multiple second feature points belong after filtering, multiple third feature points are determined. The third feature point is a second feature point that satisfies a preset index condition with the audio segments to which the adjacent second feature points are located. The preset index condition is related to the time required for two adjacent leaves to pass through the same position.

[0084] Theoretically, only one sweeping sound signal should appear within the time period required for two adjacent blades to pass the same position on the acoustic signature acquisition terminal. Taking the specific implementation scenario mentioned above as an example, within a complete cycle where the sweeping sound signal appears exactly once, that is, within the analysis area (corresponding to 12 audio segments) formed by a first window, theoretically only one second feature point should appear. However, in the complex operating environment of wind turbines, even after the aforementioned multiple rounds of screening, a small number of spurious peaks caused by noise signals or transient interference may still exist. To further eliminate such interference and ensure that only one true sweeping sound signal exists within a cycle, the previously screened second feature points need to be verified. Furthermore, considering that the sweeping sound signal has a periodic characteristic that is strictly synchronized with the blade rotation, the audio segment index position corresponding to the sweeping sound signal within each first window should be consistent and without positional shift. Therefore, the screened second feature points can be further screened based on this periodic constraint.

[0085] Based on this, such as Figure 6 As shown in step 4, this embodiment of the present disclosure first sets a preset index condition, that is, the audio segment indices of adjacent second feature points must be spaced apart by a fixed number of audio segments. This fixed number corresponds to the number of audio segments within a complete acquisition cycle of one leaf (i.e., one first window); and these second feature points forming a continuous sequence must satisfy the preset number. Then, among the multiple second feature points obtained after filtering, the index position of the audio segment to which each second feature point belongs is determined, and then it is determined whether the index position between adjacent second feature points satisfies the preset index condition. The second feature point that satisfies the preset index condition is determined as the third feature point. The preset number can be based on the minimum number of audio segments to be extracted. That is, the larger the number of audio frequency bands to be extracted, the larger the preset number can be set, and the smaller the number of audio segments to be extracted, the smaller the preset number can be set.

[0086] Taking the specific implementation scenario mentioned above as an example, in this scenario, the time period of two adjacent blades passing through the same position exactly contains a first window, corresponding to 12 audio segments. Based on this, the preset index condition is specifically set as follows: there should be an interval of 12 audio segments between adjacent second feature points, and the number of second feature points that continuously satisfy this interval rule is greater than or equal to 3. When the filtered sequence of second feature points contains points with consecutive indices and a number of no less than 3, these second feature points are considered to constitute a stable sweeping sound signal acquisition result and are identified as third feature points. Through this feature point filtering mechanism based on periodic characteristics, interference points that have passed stability and energy filtering but do not conform to the physical characteristics of the sweeping sound signal in terms of time distribution can be effectively eliminated.

[0087] In step 512, segments within a preset time period around the second feature point are extracted from the audio signal, centered on the second feature point, to obtain multiple audio segments of the blade sweeping sound.

[0088] In this embodiment of the application, a symmetrical time window is constructed by extending a preset duration to both the front and back sides of the feature point of the aforementioned confirmed wind-sweeping sound signal as the center, thereby completely extracting multiple audio segments of the wind-sweeping sound of the blades.

[0089] It should be noted that the feature points of the sweeping sound signal in this embodiment can be the second feature points after stability screening, or the third feature points after energy verification and continuity verification. Considering the different requirements for algorithm efficiency and extraction accuracy in actual application scenarios, this embodiment has flexible configuration capabilities. For example, in scenarios with high real-time requirements, segment extraction can be performed immediately after determining the second feature point to reduce computational delay; while in scenarios with high data accuracy requirements, energy threshold screening and continuity verification can be further combined to determine the third feature point before segment extraction, thereby ensuring that the extracted audio segments strictly correspond to the real sweeping sound signal.

[0090] Taking the specific implementation scenario mentioned above as an example, after the aforementioned steps, the third feature point was obtained. In this scenario, the preset duration was specifically set to 0.15 seconds. Taking the time point corresponding to the third feature point as the center, 0.15 seconds forward (to the left) was found as the start time of the sweeping segment, and 0.15 seconds backward (to the right) was found as the end time of the segment, thus extracting an audio segment with a total duration of 0.3 seconds.

[0091] To verify the effectiveness of the embodiments of this disclosure in a real wind turbine operating environment, the following is combined with... Figure 7 and Figure 8 The treatment effect is described in detail. Figure 7 This is an example illustrating the original audio signal and its time-frequency waveform under a wind turbine environment, as shown in an exemplary embodiment. Figure 1 The waveform and time-frequency diagram of the complete audio signal collected during one revolution of the three blades are shown. Observe. Figure 7 As shown in the time-frequency waveform of the image below, within the three rectangular areas marked by boxes, the energy distribution of the sweeping sound signal exhibits significant variations in color intensity. However, the signal amplitude in the corresponding waveform above is not significantly higher than the surrounding background noise. Conversely, in the waveform above, the signal amplitude of the noise signal located within the rectangular area is significantly higher than the amplitude of the aforementioned sweeping sound segment. This phenomenon intuitively demonstrates that in the complex operating environment of wind turbines, the waveform amplitude of the sweeping sound signal may be strongly interfered with. Therefore, relying solely on the amplitude of the time-domain waveform to filter sweeping sound signals is unreliable.

[0092] Figure 8 This is a processed audio signal and its voiceprint time-frequency diagram illustrated according to an exemplary embodiment, demonstrating the sweeping sound signal extracted and merged through the aforementioned steps. (Comparison) Figure 7 and Figure 8 It can be observed that the processed signal underwent significant changes in the time-frequency plot. Specifically, the changes were caused by… Figure 7 As shown on the frequency axis on the right side of the time-frequency graph below, a fixed component with a frequency of approximately 6000Hz (represented as a bright white line running the entire graph) consistently exists in the original audio signal. This component only disappears during the time segment containing the sweeping sound. However, in Figure 8 In the sweeping sound segment extracted in the embodiment of this disclosure shown, the component fixed at around 6000Hz has disappeared, which further verifies that the audio segment extracted in the embodiment of this disclosure is indeed a pure sweeping sound signal with noise signals removed.

[0093] The technical solution provided in this disclosure converts an audio signal into an audio feature sequence, slides a first window on the sequence, and extracts the feature point with the highest energy in each window as the first feature point to lock the potential sweeping sound location in each window. Then, using the first feature points of adjacent windows, points located in the same audio segment are selected as second feature points. Finally, the sweeping sound audio segment is extracted from the audio signal based on the second feature points. Since the sweeping sound signal and the noise signal have different stability in temporal characteristics, by limiting the first feature points of adjacent first windows to be located in the same audio segment to determine the second feature point, the stability characteristics of the sweeping sound signal in temporal sequence can be utilized to filter out the pseudo-peaks of random noise signals, thereby effectively extracting the audio segment of blade sweeping sound.

[0094] Figure 9 This is a block diagram illustrating an audio segment extraction apparatus according to an exemplary embodiment. See also... Figure 9 The device includes: The feature sequence determination module 901 is configured to acquire the audio feature sequence of the audio signal based on the audio signal collected during the operation of the wind turbine. The audio feature sequence is used to characterize the energy of the target frequency band in each audio segment of the audio signal. The first feature point determination module 902 is configured to slide a first window on the audio feature sequence, and for each first window, determine the first feature point of the first window, wherein the first feature point is the feature point with the highest energy in the audio feature subsequence within the first window. The second feature point determination module 903 is configured to determine multiple second feature points among multiple first feature points in adjacent first windows, wherein the second feature points are first feature points located in the same audio segment. The audio segment extraction module 904 is configured to extract audio segments of blade sweeping sound from the audio signal based on multiple second feature points.

[0095] In some embodiments, the device further includes: The window parameter setting module is configured to: determine the window length of the first window based on the first time, where the first time is the time required for two adjacent blades to pass the same position; The sliding step size of the first window is determined based on the second time, which is the time when the sweeping sound is collected once when the blade rotates.

[0096] In some embodiments, the device further includes: The audio segmentation module is configured to divide the audio signal into multiple audio segments based on time granularity, which is set based on the rotation parameters of the blades.

[0097] In some embodiments, the device further includes: The first filtering module is configured to slide a second window over the audio feature sequence. The properties, window length, and sliding step size of the second window are different from those of the first window. For each second window, the energy threshold of the second window is determined based on the second feature point in the audio feature subsequence within the second window. The energy threshold is used to characterize the energy limit of the sweeping sound in the audio signal. For each second window, filter out second feature points whose energy is greater than the energy threshold within the second window; The audio segment extraction module is configured to extract audio segments of blade sweeping sound from the audio signal based on multiple filtered second feature points.

[0098] In some embodiments, the first filtering module is configured to: for each second window, determine a reference energy value in the second window based on the second feature points in the audio feature subsequence within the second window, wherein the reference energy value is used to characterize the energy reference value of the sweeping sound in the second window; For each second window, the energy threshold of the second window is determined based on the reference energy value and energy ratio coefficient of the second window. The energy ratio coefficient is used to characterize the degree of energy deviation of the swept sound relative to the energy reference value.

[0099] In some embodiments, the first filtering module is configured to: for each second window, if the second feature point in the audio feature subsequence within the second window is not empty, determine the reference energy value in the second window based on the second feature point in the second window; For each second window, if the second feature point in the audio feature subsequence within the second window is empty, the baseline energy value in the second window is determined based on the feature points in the second window.

[0100] In some embodiments, the first filtering module is configured to: for each second window, mark feature points in the audio feature subsequence within the second window whose energy is greater than an energy threshold as a first state, and mark feature points whose energy is less than or equal to the energy threshold as a second state; Based on the labeled states of feature points in multiple second windows, a state labeling matrix for the audio feature sequence is constructed. Based on the state label matrix, the second feature points that are labeled as the first state are filtered out from each second window.

[0101] In some embodiments, the device further includes: The second filtering module is configured to determine multiple third feature points based on the audio segments to which the multiple second feature points belong after filtering. The third feature points are second feature points that satisfy a preset index condition with the audio segments to which the adjacent second feature points are located. The preset index condition is related to the time required for two adjacent leaves to pass through the same position. The audio segment extraction module is configured to extract audio segments of blade sweeping sound from the audio signal based on multiple third feature points.

[0102] In some embodiments, the audio segment extraction module is configured to extract segments within a preset time period around the second feature point in the audio signal, centered on the second feature point, to obtain multiple audio segments of the blade sweeping sound.

[0103] It should be noted that the audio segment extraction device provided in the above embodiments is only illustrated by the division of the above functional units. In practical applications, the above functions can be assigned to different functional units as needed, that is, the internal structure of the computer device can be divided into different functional units to complete all or part of the functions described above. In addition, the audio segment extraction device and the audio segment extraction method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0104] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0105] Computer equipment can be implemented as a terminal. Figure 10 This is a block diagram illustrating a terminal 1000 according to an exemplary embodiment. Figure 10A structural block diagram of a terminal 1000 provided in an exemplary embodiment of this disclosure is shown. The terminal 1000 may be a smartphone, tablet computer, MP3 player (Moving Picture Experts Group Audio Layer III), MP4 player (Moving Picture Experts Group Audio Layer IV), laptop computer, or desktop computer. The terminal 1000 may also be referred to as a user device, portable terminal, laptop terminal, desktop terminal, or other names.

[0106] Typically, terminal 1000 includes a processor 1001 and a memory 1002.

[0107] Processor 1001 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. Processor 1001 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). Processor 1001 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 1001 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the screen. In some embodiments, processor 1001 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0108] The memory 1002 may include one or more computer-readable storage media, which may be non-transitory. The memory 1002 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1002 are used to store at least one computer program, which is executed by the processor 1001 to implement the audio segment extraction method provided in the method embodiments of this application.

[0109] In some embodiments, the terminal 1000 may also optionally include a peripheral device interface 1003 and at least one peripheral device. The processor 1001, memory 1002, and peripheral device interface 1003 can be connected via a bus or signal line. Each peripheral device can be connected to the peripheral device interface 1003 via a bus, signal line, or circuit board. Specifically, the peripheral device includes at least one of the following: a radio frequency circuit 1004, a display screen 1005, a camera assembly 1006, an audio circuit 1007, and a power supply 1008.

[0110] Peripheral device interface 1003 can be used to connect at least one I / O (Input / Output) related peripheral device to processor 1001 and memory 1002. In some embodiments, processor 1001, memory 1002 and peripheral device interface 1003 are integrated on the same chip or circuit board; in some other embodiments, any one or two of processor 1001, memory 1002 and peripheral device interface 1003 can be implemented on separate chips or circuit boards, which is not limited in this embodiment.

[0111] The radio frequency (RF) circuit 1004 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The RF circuit 1004 communicates with communication networks and other communication devices via electromagnetic signals. The RF circuit 1004 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals back into electrical signals. In some embodiments, the RF circuit 1004 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, etc. The RF circuit 1004 can communicate with other terminals via at least one wireless communication protocol. This wireless communication protocol includes, but is not limited to: the World Wide Web, metropolitan area networks, intranets, various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks, and / or WiFi (Wireless Fidelity) networks. In some embodiments, the RF circuit 1004 may also include circuitry related to NFC (Near Field Communication), which is not limited in this application.

[0112] Display screen 1005 is used to display a UI (User Interface). This UI may include graphics, text, icons, videos, and any combination thereof. When display screen 1005 is a touch display screen, it also has the ability to collect touch signals on or above its surface. These touch signals can be input as control signals to processor 1001 for processing. In this case, display screen 1005 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there may be one display screen 1005, disposed on the front panel of terminal 1000; in other embodiments, there may be at least two display screens, disposed on different surfaces of terminal 1000 or in a folded design; in still other embodiments, display screen 1005 may be a flexible display screen, disposed on a curved or folded surface of terminal 1000. Furthermore, display screen 1005 may be configured as a non-rectangular, irregular shape, i.e., a non-rectangular screen. The display screen 1005 can be made of materials such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0113] The camera assembly 1006 is used to acquire images or videos. In some embodiments, the camera assembly 1006 includes a front-facing camera and a rear-facing camera. Typically, the front-facing camera is located on the front panel of the terminal, and the rear-facing camera is located on the back of the terminal. In some embodiments, there are at least two rear-facing cameras, which are any one of a main camera, a depth-sensing camera, a wide-angle camera, and a telephoto camera, to achieve background blurring by fusion of the main camera and the depth-sensing camera, panoramic shooting by fusion of the main camera and the wide-angle camera, VR (Virtual Reality) shooting, or other fusion shooting functions. In some embodiments, the camera assembly 1006 may also include a flash. The flash can be a single-color temperature flash or a dual-color temperature flash. A dual-color temperature flash is a combination of a warm-light flash and a cool-light flash, which can be used for light compensation at different color temperatures.

[0114] The audio circuit 1007 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, converting the sound waves into electrical signals that are input to the processor 1001 for processing, or input to the radio frequency circuit 1004 for voice communication. For stereo sound acquisition or noise reduction purposes, multiple microphones may be used, each positioned at a different location on the terminal 1000. The microphone may also be an array microphone or an omnidirectional microphone. The speaker is used to convert electrical signals from the processor 1001 or the radio frequency circuit 1004 into sound waves. The speaker may be a conventional diaphragm speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can convert electrical signals not only into audible sound waves but also into inaudible sound waves for purposes such as distance measurement. In some embodiments, the audio circuit 1007 may also include a headphone jack.

[0115] The power supply 1008 is used to power the various components in the terminal 1000. The power supply 1008 can be AC ​​power, DC power, a disposable battery, or a rechargeable battery. When the power supply 1008 includes a rechargeable battery, the rechargeable battery can be a wired rechargeable battery or a wireless rechargeable battery. A wired rechargeable battery is a battery that is charged via a wired line, while a wireless rechargeable battery is a battery that is charged via a wireless coil. The rechargeable battery can also be used to support fast charging technology.

[0116] Those skilled in the art will understand that Figure 10 The structure shown does not constitute a limitation on terminal 1000 and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0117] Computer equipment can be implemented as a server. Figure 11 This is a schematic diagram of a server structure provided in an embodiment of this application. The server 1100 can vary significantly due to different configurations or performance. It may include one or more Central Processing Units (CPUs) 1101 and one or more memories 1102. The one or more memories 1102 store at least one computer program, which is loaded and executed by the one or more processors 1101 to implement the audio segment extraction method provided in the above-described method embodiments. Of course, the server 1100 may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input and output. The server 1100 may also include other components for implementing device functions, which will not be elaborated here.

[0118] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including a computer program that can be executed by a processor to perform the audio segment extraction method in the above embodiments. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage device, etc.

[0119] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory including instructions, which can be executed by a processor of a computer device, such as a terminal, to complete the above-described audio segment extraction method. Optionally, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0120] This disclosure also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the above-described audio segment extraction method.

[0121] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0122] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for extracting audio segments, characterized in that, The method includes: Based on the audio signals collected during the operation of the wind turbine, an audio feature sequence of the audio signals is obtained, and the audio feature sequence is used to characterize the energy of the target frequency band in each audio segment of the audio signal; Slide a first window on the audio feature sequence. For each first window, determine a first feature point of the first window. The first feature point is the feature point with the highest energy in the audio feature subsequence within the first window. Among multiple first feature points in adjacent first windows, multiple second feature points are determined, wherein the second feature points are first feature points located in the same audio segment; Based on multiple second feature points, audio segments of blade sweeping sound are extracted from the audio signal.

2. The method for extracting audio segments according to claim 1, characterized in that, The method further includes: The window length of the first window is determined based on the first time, where the first time is the time required for two adjacent blades to pass the same position. The sliding step size of the first window is determined based on the second time, which is the time when the sweeping sound is collected once when the blade rotates.

3. The method for extracting audio segments according to claim 1, characterized in that, The method further includes: The audio signal is divided into multiple audio segments based on time granularity, which is set based on the rotation parameters of the blade.

4. The method for extracting audio segments according to claim 1, characterized in that, The method further includes: A second window is slid across the audio feature sequence. The properties, window length, and sliding step size of the second window are different from those of the first window. For each second window, an energy threshold for the second window is determined based on the second feature points in the audio feature subsequence within the second window. The energy threshold is used to characterize the energy limit of the sweeping sound in the audio signal. For each second window, filter out second feature points whose energy is greater than the energy threshold within the second window; The extraction of audio segments of blade sweeping sound from the audio signal based on each second feature point includes: Based on multiple filtered second feature points, audio segments of blade sweeping sounds are extracted from the audio signal.

5. The method for extracting audio segments according to claim 4, characterized in that, For each second window, determining the energy threshold of the second window based on the second feature points in the audio feature subsequence within the second window includes: For each second window, a reference energy value is determined based on the second feature point in the audio feature subsequence within the second window. The reference energy value is used to characterize the energy reference value of the sweeping sound in the second window. For each second window, an energy threshold is determined based on the reference energy value and energy ratio coefficient of the second window, wherein the energy ratio coefficient is used to characterize the degree of energy deviation of the sweeping sound relative to the energy reference value.

6. The method for extracting audio segments according to claim 5, characterized in that, For each second window, determining the baseline energy value within the second window based on the second feature points in the audio feature subsequence within the second window includes: For each second window, if the second feature point in the audio feature subsequence within the second window is not empty, the baseline energy value in the second window is determined based on the second feature point in the second window. For each second window, if the second feature point in the audio feature subsequence within the second window is empty, the baseline energy value in the second window is determined based on the feature points in the second window.

7. The method for extracting audio segments according to claim 4, characterized in that, For each second window, filtering out second feature points within the second window whose energy is greater than the energy threshold includes: For each second window, feature points in the audio feature subsequence within the second window whose energy is greater than the energy threshold are marked as the first state, and feature points whose energy is less than or equal to the energy threshold are marked as the second state; Based on the labeled states of feature points in multiple second windows, a state labeling matrix for the audio feature sequence is constructed. Based on the state label matrix, the second feature points marked as the first state are filtered out from each second window.

8. The method for extracting audio segments according to claim 4, characterized in that, The method further includes: Based on the audio segments to which the multiple second feature points belong after filtering, multiple third feature points are determined. The third feature points are second feature points that satisfy a preset index condition with the audio segments to which the adjacent second feature points are located. The preset index condition is related to the time required for two adjacent leaves to pass through the same position. The extraction of audio segments of blade sweeping sound from the audio signal based on multiple filtered second feature points includes: Based on multiple third feature points, audio segments of blade sweeping sound are extracted from the audio signal.

9. The method for extracting audio segments according to claim 1, characterized in that, The extraction of audio segments of blade sweeping sound from the audio signal based on multiple second feature points includes: In the audio signal, taking the second feature point as the center, extract segments within a preset time period around the second feature point to obtain multiple audio segments of the blade sweeping sound.

10. An audio segment extraction device, characterized in that, include: The feature sequence determination module is configured to acquire an audio feature sequence of an audio signal based on an audio signal collected during the operation of the wind turbine. The audio feature sequence is used to characterize the energy of the target frequency band in each audio segment of the audio signal. The first feature point determination module is configured to slide a first window on the audio feature sequence, and for each first window, determine a first feature point of the first window, wherein the first feature point is the feature point with the highest energy in the audio feature subsequence within the first window; The second feature point determination module is configured to determine multiple second feature points among multiple first feature points in adjacent first windows, wherein the second feature points are first feature points located in the same audio segment; The audio segment extraction module is configured to extract audio segments of blade sweeping sound from the audio signal based on multiple second feature points.

11. A computer device, characterized in that, The computer device includes: One or more processors; Memory used to store the executable program code of the processor; The processor is configured to execute the program code to implement the audio segment extraction method as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of a computer device, the computer device is enabled to perform the method for extracting audio segments as described in any one of claims 1 to 9.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the audio segment extraction method as described in any one of claims 1 to 9.

Images