A distributed optical fiber sound source identification and classification method fusing i-vector algorithm

Abstract

The application discloses a distributed optical fiber sound source identification and classification method fusing an i-vector algorithm, relates to the technical field of distributed optical fiber sound sources, and is used for reducing the identification efficiency of the distributed optical fiber sound sources. A plurality of environment detection points are arranged in a sound source identification environment, background noise intensity of each point is detected, a reference sound wave is emitted and sound wave absorption is collected, sound wave absorption characteristics are calculated and sound source identification states are analyzed, and target points are screened based on the identification states to identify the sound sources. Sound source information of an optical fiber to be identified is played at the target points, the number of storage types and the total occupied space are counted, an identification range is calculated, a collection time window is set, volume is collected at a plurality of time points, volume fluctuation is calculated, and whether the sound source information is input into an i-vector algorithm for processing is judged in combination with the identification range, so that the i-vector algorithm processing environment is ensured, and the identification efficiency of the distributed optical fiber sound sources is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of distributed optical fiber sound source, more particularly, the present application relates to a kind of distributed optical fiber sound source identification and classification method of fusion i-vector algorithm. BACKGROUND

[0002] With the development of optical fiber sensing technology, distributed optical fiber sensing system (Distributed Optical Fiber Sensing System, DOFSS) has been widely used in structural health monitoring, perimeter intrusion detection and oil and gas pipeline anomaly monitoring and other fields. By introducing laser signal in optical fiber, the system can realize real-time sensing and acquisition of vibration, temperature and strain information along the optical fiber layout path, especially in sound source positioning and identification. It shows unique advantages.

[0003] The prior art has the following shortcomings:

[0004] At present, due to environmental noise interference, sound absorption characteristics change and signal attenuation on optical fiber transmission path, the recognition accuracy of the prior art in complex background environment has a downward trend. In large-scale optical fiber network environment with multiple node deployment, the processing environment of i-vector algorithm is variable, which limits the algorithm operation and reduces the distributed optical fiber sound source recognition efficiency. Therefore, a distributed optical fiber sound source recognition and classification method based on i-vector algorithm is proposed.

[0005] The above information disclosed in the background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a distributed optical fiber sound source recognition and classification method based on i-vector algorithm. By using a target point screening mechanism based on sound absorption characteristics and background noise intensity, combining a dynamic recognition range determination method based on stored feature evaluation, and a sound source feature processing model based on fusion of volume fluctuation analysis and i-vector algorithm, the problems raised in the above background technology are solved.

[0007] To achieve the above purpose, the present application provides the following technical scheme, a distributed optical fiber sound source recognition and classification method based on i-vector algorithm, comprising the following steps:

[0008] Step S1: In the sound source recognition environment, a plurality of environment detection points are set, the background noise intensity of each environment detection point is detected, a reference sound wave is emitted to each environment detection point, and the sound absorption amount of each environment detection point is collected.

[0009] Step S2: Calculate the sound wave absorption characteristics of each environment detection point according to the sound wave absorption amount, analyze the sound source recognition state of each environment detection point by combining the background noise intensity of the corresponding environment detection point, and select a target point in the environment detection point based on the sound source recognition state for sound source recognition;

[0010] Step S3: Play the to-be-identified optical fiber sound source information at the target point, store the to-be-identified optical fiber sound source information in the sound source database, count the storage type quantity and total storage space of the to-be-identified optical fiber sound source information, and calculate the to-be-identified range of the to-be-identified optical fiber sound source information by comprehensively considering the storage type quantity and total storage space of the to-be-identified optical fiber sound source information.

[0011] Step S4: Set a collection time window, collect the to-be-identified optical fiber sound source information release volume at multiple time points in the collection time window, calculate the volume fluctuation of the to-be-identified optical fiber sound source information according to the to-be-identified optical fiber sound source information release volume at each time point, and determine whether to pass the to-be-identified optical fiber sound source information into the i-vector algorithm for processing in combination with the to-be-identified range of the to-be-identified optical fiber sound source information.

[0012] In a preferred embodiment, in step S1, a plurality of environment detection points are randomly set in the sound source recognition environment, and the background noise of each environment detection point is detected;

[0013] A distance is randomly selected as the transmission distance, a reference sound wave is transmitted to each environment detection point based on the transmission distance, the echo intensity of the corresponding environment detection point is received, and the difference between the sound wave intensity of the reference sound wave and the echo intensity of the environment detection point is taken as the sound wave absorption amount of the corresponding environment detection point.

[0014] In a preferred embodiment, in step S2, the absolute difference value operation is performed on the sound wave absorption amount of each environment detection point and the preset calibration sound wave, and the operation result is taken as the sound wave absorption characteristic of the corresponding environment detection point;

[0015] The ratio of the background noise intensity of each environment detection point to the maximum background noise intensity of all environment detection points is taken as the background noise intensity coefficient of the corresponding environment detection point, and a polynomial regression model is constructed by combining the sound wave absorption characteristic of the corresponding environment detection point to analyze the sound source recognition state of each environment detection point.

[0016] In a preferred embodiment, in step S2, a polynomial regression model is constructed by comprehensively considering the sound wave absorption characteristics and background noise intensity coefficients of each environment detection point to analyze the sound source recognition state, and the specific steps are as follows:

[0017] The calculation of the influencing factor: the sound wave absorption characteristics and the preset first influence weight are multiplied to generate the first influence factor; the square of the background noise intensity coefficient and the preset second influence weight are multiplied to generate the second influence factor;

[0018] Generating a sound source recognition state index: summing the first influence factor and the second influence factor of each environment detection point to obtain the sound source recognition state index of the corresponding environment detection point;

[0019] Screening target points: among all the environment detection points, the environment detection point corresponding to the minimum value of the sound source recognition state index is selected as the target point.

[0020] In a preferred embodiment, in step S3, the to-be-identified optical fiber sound source information is played at the target point and stored in the sound source database by the optical fiber sound wave sensing device;

[0021] The to-be-identified optical fiber sound source information includes the storage type of the optical fiber sound source and the total storage space of various types. The count of different types of optical fiber sound sources is taken as the storage type number, and the storage space of all optical fiber sound source types is added up as the total storage space.

[0022] In a preferred embodiment, in step S3, the storage type number and the total storage space are standardized and then introduced into the geometric mean method to obtain the to-be-identified range score of the to-be-identified optical fiber sound source information;

[0023] The to-be-identified range score of the to-be-identified optical fiber sound source information is compared with the preset range threshold;

[0024] If the to-be-identified range score of the to-be-identified optical fiber sound source information exceeds the range threshold, the to-be-identified range of the to-be-identified optical fiber sound source information is marked as large;

[0025] If the to-be-identified range score of the to-be-identified optical fiber sound source information is lower than the range threshold, the to-be-identified range of the to-be-identified optical fiber sound source information is marked as small.

[0026] In a preferred embodiment, in step S4, the to-be-identified optical fiber sound source information release volume of the target point at multiple time points in the set collection time window is collected by the optical fiber sound source demodulation collector.

[0027] In a preferred embodiment, in step S4, the to-be-identified optical fiber sound source information release volume standard deviation of each time point is calculated as the volume fluctuation of the to-be-identified optical fiber sound source information according to the to-be-identified optical fiber sound source information release volume of each time point.

[0028] In a preferred embodiment, in step S4, the to-be-identified range and the volume fluctuation are defined as input variables, which are respectively divided into different fuzzy sets;

[0029] The to-be-identified optical fiber sound source information input result is defined as an output variable, which is divided into a fuzzy set;

[0030] Fuzzy rules are formulated to describe the influence of the to-be-identified range and the volume fluctuation on the to-be-identified optical fiber sound source information input result;

[0031] According to the fuzzy rules, fuzzy reasoning is performed to determine the to-be-identified optical fiber sound source information input result.

[0032] In a preferred embodiment, in step S4, the to-be-identified optical fiber sound source information input result includes processing the to-be-identified optical fiber sound source information into an i-vector algorithm and processing the to-be-identified optical fiber sound source information without an i-vector algorithm.

[0033] Technical effects and advantages of the present application:

[0034] The present application sets multiple environment detection points in the sound source identification environment, detects the background noise of each environment detection point and obtains the background noise intensity, emits a reference sound wave to each environment detection point, collects the sound wave absorption amount of each environment detection point to calculate the sound wave absorption characteristics and analyze the sound source identification state of each environment detection point, selects a target point from the environment detection points based on the sound source identification state for sound source identification, plays the to-be-identified optical fiber sound source information at the target point, counts the storage type number and total storage space of the to-be-identified optical fiber sound source information, calculates the to-be-identified range of the to-be-identified optical fiber sound source information, sets a collection time window, collects the release volume of the to-be-identified optical fiber sound source information of the target point at multiple time points in the collection time window, calculates the volume fluctuation of the sound source information according to the release volume of the to-be-identified optical fiber sound source information at each time point, and determines whether to process the to-be-identified optical fiber sound source information into an i-vector algorithm in combination with the to-be-identified range of the to-be-identified optical fiber sound source information, thereby guaranteeing the processing environment of the i-vector algorithm, improving the distributed optical fiber sound source identification efficiency, and realizing high-precision classification and intelligent scheduling processing of the optical fiber sound source under the conditions of multiple noise interference and sound source complexity. BRIEF DESCRIPTION OF DRAWINGS

[0035] Fig. 1 The present application is a realization flowchart of a distributed optical fiber sound source identification and classification method fused with an i-vector algorithm.

[0036] Fig. 2 The present application is a step schematic diagram of a distributed optical fiber sound source identification and classification method fused with an i-vector algorithm. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0038] The present application sets multiple environment detection points in a sound source recognition environment, detects the background noise of each environment detection point and obtains the background noise intensity, emits reference sound waves to each environment detection point, collects the sound wave absorption amount of each environment detection point, calculates the sound wave absorption characteristics, analyzes the sound source recognition state of each environment detection point, filters out a target point from the environment detection points based on the sound source recognition state for sound source recognition, plays the to-be-recognized fiber sound source information at the target point, counts the storage type quantity and total storage space of the to-be-recognized fiber sound source information, calculates the to-be-recognized range of the to-be-recognized fiber sound source information, sets a collection time window, collects the to-be-recognized fiber sound source information release volume of the target point at multiple time points in the collection time window, calculates the volume fluctuation of the sound source information according to the to-be-recognized fiber sound source information release volume at each time point, judges whether to pass the to-be-recognized fiber sound source information into the i-vector algorithm for processing in combination with the to-be-recognized range of the to-be-recognized fiber sound source information, so as to guarantee the processing environment of the i-vector algorithm and improve the distributed fiber sound source recognition efficiency.

[0039] Embodiment 1, a distributed fiber sound source recognition and classification method fusing an i-vector algorithm, as shown in Figs. 1-2 The method comprises the following steps:

[0040] Step S1: multiple environment detection points are set in a sound source recognition environment, the background noise of each environment detection point is detected to obtain the background noise intensity, reference sound waves are emitted to each environment detection point, and the sound wave absorption amount of each environment detection point is collected.

[0041] Step S2: the sound wave absorption characteristics of each environment detection point are calculated according to the sound wave absorption amount, the sound source recognition state of each environment detection point is analyzed in combination with the background noise intensity of the corresponding environment detection point, and a target point is filtered out from the environment detection points based on the sound source recognition state for sound source recognition.

[0042] Step S3: to-be-recognized fiber sound source information is played at the target point, the to-be-recognized fiber sound source information is stored in a sound source database, the storage type quantity and total storage space of the to-be-recognized fiber sound source information are counted, and the to-be-recognized range of the to-be-recognized fiber sound source information is calculated in combination with the storage type quantity and total storage space of the to-be-recognized fiber sound source information.

[0043] Step S4: setting a collection time window, collecting the to-be-identified optical fiber sound source information release volume of the target point at multiple time points in the collection time window, calculating the volume fluctuation of the to-be-identified optical fiber sound source information according to the to-be-identified optical fiber sound source information release volume at each time point, and combining the to-be-identified range of the to-be-identified optical fiber sound source information to determine whether to pass the to-be-identified optical fiber sound source information into the i-vector algorithm for processing.

[0044] The specific implementation is as follows:

[0045] In step S1, since the background noise at different positions in the same environment is different, which further causes different effects of receiving sound sources at different positions, a plurality of environment detection points are randomly set in the sound source identification environment, the background noise at each environment detection point is detected by a noise monitoring sensor, and the value displayed in the noise monitoring sensor is taken as the background noise intensity of the corresponding environment detection point.

[0046] A distance is randomly selected as the transmission distance, and a reference sound wave is transmitted to each environment detection point by using a sound wave sensor. When the reference sound wave is transmitted to different environment detection points, the distance between the sound wave sensor and the corresponding environment detection point is ensured to be the transmission distance.

[0047] The echo intensity of the corresponding environment detection point is received by using the sound wave sensor, and the difference between the sound wave intensity of the reference sound wave and the echo intensity of the environment detection point is taken as the sound wave absorption amount of the corresponding environment detection point.

[0048] It should be noted that the noise monitoring sensor is a device for collecting and recording environmental noise, which is used to detect sound sources at each environment detection point in the sound source identification environment in this example; the sound wave sensor is a device for detecting and measuring sound waves, which is used to transmit a reference sound wave to the environment detection point and detect the echo intensity of the environment detection point. The reference sound wave is set by a person skilled in the art according to the actual situation, which is not described here.

[0049] In step S2, when calculating the sound wave absorption characteristics of each environment detection point according to the sound wave absorption amount, the absolute difference value operation is performed on the sound wave absorption amount of each environment detection point and the preset calibration sound wave, and the operation result is taken as the sound wave absorption characteristics of the corresponding environment detection point.

[0050] It should be explained that too large or too small sound wave absorption amount will affect the sound source identification state of the corresponding environment detection point. When the sound wave absorption amount is too small, the echo and reverberation effect of the sound source is large, and the sound source feature extraction effect is worse. When the sound wave absorption amount is too large, the sound source signal will be excessively attenuated, resulting in that the captured features are not obvious. The calibration sound wave is the ideal value of the sound wave absorption amount in the sound source identification environment, which can be set according to the actual situation, and is not analyzed too much here.

[0051] The ratio of the background noise intensity of each environment detection point to the maximum value of the background noise intensity of all environment detection points is taken as the background noise intensity coefficient of the corresponding environment detection point, and a polynomial regression model is constructed combining the sound wave absorption characteristics of the corresponding environment detection point to analyze the sound source identification state of each environment detection point.

[0052] The greater the sound wave absorption characteristics of the environment detection point, the greater or smaller the sound wave absorption of the corresponding environment detection point, and the worse the sound source identification state of the corresponding environment detection point. The greater the background noise intensity coefficient of the environment detection point, the higher the background noise intensity of the environment detection point, the stronger the sound source identification interference of the corresponding environment detection point, and the worse the sound source identification state of the corresponding environment detection point.

[0053] The sound wave absorption characteristics and the background noise intensity coefficient of each environment detection point are combined to construct a polynomial regression model to analyze the sound source identification state, and the specific steps are as follows:

[0054] Calculate the influencing factors: multiply the sound wave absorption characteristics by the preset first influence weight to generate the first influence factor; square the background noise intensity coefficient and multiply it by the preset second influence weight to generate the second influence factor;

[0055] Generate the sound source identification state index: sum the first influence factor and the second influence factor of each environment detection point to obtain the sound source identification state index of the corresponding environment detection point;

[0056] It should be noted that the influence weight in the above is not unique and can be set according to actual conditions, for example, the first influence weight is set to 0.4 and the second influence weight is set to 0.6.

[0057] Screen the target point: among all the environment detection points, the environment detection point corresponding to the minimum value of the sound source identification state index is selected as the target point.

[0058] In step S3, the target point plays the to-be-identified fiber sound source information and stores the to-be-identified fiber sound source information into the sound source database through the fiber sound wave sensing device;

[0059] It should be noted that the fiber sound wave sensing device is a sound wave acquisition module constructed based on the principle of distributed optical fiber acoustic sensing; the acquired sound source signal is preprocessed through a signal coding module and then stored in the sound source database in the form of a time sequence, denoted as:

[0060] ;

[0061] Where, S(t i ) represents the to-be-identified sound source signal collected at time point t i , D srcThe fiber acoustic source information dataset is i = 1, 2, 3, … n, i is the i th time sampling point, corresponding to the i th acoustic source sampling data in the acquisition time window;

[0062] Further, the acoustic source database is a multi-modal acoustic information storage system with a structured tag indexing mechanism, including: a time indexing unit, a spectral feature indexing unit, an acoustic source label mapping unit, and a storage management unit, for supporting efficient retrieval and classification of acoustic source data;

[0063] The time indexing unit is used to record the acquisition time stamp of each acoustic source signal; the spectral feature indexing unit is used to store the acoustic source frequency components extracted by FFT, MFCC, etc. spectral processing algorithm; the acoustic source label mapping unit pre-classifies part of the acoustic source information according to the historical training model, and labels the acoustic source type (such as "mechanical knocking", "liquid dropping", "human voice", etc.); The storage management unit dynamically configures the compression rate, storage path and redundancy backup mechanism of the signal, supports concurrent access strategy of writing and retrieving at the same time, and counts the storage space of the fiber acoustic source storage type and various types;

[0064] It should be noted that the fiber acoustic source information refers to the acoustic response data corresponding to the perturbation signal excited on the optical fiber body after the external excitation source (such as a sound wave generator, an environmental natural sound or a simulated event source) acts on the distributed optical fiber sensing channel, which is not described here;

[0065] The fiber acoustic source information to be identified includes the fiber acoustic source storage type and the total storage space of various types, and the count results are used as the storage type number. The storage space of all fiber acoustic source types is accumulated as the total storage space.

[0066] It should be noted that the storage type number refers to the number of different categories formed by the current collected fiber acoustic source information to be identified in the acoustic source database according to the signal feature dimension;

[0067] It should be noted that the total storage space is the cumulative value of the total physical storage capacity occupied by the current fiber acoustic source information to be identified in the acoustic source database, which is obtained by counting the classified and stored acoustic source data information by the acoustic source database, and is used to reflect the storage scale and complexity of the fiber acoustic source data as a whole.

[0068] The storage type number and the total storage space are standardized to make the dimensions of the storage type number and the total storage space consistent and the numerical range between 0 and 1, maintaining the original logical correlation;

[0069] It should be noted that the standardization processing mode includes but is not limited to standard linear transformation based on interval scaling, Z-Score standardization method based on statistics or normalization method based on nonlinear mapping function, and the application method of standardization processing is not described here;

[0070] The normalized storage category quantity and the total storage space are substituted into the geometric mean method to obtain the to-be-identified range score of the to-be-identified optical fiber sound source information, and the specific geometric mean method formula is as follows:

[0071] ;

[0072] In the formula, C is the to-be-identified range score of the to-be-identified optical fiber sound source information, a is the normalized storage category quantity, and b is the normalized total storage space.

[0073] The to-be-identified range score of the to-be-identified optical fiber sound source information is compared with the preset range threshold value;

[0074] If the to-be-identified range score of the to-be-identified optical fiber sound source information exceeds the range threshold value, the to-be-identified range of the to-be-identified optical fiber sound source information is marked as a large to-be-identified range.

[0075] If the to-be-identified range score of the to-be-identified optical fiber sound source information is lower than the range threshold value, the to-be-identified range of the to-be-identified optical fiber sound source information is marked as a small to-be-identified range.

[0076] It should be noted that the preset range threshold value is set by the experimenters according to the maximum sound source classification complexity that the system can support and the database storage capacity in the actual deployment environment, and the specific setting mode includes but is not limited to this, and details are not described here.

[0077] Optionally, according to the maximum storage category quantity and the average storage space that can be stably processed in the past experiments, considering the actual hardware performance indicators (including memory upper limit, concurrent IO capability, etc.) of the deployment platform (such as edge node, server cluster), and combining error tolerance, classification accuracy requirement, training data size and other parameters to adjust the range threshold value, etc.

[0078] In step S4, the to-be-identified optical fiber sound source information release volume of the target point at multiple time points in the set collection time window is collected by the optical fiber sound source demodulation collector.

[0079] The to-be-identified optical fiber sound source information release volume standard deviation of each time point is calculated according to the to-be-identified optical fiber sound source information release volume of each time point as the volume fluctuation of the to-be-identified optical fiber sound source information.

[0080] The optical fiber acoustic source demodulation acquisition instrument is a common acquisition device for technicians in the field, which is used to acquire and quantify the amplitude information of the acoustic disturbance signal in the distributed optical fiber sensing channel. The specific device operation method is common knowledge for technicians in the field, and will not be described here.

[0081] The skilled person can realize that through the optical fiber acoustic source demodulation acquisition instrument, the release volume at multiple time points can be acquired with high precision without affecting the optical fiber main sensing link, and output in time sequence form for subsequent volume fluctuation analysis.

[0082] It should be noted that the time length of the acquisition time window is set by the experimenters according to the average duration of the acoustic source event and the real-time processing capability of the system; the number of time points in the acquisition time window is set according to the required time resolution accuracy.

[0083] Specifically, the setting rule of the number of time points is not fixed, and can be adjusted according to the complexity of the acoustic source signal change and the target recognition accuracy requirement, so as to flexibly set the acquisition time window and the number of time points in the acquisition time window, which will not be described here.

[0084] The volume fluctuation of the to-be-identified optical fiber acoustic source information is compared with the preset fluctuation threshold. If the volume fluctuation of the to-be-identified optical fiber acoustic source information exceeds the fluctuation threshold, the volume fluctuation of the to-be-identified optical fiber acoustic source information is marked as volume fluctuation high.

[0085] If the volume fluctuation of the to-be-identified optical fiber acoustic source information is lower than the fluctuation threshold, the volume fluctuation of the to-be-identified optical fiber acoustic source information is marked as volume fluctuation low.

[0086] It should be noted that the preset fluctuation threshold is set by the experimenters according to the standard deviation distribution characteristics of the historical acoustic source signal and the sensitive discrimination ability of the acoustic source stability difference, which will not be described here.

[0087] The to-be-identified optical fiber acoustic source information is determined according to the to-be-identified range and the volume fluctuation using fuzzy logic.

[0088] Since the known to-be-identified range is marked as to-be-identified range large and to-be-identified range small, and the volume fluctuation is marked as volume fluctuation high and volume fluctuation low.

[0089] A set of fuzzy rules is developed to describe the influence of different input variables on the output variable. The definition of the rule can be based on professional knowledge, or obtained through data analysis and experiment.

[0090] Then it can be defined as:

[0091] Rule 1: If the to-be-identified range is large and the volume fluctuation is low, the to-be-identified optical fiber sound source information input result is to input the to-be-identified optical fiber sound source information into the i-vector algorithm for processing.

[0092] Otherwise, the to-be-identified optical fiber sound source information input result is that the to-be-identified optical fiber sound source information is not input into the i-vector algorithm for processing.

[0093] According to the fuzzy rule, the fuzzy reasoning is performed to determine the to-be-identified optical fiber sound source information input result.

[0094] Further, the to-be-identified optical fiber sound source information input result includes inputting the to-be-identified optical fiber sound source information into the i-vector algorithm for processing and not inputting the to-be-identified optical fiber sound source information into the i-vector algorithm for processing.

[0095] It should be noted that the division of the fuzzy set can be adjusted according to actual conditions. For example, although three fuzzy sets are taken as examples in the embodiment, the to-be-identified range and the volume fluctuation can be divided into more than three sets to facilitate better and more accurate adjustment according to different similarities.

[0096] The i-vector algorithm is a low-dimensional space modeling algorithm currently widely used in the fields of speech recognition, speaker recognition and environmental sound classification. Its core principle is to map high-dimensional acoustic features (such as GMM hyper vectors) to a low-dimensional constant feature space to achieve feature compression and class discrimination ability enhancement. It is usually suitable for scenarios where the input sound source data is large (such as sound source storage with rich categories and high total storage space), the sound source information has high stability (such as small volume fluctuation), and the sound source features have distinguishability but high dimensionality and need to be reduced in dimension.

[0097] Therefore, when the number of storage categories of the to-be-identified sound source information is larger and the total storage space is larger (i.e., the to-be-identified range is large), and the volume fluctuation is smaller, it indicates that the sound source has good statistical aggregation and feature stability, and is more suitable for using the i-vector algorithm for modeling and recognition.

[0098] Further, when the to-be-identified range is small (small data volume and few categories) but the volume fluctuation is low, although the sound source is stable, the limited data source cannot support the low-dimensional space modeling requirement of the i-vector algorithm, which may lead to model overfitting (only adapting to a small amount of sample features) and cannot be generalized to actual classification tasks. At this time, the i-vector algorithm loses its advantage due to insufficient data, and is not as efficient as a simple classification algorithm, so in this scenario, the i-vector algorithm is not suitable.

[0099] Finally, it should be noted that the terms "first" and "second", and the like, herein do not denote any order, quantity, combination or important / primary / secondary status, but are used to merely distinguish one element from another, and do not imply any actual relationship or sequence between or among the elements.

[0100] Also, the use of "including," "comprising," or "having" and variations thereof herein is meant to encompass the items listed thereafter and equivalents thereof as well as additional items. Unless otherwise specified, "or" means "and / or". Unless otherwise noted, the use of the singular includes the plural.

[0101] It should also be understood that, unless clearly indicated otherwise, terms such as "comprise", "comprising", "include", "including", "contain", "containing", "have", "having" or variants thereof are used inclusively and that terms such as "consist of" or "consisting of" are used exclusively.

[0102] The various embodiments described in this specification are presented by way of example, and are not intended to limit the scope of the disclosure. Each embodiment described herein can be implemented in combination with one or more other embodiments described herein. The description herein of any contemplated embodiment includes textual references listing one or more components that are included in that embodiment. Such textual references should be understood as describing a single embodiment that includes all of the components listed in that reference, and as describing alternative embodiments that include only those components listed in the reference. For example, a first embodiment is described as including components A, B, and C, and a second embodiment is described as including components B and C. The description of the first embodiment as including components A, B, and C should be understood as describing a single embodiment that includes all of the components A, B, and C, and as describing alternative embodiments that include only components B and C, or only components B and A. Similarly, the description of the second embodiment as including components B and C should be understood as describing a single embodiment that includes both components B and C, and as describing alternative embodiments that include only component B, or only component C. The description of any embodiment as including a component should be understood as describing a single embodiment that includes only that component, and as describing alternative embodiments that include an additional component or components in addition to that component.

[0103] The above description of disclosed embodiments is intended to be illustrative and not restrictive. Many embodiments will be apparent to those of skill in the art upon reading this disclosure. The scope of the application should, therefore, be determined not with reference to the above description, but instead with reference to the appended claims, along with their full scope of equivalents. All publications and patent documents cited herein are incorporated by reference in their entirety for the purpose of explaining and disclosing the concepts underlying the present application.

Claims

1. A distributed optical fiber sound source identification and classification method integrating the i-vector algorithm, characterized in that: Includes the following steps: Step S1: Set up multiple environmental detection points in the sound source identification environment, detect the background noise of each environmental detection point to obtain the background noise intensity, emit reference sound waves to each environmental detection point, and collect the sound wave absorption of each environmental detection point. Step S2: Calculate the sound absorption characteristics of each environmental detection point based on the sound absorption amount, analyze the sound source identification status of each environmental detection point in combination with the background noise intensity of the corresponding environmental detection point, and select target points from the environmental detection points for sound source identification based on the sound source identification status. Step S3: Play the fiber optic sound source information to be identified at the target location, store the fiber optic sound source information to be identified in the sound source database, count the number of storage types and the total storage space occupied by the fiber optic sound source information to be identified, and calculate the identification range of the fiber optic sound source information to be identified by combining the number of storage types and the total storage space occupied by the fiber optic sound source information to be identified. Step S4: Set the acquisition time window, and acquire the release volume of the fiber optic sound source information to be identified at the target location at multiple time points within the acquisition time window. Calculate the volume fluctuation of the fiber optic sound source information to be identified based on the release volume of the fiber optic sound source information to be identified at each time point, and determine whether to feed the fiber optic sound source information to be identified into the i-vector algorithm for processing based on the identification range of the fiber optic sound source information to be identified.

2. The distributed optical fiber sound source identification and classification method integrating the i-vector algorithm according to claim 1, characterized in that: In step S1, multiple environmental detection points are randomly set in the sound source identification environment, and the background noise at each environmental detection point is detected. A distance is randomly selected as the transmission distance. Reference sound waves are transmitted to each environmental detection point based on the transmission distance. The echo intensity of the corresponding environmental detection point is received. The difference between the sound wave intensity of the reference sound wave and the echo intensity of the environmental detection point is taken as the sound wave absorption of the corresponding environmental detection point.

3. The distributed optical fiber sound source identification and classification method integrating the i-vector algorithm according to claim 2, characterized in that: In step S2, the absolute difference between the sound wave absorption at each environmental detection point and the preset calibration sound wave is calculated, and the calculation result is used as the sound wave absorption characteristic of the corresponding environmental detection point. The ratio of the background noise intensity of each environmental monitoring point to the maximum background noise intensity of all environmental monitoring points is used as the background noise intensity coefficient of the corresponding environmental monitoring point. A multinomial regression model is constructed in combination with the sound wave absorption characteristics of the corresponding environmental monitoring point to analyze the sound source identification status of each environmental monitoring point.

4. The distributed optical fiber sound source identification and classification method integrating the i-vector algorithm according to claim 3, characterized in that: In step S2, a multinomial regression model is constructed by combining the sound wave absorption characteristics of each environmental detection point and the background noise intensity coefficient to analyze the sound source identification status. The specific steps are as follows: Calculate influencing factors: Multiply the sound wave absorption characteristics with a preset first influencing weight to generate the first influencing factor; multiply the square of the background noise intensity coefficient with a preset second influencing weight to generate the second influencing factor; Generate the sound source identification status index: sum the first influencing factor and the second influencing factor of each environmental detection point to obtain the sound source identification status index of the corresponding environmental detection point; Target location selection: Among all environmental detection points, select the environmental detection point corresponding to the minimum sound source recognition state index as the target location.

5. The distributed optical fiber sound source identification and classification method integrating the i-vector algorithm according to claim 1, characterized in that: In step S3, the optical fiber sound source information to be identified is played at the target location and stored in the sound source database through the optical fiber acoustic wave sensing device. The information of the fiber optic sound source to be identified includes the types of fiber optic sound sources stored and the total storage space occupied by each type. The different types of fiber optic sound sources are counted, and the count results are used as the number of storage types. The storage space occupied by all types of fiber optic sound sources is accumulated to obtain the total storage space.

6. The distributed optical fiber sound source identification and classification method integrating the i-vector algorithm according to claim 5, characterized in that: In step S3, the number of storage types and the total storage space are standardized and then the geometric mean method is used to obtain the identification range score of the optical fiber sound source information to be identified. The score of the range to be identified for the optical fiber sound source information to be identified is compared with the preset range threshold. If the score of the identification range of the optical fiber sound source information to be identified exceeds the range threshold, the identification range of the optical fiber sound source information to be identified will be marked as large. If the score of the identification range of the optical fiber sound source information to be identified is lower than the range threshold, then the identification range of the optical fiber sound source information to be identified will be marked as small.

7. The distributed optical fiber sound source identification and classification method integrating the i-vector algorithm according to claim 1, characterized in that: In step S4, the fiber optic sound source demodulation and acquisition instrument acquires the release volume of the fiber optic sound source information of the target location at multiple time points within the set acquisition time window.

8. The distributed optical fiber sound source identification and classification method integrating the i-vector algorithm according to claim 7, characterized in that: In step S4, the standard deviation of the release volume of the optical fiber sound source information to be identified at each time point is calculated as the volume fluctuation of the optical fiber sound source information to be identified based on the release volume of the optical fiber sound source information to be identified at each time point.

9. A distributed optical fiber sound source identification and classification method incorporating the i-vector algorithm according to claim 8, characterized in that: In step S4, the range to be identified and the volume fluctuation are defined as input variables, and they are divided into different fuzzy sets respectively; The input result of the optical fiber sound source information to be identified is defined as the output variable, and it is divided into a fuzzy set. Formulate fuzzy rules to describe the range to be identified and the impact of volume fluctuations on the input results of the fiber optic sound source information to be identified; Fuzzy reasoning is performed based on fuzzy rules to determine the input result of the optical fiber sound source information to be identified.

10. A distributed optical fiber sound source identification and classification method incorporating the i-vector algorithm according to claim 9, characterized in that: In step S4, the input result of the optical fiber sound source information to be identified includes inputting the optical fiber sound source information to be identified into the i-vector algorithm for processing and not inputting the optical fiber sound source information to be identified into the i-vector algorithm for processing.

Images