CN117727332B - Ecological population assessment method based on language spectrum feature analysis - Google Patents
Ecological population assessment method based on language spectrum feature analysis Download PDFInfo
- Publication number
- CN117727332B CN117727332B CN202410179279.8A CN202410179279A CN117727332B CN 117727332 B CN117727332 B CN 117727332B CN 202410179279 A CN202410179279 A CN 202410179279A CN 117727332 B CN117727332 B CN 117727332B
- Authority
- CN
- China
- Prior art keywords
- data point
- spectrogram
- ringing
- masking
- ecological
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001228 spectrum Methods 0.000 title claims abstract description 85
- 238000000034 method Methods 0.000 title claims abstract description 75
- 238000004458 analytical method Methods 0.000 title claims abstract description 31
- 238000012544 monitoring process Methods 0.000 claims abstract description 103
- 230000000873 masking effect Effects 0.000 claims abstract description 92
- 230000005236 sound signal Effects 0.000 claims abstract description 72
- 238000011156 evaluation Methods 0.000 claims abstract description 42
- 230000003595 spectral effect Effects 0.000 claims abstract description 32
- 230000008569 process Effects 0.000 claims description 17
- 238000010606 normalization Methods 0.000 claims description 11
- 238000009825 accumulation Methods 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 description 12
- 238000012545 processing Methods 0.000 description 10
- 238000004364 calculation method Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 241000894007 species Species 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 3
- 238000002790 cross-validation Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000001788 irregular Effects 0.000 description 3
- 230000005237 high-frequency sound signal Effects 0.000 description 2
- 230000005923 long-lasting effect Effects 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000002759 z-score normalization Methods 0.000 description 1
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The application relates to the technical field of language spectrum feature analysis, and provides an ecological population assessment method based on language spectrum feature analysis, which comprises the following steps: acquiring a spectrogram of an ecological audio signal, constructing a spectrum characteristic window by the center of each data point in the spectrogram, and constructing a natural ringing index according to the connected domain characteristics, the energy gradient amplitude and the energy value characteristics of the data points in the spectrum characteristic window of each data point in the spectrogram; and constructing a masking monitoring window by taking each data point in the spectrogram as a center, calculating a bird song enhancement factor according to the distribution characteristics of the data points in the masking monitoring window, calculating a natural song enhancement index according to the natural song index and the bird song enhancement factor, constructing a weight coefficient according to the natural song enhancement index, and acquiring an ecological population evaluation result by utilizing a spectral clustering algorithm based on the weight coefficient. According to the application, the weight coefficient in the spectral clustering algorithm is obtained by constructing the natural ringing enhancement index, so that the accuracy of ecological population evaluation based on the spectral clustering algorithm is improved.
Description
Technical Field
The application relates to the technical field of language spectrum feature analysis, in particular to an ecological population assessment method based on language spectrum feature analysis.
Background
Biodiversity is an important component of the ecosystem, having a key role in maintaining the stability and function of the ecosystem. By monitoring the biodiversity, the number, distribution, structure and interrelation of different species in the ecosystem can be understood, thereby assessing the health of the ecosystem. The biodiversity monitoring can provide scientific basis for formulating and adjusting biodiversity protection policies and management measures. By monitoring and analyzing data such as species richness, population quantity, trend and the like, the areas of endangered species and damaged ecosystems can be determined, so that corresponding protection measures are adopted; it can also help us understand the impact of environmental changes on biodiversity. With climate change, land use change and expansion of human activity, biodiversity is facing serious threat. By monitoring the indexes such as species migration, distribution range change and the like, the influence of environmental change on species adaptation and ecological system functions can be known. Therefore, there is an urgent need to develop rapid biodiversity monitoring, evaluation, protection, management, planning, and the like.
A spectrogram is an image describing the time-frequency-energy variation of sound, and is an important method of sound analysis. Because birds are active sounding animals and are sensitive to the environment, the ecological characteristics can be analyzed by extracting the characteristics in the spectrogram after converting the bird audio signals into the spectrogram, the traditional method for extracting the spectrogram characteristics, such as a spectral clustering algorithm, has strong adaptability, namely has strong adaptability to complex clustering problems in the ecological audio spectrogram, but the effect of the algorithm is sensitive to parameter selection, such as the clustering result of the spectral clustering algorithm is highly dependent on the weight among data points, and because the frequency spectrum structure of the bird audio signals is complex, the bird audio signals have obvious timeliness, the weight is not easy to determine, the adjacency matrix cannot accurately reflect the real similarity among the data points, so that the clustering result is distorted, and the accuracy of ecological characteristic analysis is affected.
Disclosure of Invention
The application provides an ecological population assessment method based on language spectrum feature analysis, which aims to solve the problem of low accuracy of ecological population analysis by using language spectrum features, and adopts the following technical scheme:
One embodiment of the application provides an ecological population assessment method based on language spectrum feature analysis, which comprises the following steps:
Acquiring an ecological audio signal, converting the ecological audio signal into a spectrogram, carrying out graying treatment on the spectrogram by adopting a gray level average method, and taking the grayed spectrogram as the spectrogram of each ecological audio signal;
A spectrum characteristic window is constructed by taking each data point in the spectrogram as a center, a normal ringing index and a characteristic ringing index are respectively obtained according to the energy value distribution characteristic and the connected domain characteristic in the spectrum characteristic window, and the natural ringing index of each data point in the spectrogram is calculated based on the normal ringing index and the characteristic ringing index;
constructing a masking monitoring window by taking each data point in the spectrogram as a center, and calculating a bird song enhancement factor of each data point in the spectrogram according to the region growth characteristics of the central data point in the masking monitoring window;
Calculating a natural ringing enhancement index according to the natural ringing index and the bird ringing enhancement factor of each data point in the spectrogram, calculating a weight coefficient according to the difference of the natural ringing enhancement indexes between different data points in the spectrogram, and acquiring a clustering result of the spectrogram by using a spectral clustering algorithm based on the weight coefficient;
And obtaining an evaluation result of the ecological population according to the clustering result of the spectrogram.
Preferably, the method for respectively obtaining the normal ringing index and the characteristic ringing index according to the energy value distribution characteristic and the connected domain characteristic in the spectrum characteristic window and calculating the natural ringing index of each data point in the spectrogram based on the normal ringing index and the characteristic ringing index comprises the following steps:
For a spectrum characteristic window of each data point in a spectrogram, acquiring energy gradient amplitude values of all data points in the spectrum characteristic window and edge data points in the spectrum characteristic window by adopting a Sobel algorithm, acquiring connected domains in the spectrum characteristic window by adopting a connected domain analysis algorithm based on the edge data points in the spectrum characteristic window, acquiring a minimum circumscribed rectangle of each connected domain in the spectrum characteristic window, acquiring a central skeleton of each connected domain in the spectrum characteristic window by adopting a skeleton extraction algorithm, and acquiring fitting goodness of all points in the central skeleton by adopting a curve fitting algorithm;
Calculating a normal ringing index according to the minimum circumscribed rectangle, the edge energy value and the energy gradient amplitude of the connected domain in the spectrum characteristic window of each data point in the spectrogram;
Calculating a characteristic ringing index according to the fitting goodness of all points in the central skeleton of the connected domain in the spectrum characteristic window of each data point in the spectrogram and the length of the minimum circumscribed rectangle;
and taking the sum of the normal ringing index and the characteristic ringing index corresponding to each data point in the spectrogram as the natural ringing index of each data point.
Preferably, the method for calculating the normal ringing index according to the minimum circumscribed rectangle, the edge energy value and the energy gradient amplitude of the connected domain in the spectrum characteristic window of each data point in the spectrogram comprises the following steps:
For each edge data point in a spectrum characteristic window of each data point in the spectrogram, taking the product of the energy value of the edge data point and the corresponding energy gradient amplitude as a molecule, taking the product of the length and width ratio of the minimum circumscribed rectangle of the connected domain where the edge data point is located and the length of the minimum circumscribed rectangle as a denominator, and taking the average value of the accumulation result of the ratio of the molecule and the denominator on all the edge data points in the spectrum characteristic window as the normal ringing index of each data point.
Preferably, the method for calculating the characteristic ringing index according to the goodness of fit of all points in the central skeleton of the connected domain in the spectrum characteristic window of each data point in the spectrogram and the length of the minimum circumscribed rectangle comprises the following steps:
And for a spectrum characteristic window of each data point in the spectrogram, taking the product of the fitting goodness corresponding to each connected domain in the spectrum characteristic window and the length of the minimum circumscribed rectangle as an accumulation factor, and taking the average value of the accumulation result of the accumulation factor on the spectrum characteristic window as the characteristic ringing index of each data point.
Preferably, the method for calculating the bird song enhancement factor of each data point in the spectrogram according to the region growth characteristics of the central data point in the masking monitoring window comprises the following steps:
For a masking monitoring window of each data point in the spectrogram, taking a central data point in the masking monitoring window as an initial seed point, acquiring a region growing result of the masking monitoring window by adopting a region growing algorithm based on the initial seed point, taking a region where the initial seed point is positioned as a suspected high-frequency ringing region, and taking the angles of all growing directions of each data point in the suspected high-frequency ringing region in the masking monitoring window as a growing sequence according to a sequence formed by a sequence from small to large;
in the process of region growth, taking the difference value between the maximum value and the minimum value of the frequencies corresponding to all data points in the growth region when the data points in the suspected high-frequency ringing region in the masking monitoring window are grown as the growth frequency bandwidth;
calculating a ringing matching index according to a growth sequence corresponding to each data point in a masking monitoring window of each data point in the spectrogram and energy values of different areas in the masking monitoring window;
And calculating a bird song enhancement factor according to the song matching index and the growth frequency bandwidth corresponding to each data point in the masking monitoring window of each data point in the spectrogram.
Preferably, the method for calculating the chirp matching index according to the growth sequence corresponding to each data point in the masking monitoring window of each data point in the spectrogram and the energy values of different areas in the masking monitoring window comprises the following steps:
for a masking monitoring window of each data point in the spectrogram, taking a difference value of a preset parameter and each element in a growth sequence of each data point in the masking monitoring window as a first characteristic coefficient of each element, and taking a minimum value of each element in the growth sequence of each data point in the masking monitoring window and a corresponding first characteristic coefficient as a longitudinal growth length of each element;
Taking the reciprocal of the sum of the longitudinal growth length of each element in the growth sequence of each data point in the masking monitoring window and the first preset parameter as a first matching factor, taking the sum of the first matching factor and the second preset parameter as a base, taking the difference between the average value of the energy values of all the data points in the masking monitoring window and the average value of the energy values of all the data points in the suspected high-frequency ringing region in the masking monitoring window as an index, taking the mapping result of the base on the index as a second matching factor, and taking the accumulated result of the second matching factor on the growth sequence of each data point in the masking monitoring window as the ringing matching index of each data point in the masking monitoring window.
Preferably, the specific method for calculating the bird song enhancement factor according to the song matching index and the growth frequency bandwidth corresponding to each data point in the masking monitoring window of each data point in the spectrogram comprises the following steps:
For a masking monitoring window of each data point in the spectrogram, taking the energy value of the central data point in the masking monitoring window as a first judgment coefficient, and taking the average value of the energy values of all data points in a suspected high-frequency ringing area in the masking monitoring window as a second judgment coefficient;
If the first judgment coefficient corresponding to the masking monitoring window is larger than or equal to the second judgment coefficient, taking the accumulated result of the ratio of the ringing matching index of each data point in the suspected high-frequency ringing area in the masking monitoring window to the growth frequency bandwidth on the suspected high-frequency ringing area in the masking monitoring window as a bird-ringing enhancement factor;
And if the first judgment coefficient corresponding to the masking monitoring window is smaller than the second judgment coefficient, taking the number of data points in the suspected high-frequency ringing region in the masking monitoring window as a bird song enhancement factor.
Preferably, the method for calculating the natural ringing enhancement index according to the natural ringing index and the bird ringing enhancement factor of each data point in the spectrogram comprises the following steps:
For each data point in the spectrogram, taking the sum of the normalization result of the bird song enhancement factor corresponding to the data point and the preset parameter as a product factor of the data point, and taking the product of the product factor and the natural song enhancement index corresponding to the data point as the natural song enhancement index of the data point.
Preferably, the method for calculating the weight coefficient according to the difference of natural ringing enhancement indexes between different data points in the spectrogram and obtaining the clustering result of the spectrogram by using a spectral clustering algorithm based on the weight coefficient comprises the following steps:
Calculating the absolute value of the difference between natural ringing enhancement indexes of any two data points in the spectrogram, taking the reciprocal of the sum of the absolute value and preset parameters as a weight coefficient between any two data points, taking the weight coefficient between the data points in the spectrogram as the weight between different nodes in a spectral clustering algorithm, and obtaining the clustering result of the spectrogram by using the spectral clustering algorithm.
Preferably, the method for obtaining the evaluation result of the ecological population according to the clustering result of the spectrogram comprises the following steps:
The method comprises the steps of taking the number of clusters in a clustering result of a spectrogram as ecological population characteristic values of ecological audio signals corresponding to the spectrogram, taking a set formed by ecological population characteristic values of all the ecological audio signals acquired in one day in an ecological population monitoring area as an ecological population evaluation set, adopting a normalization algorithm to obtain normalization results of all elements in the ecological population evaluation set, and taking the average value of the normalization results of all the elements in the ecological population evaluation set as an ecological population evaluation coefficient;
And taking a sequence formed by all the ecological population evaluation coefficients corresponding to the ecological population monitoring area in a preset time period according to the time sequence as an ecological population evaluation coefficient sequence, taking the average value of the ecological population evaluation coefficient sequence as an ecological population evaluation characteristic value, and taking the judgment result of the ecological population evaluation characteristic value in the preset interval as an evaluation result of the ecological population.
The beneficial effects of the application are as follows: the method comprises the steps of converting an ecological audio signal into a spectrogram, acquiring characteristics of the bird audio signal in the spectrogram according to analysis results of characteristics of frequency and duration of bird sounds in the ecological audio signal, constructing a natural sounding index according to the characteristics of the bird audio signal in the spectrogram, reflecting the degree that data points in the spectrogram accord with bird sounds characteristics according to the natural sounding index, and more accurately identifying the data points accord with bird sounds characteristics; the method has the advantages that the improved weight coefficient can reflect the correlation degree among bird characteristic data points more truly, the analysis of the clustering spectrogram characteristics is carried out accurately through the spectral clustering algorithm, the ecological population is evaluated and analyzed according to the clustering analysis result, and the accuracy of ecological population analysis through the spectral feature is improved.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the application, and that other drawings can be obtained according to these drawings without inventive faculty for a person skilled in the art.
FIG. 1 is a schematic flow chart of an ecological population assessment method based on semantic analysis according to an embodiment of the present application;
Fig. 2 is a schematic diagram of an implementation process for obtaining an evaluation result of an ecological population according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Referring to fig. 1, a flowchart of an ecological population assessment method based on a language spectrum feature analysis according to an embodiment of the present application is shown, and the method includes the following steps:
And S001, acquiring a spectrogram of the ecological audio signal.
An ecological audio signal acquisition point is arranged in an ecological population monitoring area, audio acquisition equipment is arranged at the ecological audio signal acquisition point, and an ecological audio signal of the ecological monitoring area is acquired through the audio acquisition equipment. In particular, for exampleThe region is a ecological population monitoring region, pair/>Monitoring and analyzing ecological population in region, at/>Regional setting/>(The size is 100 tested values) the ecological audio signal acquisition points, a microphone array formed by four paths of sound pick-up devices is arranged at each ecological audio signal acquisition point, wherein the microphone array formed by the four paths of sound pick-up devices can effectively acquire audio signals in a round area with the diameter of 200m, and the sampling frequency is/>(Size-checked value 44.1 kHz), and the duration of each acquisition of an audio signal is/>(Size checked 1 h); because the activity time of different organisms is different, the acquisition frequency of each ecological audio signal acquisition point is/>(Size tested 3) and the audio signal is acquired at different time points, e.g. each of the ecological signal acquisition points acquires the audio signal once per day at 0:00, 8:00, 16:00.
Further, due toThe environment in the region may be complex, so that noise signals such as wind noise and rain noise may exist in the collected original ecological audio signals, in order to reduce the influence of the noise signals on subsequent audio processing, the collected ecological audio signals are subjected to denoising processing, the common audio denoising technology includes wavelet transform denoising, a spectrum subtraction algorithm, adaptive filtering denoising and the like, in order to adapt to the noise of different environments, a better denoising effect is achieved, the collected ecological audio signals are subjected to denoising processing by adopting an adaptive filtering algorithm, the adaptive filtering algorithm is a known technology, and a detailed description is omitted in a specific calculation process.
Further, each ecological audio signal is converted into a spectrogram for analysis, specifically, the ecological audio signal is subjected to pre-emphasis processing and frame windowing processing in sequence, wherein the frame length is 20ms, the frame is shifted to 10ms, the window function is a hamming window function, the signals subjected to frame windowing processing are used as short-time Fourier transform input, the signals are output as spectrograms of each frame of signals, the spectrograms are sequentially rotated and mapped, all the converted frame spectrums are spliced to form the spectrogram, meanwhile, in order to improve the accuracy of extracting ecological features from the spectrogram in the subsequent steps and improve the signal quality, the obtained spectrogram is required to be subjected to denoising processing, the spectrogram is subjected to denoising processing by adopting spectral subtraction, the denoised spectrogram is obtained, the structure and the pattern in the salient image are more obvious, the characteristics under certain specific time and frequency are more obvious, the spectrogram is subjected to graying processing by adopting a gray-level average method, the graying spectrogram is obtained, the graying spectrogram is used as the spectrogram of each ecological audio signal, the Fourier transform, the spectral subtraction and the average method is known in a short-time, and the repeated implementation is not repeated.
Thus, a spectrogram of each of the eco-audio signals is acquired.
Step S002, a spectrum characteristic window is constructed by the center of each data point in the spectrogram, and a natural ringing index is constructed according to the connected domain characteristics, the energy gradient amplitude and the energy value characteristics of the data points in the spectrum characteristic window of each data point in the spectrogram.
The abscissa in the spectrogram of each eco-audio signal represents time, the ordinate represents frequency, and the energy value of each data point in the spectrogram represents the energy value at the corresponding frequency and time. The bird audio features of the local area of the spectrogram are analyzed according to the frequency and energy features of the bird audio.
Specifically, the bird sounds generally have two states, one is irregular sounds, the sounds are sharp due to the fact that the sound-producing organs of the birds are small, the audio signals of the bird sounds are high in frequency and small in frequency range, bright thin lines are displayed in the spectrogram, and the high-frequency bright lines in the spectrogram are short due to the fact that the duration of the bird sounds is short; the other is regular singing, such as regular long-lasting sounds generated during the actions of coupling, warning and the like, and the regular long-lasting sounds are shown in a spectrogram. And constructing a natural ringing index based on the spectrogram of the ecological audio signal, and reflecting the degree that the data points in the spectrogram belong to the audio characteristics of birds through the natural ringing index.
Specifically, for each spectrogram of the ecological audio signal, each data point in the spectrogram is taken as the center to constructA rectangular window of size, the rectangular window being taken as a spectral signature window for each data point, wherein the length/> -of the spectral signature window for each data pointSum width/>The length of the spectral feature window input for each data point is determined by cross-validationCandidate value/>Sum width/>Candidate value/>Obtaining the optimal length/>, of each spectral feature window by adopting a cross-validation methodSum width/>The specific calculation process of the cross-validation method is a well-known technique and will not be described in detail.
Further, for a spectrum characteristic window of each data point in the spectrogram, inputting all data points in the spectrum characteristic window, acquiring energy gradient amplitude values of all data points in the spectrum characteristic window and edge data points in the spectrum characteristic window by adopting a Sobel algorithm, and acquiring a connected domain in the spectrum characteristic window by adopting a connected domain analysis algorithm based on the edge data points in the spectrum characteristic window. Further, inputting each connected domain in a spectrum characteristic window, and acquiring all data points of a central skeleton of each connected domain by adopting a K3M skeleton extraction algorithm; inputting all data points of the central skeleton of each connected domain in the spectrum characteristic window, and acquiring the goodness of fit of all data points of the central skeleton of each connected domain in the spectrum characteristic window by adopting a least square method. Further, the minimum circumscribed rectangle algorithm is adopted to obtain the minimum circumscribed rectangle of each connected domain in the spectrum characteristic window, and the specific calculation processes of the Sobel algorithm, the connected domain analysis algorithm, the K3M skeleton extraction algorithm and the least square method are all known techniques and are not repeated.
Further, a normal ringing index and a characteristic ringing index are obtained according to the energy value distribution characteristic and the connected domain characteristic in the spectrum characteristic window, and the natural ringing index of each data point in the spectrogram is calculated based on the normal ringing index and the characteristic ringing index, specifically, for example, the first point in the spectrogramThe natural ringing index of the data points was obtained as follows:
In the method, in the process of the invention, Representing the/>, in the spectrogramNormal ringing index for each data point; /(I)And/>Respectively represent the/>, in the spectrogramFirst/>, in spectral feature window of data pointsThe energy gradient magnitude and energy value of the individual edge data points; /(I)Representing the/>, in the spectrogramFirst/>, in spectral feature window of data pointsThe ratio of the length to the width of the smallest circumscribed rectangle of the connected domain where the edge data points are located; /(I)Representing the/>, in the spectrogramFirst/>, in spectral feature window of data pointsThe length of the smallest circumscribed rectangle of the connected domain where the edge data points are located; /(I)Representing the/>, in the spectrogramThe number of edge data points in the spectral feature window of the data points;
representing the/>, in the spectrogram Characteristic ringing index of data points; /(I)Representing the/>, in the spectrogramFirst/>, in spectral feature window of data pointsFitting goodness of curve fitting corresponding to the connected domains; /(I)Representing the/>, in the spectrogramFirst/>, in spectral feature window of data pointsThe length of the smallest circumscribed rectangle of each connected domain; /(I)Representing the/>, in the spectrogramThe number of connected domains in the spectral feature window of the data points;
representing the/>, in the spectrogram Natural ringing index of data points.
The first of the speech patternsThe larger the energy gradient amplitude and energy value of the edge data point in the spectrum characteristic window of the data point, the calculated/>The larger the value of (2) is, the more/>, the expression in the spectrogram isThe point in time at which the data point is located may be a sharp, high frequency sound signal; simultaneous speech and spectrogram of the first and second aspectsThe ratio of the length to the width of the smallest circumscribed rectangle of the connected domain in the spectrum characteristic window of the data point and the length of the smallest circumscribed rectangle are smaller, and the/>The smaller the value of (2) is, the more/>, the expression in the spectrogram isSharp, high frequency sound signal at the point in time of the data point, i.e./>The larger the value of (2), the greater the/>, in the calculated spectrogramNormal ringing index of individual data points/>The larger the value of (2) is, the more/>, the expression in the spectrogram isThe more the audio signal characteristics of the time point where the data points are located are in line with the characteristics of irregular bird song.
Further, if the first is in the spectrogramThe larger the length of the smallest circumscribed rectangle of the connected domain in the spectrum characteristic window of the data point is, the calculated/>The larger the value of (2) is, the more/>, the expression in the spectrogram isThe longer the regular beep duration at the point in time where the data point is located; simultaneous speech and spectrogram of the first and second aspectsThe higher the goodness of fit when fitting the data points on the central skeleton of the connected domain in the spectral feature window of the data points, namely/>The larger the value of (2) is, the more/>, the expression in the spectrogram isThe time points at which the data points are located may have regular and periodic acoustic signature, i.e. accumulation factor/>The larger the value of (2), the greater the/>, in the calculated spectrogramData points for evidence ringing index/>The larger the value of (2) is, the more/>, the expression in the spectrogram isThe more the audio signal characteristics of the time point where the data points are located are in line with the characteristics of regular bird beeps.
Further, if the first is in the spectrogramThe greater the possibility that the audio signal characteristics of the time point where the data point is located show irregular bird song or the characteristics of regular bird song, the more/>, in the calculated spectrogramThe greater the natural ringing index of the data points.
Thus, the natural ringing index of each data point in the spectrogram is obtained.
And S003, constructing a masking monitoring window by taking each data point in the spectrogram as a center, calculating a bird song enhancement factor according to the distribution characteristics of the data points in the masking monitoring window, and calculating a natural song enhancement index according to the natural song index and the bird song enhancement factor.
At the position ofThere are various human activity disturbances in the process of monitoring sounds over time in a region, such as sounds of vehicles and mechanical motors of logging tools where human activity may be present, etc., in particular, when/>The microphone array formed by four paths of pickups of the ecological audio signal collection points can collect background noise generated in the processes of vehicle and felling, and the background noise is louder than the sound of birds.
Furthermore, if the ecological audio signal collection points have larger background sound and bird song sound at the same time when the sound signals are collected, the masking effect of the sound possibly occurs, namely, the bird song sound is masked by the background noise, and the difficulty of identifying the bird song sound through the spectrogram is larger at the moment, so that the natural song index of each data point in the spectrogram of each ecological audio signal needs to be corrected, and the audio signal characteristics of the bird song in the spectrogram are more accurately reflected.
Further, for each spectrogram of the ecological audio signal, each data point in the spectrogram is taken as the center to constructThe rectangular window is used as a masking monitoring window of each data point, strong background noise and weak bird song can be overlapped when masking effect occurs, the energy value of the data point is higher when a bird song audio signal exists in the masking monitoring window of each data point than that of the data point when the bird song audio signal does not exist, and meanwhile, the sound is stable when the bird song is generated, so that the difference of the energy values of the data points is smaller when the bird song audio signal exists in a spectrogram, and the bird song enhancement factor of each data point in the spectrogram is calculated according to the characteristics of the masking effect in the ecological audio signal.
Specifically, for a masking monitoring window of each data point in a spectrogram of each ecological audio signal, a central data point of the masking monitoring window is used as an initial seed point, all data points in the masking monitoring window are input, a region growing algorithm is adopted to obtain a region growing result of the masking monitoring window, wherein the growing condition is that the energy value of the data point in an 8-neighborhood is larger than or equal to the average value of the energy values of all the data points in the masking monitoring window, the stopping condition is that the data point in the 8-neighborhood does not exist data points meeting the growing condition, the growing area where the central data point in the masking monitoring window is located is used as a suspected high-frequency ringing area, and the angles of all growing directions of each data point in the suspected high-frequency ringing area in the masking monitoring window are used as a growing sequence according to a sequence consisting of small to large orders.
Specifically, for example, the growth direction of the initial seed point is 0 °, 45 °, 135 °, the growth sequence corresponding to the initial seed point is; Further, in the process of the region growth, taking the difference value between the maximum value and the minimum value of the frequencies corresponding to all data points in the growth region when the data points are grown to each data point in the suspected high-frequency ringing region in the masking monitoring window as the growth frequency bandwidth; since birds are in a smaller frequency variation range and possibly longer duration when they are ringing, the growth direction is mostly horizontal when they grow from the initial seed point, and the bird ringing enhancement factor is calculated according to the region growth characteristics in the masking window, and the specific calculation process is as follows:
In the method, in the process of the invention, The/>, in the spectrogram representing the eco-audio signalFirst/>, in suspected high-frequency ringing area in data point masking monitoring windowGrowth sequence of data points/>Longitudinal length of the individual elements,/>Representing a minimum function,/>The/>, in the spectrogram representing the eco-audio signalFirst/>, in suspected high-frequency ringing area in data point masking monitoring windowGrowth sequence of data points/>Values of individual elements,/>Is a first characteristic coefficient;
the/>, in the spectrogram representing the eco-audio signal First/>, in suspected high-frequency ringing area in data point masking monitoring windowData point ringing match index,/>The/>, in the spectrogram representing the eco-audio signalFirst/>, in suspected high-frequency ringing area in data point masking monitoring windowNumber of elements in growth sequence of data points,/>Representing the adjustment parameters, and taking a checked value 45; /(I)The/>, in the spectrogram representing the eco-audio signalA mean of energy values for all data points within a masked monitoring window of data points; /(I)The/>, in the spectrogram representing the eco-audio signalThe average value of the energy values of all data points in the suspected high-frequency ringing area in the masking monitoring window of the data points;
the/>, in the spectrogram representing the eco-audio signal A data point of a bird song enhancement factor; /(I)The/>, in the spectrogram representing the eco-audio signalThe number of data points in the suspected high-frequency ringing area in the masking monitoring window of the data points; Representing the/>, in a spectrogram of an eco-audio signal First/>, within a masked monitoring window of data pointsGrowth frequency bandwidth of data points,/>The/>, in the spectrogram representing the eco-audio signalEnergy values of data points.
If the speech spectrogram of the ecological audio signal is the firstThe data points in the suspected high-frequency ringing area in the masking monitoring window of the data points grow along the time direction, and the calculated/>The smaller the value of the first matching factor/>The larger the value of (a), the larger the average value of the energy value of the data point in the suspected high-frequency ringing region is, namelyThe larger the value of (2) the second matching factor/>The larger the value of (2) is, the/>, the calculated the spectrogram of the ecological audio signal isAlarm matching index of data points in suspected high-frequency alarm region in data point masking monitoring windowThe greater the value of (c) is, the more likely the data points in the suspected high frequency beep region are data points of the audio signature of the presence of bird beeps when the masking effect occurs.
Further, if the frequency variation amplitude corresponding to the data points in the region growing process is smaller, the calculated spectrogram of the ecological audio signal is the firstGrowth frequency bandwidth of data points within a masking monitoring window of data points/>The smaller the value of (2), the/>, in the spectrogram of the physiological audio signalIn the case that the difference between the energy value of the data point and the average value of all the data points in the suspected high-frequency ringing region is large, that is, the first judgment coefficient is greater than or equal to the second judgment coefficient (/ >)) Calculated/>The larger the value of (a) is, namely the/>, in the spectrogram of the ecological audio signalData point bird song enhancement factorThe larger the value of (2) is, the greater the/>, in the spectrogram representing the eco-audio signalThe smaller the frequency range corresponding to the suspected high-frequency beeping area in the masking monitoring window of the data point is, the more possible audio features of bird beeping are present.
Further, in the spectrogram of the ecological audio signalIn the case that the difference between the energy value of the data point and the average value of all the data points in the suspected high-frequency ringing region is small, i.e. the first judgment coefficient is smaller than the second judgment coefficient (/ >)) The/>, in the spectrogram of the ecological audio signalThe number of data points in the suspected high-frequency ringing area in the masking observation window of the data points is used as the/>, in the spectrogram of the ecological audio signalData points of the bird song enhancement factor/>If the/>, in the spectrogram of the ecological audio signalThe larger the range of the suspected high-frequency ringing region in the masking observation window of the data point is, the calculated/>The larger the value of (2) is, the more/>, represented in the spectrogram of the eco-audio signalThe smaller the difference in the energy value of the data points from the average of all data points in the suspected high frequency beeping area, the greater the likelihood of masking the audio features of bird beeps in the monitoring window.
Further, the natural ringing index of each data point is enhanced according to the bird song enhancement factor of each data point in the spectrogram of the ecological audio signal, and the enhanced result is used as the natural ringing enhancement index of each data point, and the specific calculation formula is as follows:
In the method, in the process of the invention, The/>, in the spectrogram representing the eco-audio signalNatural ringing enhancement index for data points; /(I)Representing a Z-score normalization function,/>The/>, in the spectrogram representing the eco-audio signalData points of the bird song enhancement factor,/>The/>, in the spectrogram representing the eco-audio signalNatural ringing index of data points.
If the speech spectrogram of the ecological audio signal is the firstThe greater the bird song enhancement factor of the data point, i.e. the product factorThe larger the value of (2), the more/>, in the spectrogram of the ecological audio signalThe greater the enhancement degree of the natural ringing index of the data point, namely the/>, in the calculated spectrogram of the obtained ecological audio signalData point natural ringing enhancement indexThe greater the value of (2).
Thus, the natural ringing enhancement index of each data point in the spectrogram of each ecological audio signal is obtained.
And S004, constructing a weight coefficient according to the natural ringing enhancement index, and acquiring an ecological population evaluation result by using a spectral clustering algorithm based on the weight coefficient.
And obtaining the weight coefficient between different data points in the spectrogram according to the natural ringing enhancement index of each data point in the spectrogram of each ecological audio signal. Specifically, the natural ringing enhancement index of each data point in the spectrogram of each ecological audio signal reflects the possibility that the bird song appears at the corresponding time point of the data point, so the weight coefficient is calculated through the difference of the natural ringing enhancement indexes among the data points in the spectrogram of each ecological audio signal, and a specific calculation formula is as follows:
In the method, in the process of the invention, The/>, in the spectrogram representing the eco-audio signalData points and/>Weight coefficient between data points,/>And/>The/>, in the spectrograms respectively representing the ecological audio signalsSum/>Natural ringing enhancement index of data points,/>The regulating parameter is expressed, and the magnitude takes an empirical value of 1.
If the speech spectrogram of the ecological audio signal is the firstData points and/>The smaller the natural ringing enhancement index difference between data points, i.e. >, iThe smaller the one representing the/>, in the spectrogram of the ecological audio signalData points and/>The data points have similar audio signal characteristics, namely the/>, in the spectrogram of the calculated ecological audio signalData points and/>Weight coefficient between data points/>The larger the speech pattern representing the ecological audio signal is, the more/>Data points and/>The greater the likelihood that the data points will each be characterized by bird song audio signals or by environmental background audio signals.
Further, taking all data points in the spectrogram of each ecological audio signal and weight coefficients among the data points as inputs of a spectral clustering algorithm, outputting a clustering result of the data points in the spectrogram of each ecological audio signal, specifically, constructing an adjacent matrix according to the weight coefficients among the data points in the spectrogram of the ecological audio signal, wherein the adjacent matrix is the first oneLine 1Elements of the columns are the/>, in the spectrogram of the physiological audio signalData points and/>The weight coefficient among the data points takes the adjacent matrix as the input of the degree matrix calculation method and outputs the adjacent matrix as the degree matrix corresponding to the adjacent matrix; taking the difference result between the degree matrix and the adjacent matrix as a Laplacian matrix, and obtaining/>, through the Laplacian matrix(The size can be determined according to the dimension of the actual data) feature vectors based on/>The individual feature vector acquisition dimension is/>Is/>, for dimension by using Ncut graph cutting methodThe weighted undirected graph formed by the samples is subjected to graph cutting processing, a clustering result of data points in a spectrogram of the ecological audio signal is output, and the detailed calculation process of the spectral clustering algorithm is a known technology and is not repeated.
Further, the number of clusters in the clustering result of the spectrogram of each ecological monitoring audio signal is used as the ecological population characteristic value of each ecological audio signalThe method comprises the steps that a set formed by ecological population characteristic values of all ecological audio signals collected in the same day in a region is used as an ecological population evaluation set, a normalization algorithm is adopted to obtain normalization results of all elements in the ecological population evaluation set, and the average value of the normalization results of all elements in the ecological population evaluation set is used as/>Ecological population evaluation coefficients in the same day of the region; for better pair/>Evaluation of ecological populations in regions while avoiding longer periods of timeThe influence of the regional monitoring data on the ecological population evaluation error is stopped within one year/>, up to the current timeThe sequence of the ecological population evaluation coefficients corresponding to the region according to the time sequence is used as an ecological population evaluation coefficient sequence, the average value of the ecological population evaluation coefficient sequence is used as an ecological population evaluation characteristic value, and an ecological population evaluation interval is setEcological population evaluation characteristic value is in/>Respectively expressed/>The ecological population richness of the region is low, the ecological population richness is medium, the ecological population richness is high, and the specific acquisition/>The implementation process of the evaluation result of the regional ecological population is shown in figure 2.
So far, the evaluation result of the ecological population is obtained.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. The above description is only of the preferred embodiments of the present application and is not intended to limit the application, but any modifications, equivalent substitutions, improvements, etc. within the principles of the present application should be included in the scope of the present application.
Claims (9)
1. The ecological population assessment method based on the language spectrum feature analysis is characterized by comprising the following steps of:
Acquiring an ecological audio signal, converting the ecological audio signal into a spectrogram, carrying out graying treatment on the spectrogram by adopting a gray level average method, and taking the grayed spectrogram as the spectrogram of each ecological audio signal;
A spectrum characteristic window is constructed by taking each data point in the spectrogram as a center, a normal ringing index and a characteristic ringing index are respectively obtained according to the energy value distribution characteristic and the connected domain characteristic in the spectrum characteristic window, and the natural ringing index of each data point in the spectrogram is calculated based on the normal ringing index and the characteristic ringing index;
constructing a masking monitoring window by taking each data point in the spectrogram as a center, and calculating a bird song enhancement factor of each data point in the spectrogram according to the region growth characteristics of the central data point in the masking monitoring window;
Calculating a natural ringing enhancement index according to the natural ringing index and the bird ringing enhancement factor of each data point in the spectrogram, calculating a weight coefficient according to the difference of the natural ringing enhancement indexes between different data points in the spectrogram, and acquiring a clustering result of the spectrogram by using a spectral clustering algorithm based on the weight coefficient;
Acquiring an evaluation result of the ecological population according to the clustering result of the spectrogram;
The method for respectively obtaining the normal ringing index and the characteristic ringing index according to the energy value distribution characteristic and the connected domain characteristic in the spectrum characteristic window and calculating the natural ringing index of each data point in the spectrogram based on the normal ringing index and the characteristic ringing index comprises the following steps:
For a spectrum characteristic window of each data point in a spectrogram, acquiring energy gradient amplitude values of all data points in the spectrum characteristic window and edge data points in the spectrum characteristic window by adopting a Sobel algorithm, acquiring connected domains in the spectrum characteristic window by adopting a connected domain analysis algorithm based on the edge data points in the spectrum characteristic window, acquiring a minimum circumscribed rectangle of each connected domain in the spectrum characteristic window, acquiring a central skeleton of each connected domain in the spectrum characteristic window by adopting a skeleton extraction algorithm, and acquiring fitting goodness of all points in the central skeleton by adopting a curve fitting algorithm;
Calculating a normal ringing index according to the minimum circumscribed rectangle, the edge energy value and the energy gradient amplitude of the connected domain in the spectrum characteristic window of each data point in the spectrogram;
Calculating a characteristic ringing index according to the fitting goodness of all points in the central skeleton of the connected domain in the spectrum characteristic window of each data point in the spectrogram and the length of the minimum circumscribed rectangle;
and taking the sum of the normal ringing index and the characteristic ringing index corresponding to each data point in the spectrogram as the natural ringing index of each data point.
2. The ecological population assessment method based on the speech spectrum feature analysis according to claim 1, wherein the method for calculating the normal ringing index according to the minimum circumscribed rectangle, the edge energy value and the energy gradient amplitude of the connected domain in the spectrum feature window of each data point in the speech spectrum is as follows:
For each edge data point in a spectrum characteristic window of each data point in the spectrogram, taking the product of the energy value of the edge data point and the corresponding energy gradient amplitude as a molecule, taking the product of the length and width ratio of the minimum circumscribed rectangle of the connected domain where the edge data point is located and the length of the minimum circumscribed rectangle as a denominator, and taking the average value of the accumulation result of the ratio of the molecule and the denominator on all the edge data points in the spectrum characteristic window as the normal ringing index of each data point.
3. The ecological population assessment method based on the speech spectrum feature analysis according to claim 1, wherein the method for calculating the feature ringing index according to the goodness of fit and the length of the minimum circumscribed rectangle of all points in the central skeleton of the connected domain in the spectrum feature window of each data point in the speech spectrum is as follows:
And for a spectrum characteristic window of each data point in the spectrogram, taking the product of the fitting goodness corresponding to each connected domain in the spectrum characteristic window and the length of the minimum circumscribed rectangle as an accumulation factor, and taking the average value of the accumulation result of the accumulation factor on the spectrum characteristic window as the characteristic ringing index of each data point.
4. The method for ecological population assessment based on speech spectrum feature analysis according to claim 1, wherein the method for calculating the bird song enhancement factor of each data point in the speech spectrum according to the region growth feature of the central data point in the masking monitoring window is as follows:
For a masking monitoring window of each data point in the spectrogram, taking a central data point in the masking monitoring window as an initial seed point, acquiring a region growing result of the masking monitoring window by adopting a region growing algorithm based on the initial seed point, taking a region where the initial seed point is positioned as a suspected high-frequency ringing region, and taking the angles of all growing directions of each data point in the suspected high-frequency ringing region in the masking monitoring window as a growing sequence according to a sequence formed by a sequence from small to large;
in the process of region growth, taking the difference value between the maximum value and the minimum value of the frequencies corresponding to all data points in the growth region when the data points in the suspected high-frequency ringing region in the masking monitoring window are grown as the growth frequency bandwidth;
calculating a ringing matching index according to a growth sequence corresponding to each data point in a masking monitoring window of each data point in the spectrogram and energy values of different areas in the masking monitoring window;
And calculating a bird song enhancement factor according to the song matching index and the growth frequency bandwidth corresponding to each data point in the masking monitoring window of each data point in the spectrogram.
5. The ecological population assessment method based on the speech spectrum feature analysis according to claim 4, wherein the method for calculating the ringing matching index according to the growth sequence corresponding to each data point in the masking monitoring window of each data point in the speech spectrum and the energy values of different areas in the masking monitoring window is as follows:
for a masking monitoring window of each data point in the spectrogram, taking a difference value of a preset parameter and each element in a growth sequence of each data point in the masking monitoring window as a first characteristic coefficient of each element, and taking a minimum value of each element in the growth sequence of each data point in the masking monitoring window and a corresponding first characteristic coefficient as a longitudinal growth length of each element;
Taking the reciprocal of the sum of the longitudinal growth length of each element in the growth sequence of each data point in the masking monitoring window and the first preset parameter as a first matching factor, taking the sum of the first matching factor and the second preset parameter as a base, taking the difference between the average value of the energy values of all the data points in the masking monitoring window and the average value of the energy values of all the data points in the suspected high-frequency ringing region in the masking monitoring window as an index, taking the mapping result of the base on the index as a second matching factor, and taking the accumulated result of the second matching factor on the growth sequence of each data point in the masking monitoring window as the ringing matching index of each data point in the masking monitoring window.
6. The ecological population assessment method based on the semantic profile analysis according to claim 4, wherein the specific method for calculating the bird song enhancement factor according to the song matching index and the growth frequency bandwidth corresponding to each data point in the masking monitoring window of each data point in the semantic profile is as follows:
For a masking monitoring window of each data point in the spectrogram, taking the energy value of the central data point in the masking monitoring window as a first judgment coefficient, and taking the average value of the energy values of all data points in a suspected high-frequency ringing area in the masking monitoring window as a second judgment coefficient;
If the first judgment coefficient corresponding to the masking monitoring window is larger than or equal to the second judgment coefficient, taking the accumulated result of the ratio of the ringing matching index of each data point in the suspected high-frequency ringing area in the masking monitoring window to the growth frequency bandwidth on the suspected high-frequency ringing area in the masking monitoring window as a bird-ringing enhancement factor;
And if the first judgment coefficient corresponding to the masking monitoring window is smaller than the second judgment coefficient, taking the number of data points in the suspected high-frequency ringing region in the masking monitoring window as a bird song enhancement factor.
7. The ecological population assessment method based on the speech spectrum feature analysis according to claim 1, wherein the method for calculating the natural ringing enhancement index according to the natural ringing index and the bird ringing enhancement factor of each data point in the speech spectrum is as follows:
For each data point in the spectrogram, taking the sum of the normalization result of the bird song enhancement factor corresponding to the data point and the preset parameter as a product factor of the data point, and taking the product of the product factor and the natural song enhancement index corresponding to the data point as the natural song enhancement index of the data point.
8. The ecological population assessment method based on the speech spectrum feature analysis according to claim 1, wherein the method for calculating the weight coefficient according to the difference of natural ringing enhancement indexes among different data points in the speech spectrum and obtaining the clustering result of the speech spectrum by using a spectral clustering algorithm based on the weight coefficient is as follows:
Calculating the absolute value of the difference between natural ringing enhancement indexes of any two data points in the spectrogram, taking the reciprocal of the sum of the absolute value and preset parameters as a weight coefficient between any two data points, taking the weight coefficient between the data points in the spectrogram as the weight between different nodes in a spectral clustering algorithm, and obtaining the clustering result of the spectrogram by using the spectral clustering algorithm.
9. The ecological population assessment method based on the language spectrum feature analysis according to claim 1, wherein the method for obtaining the assessment result of the ecological population according to the clustering result of the language spectrum is as follows:
The method comprises the steps of taking the number of clusters in a clustering result of a spectrogram as ecological population characteristic values of ecological audio signals corresponding to the spectrogram, taking a set formed by ecological population characteristic values of all the ecological audio signals acquired in one day in an ecological population monitoring area as an ecological population evaluation set, adopting a normalization algorithm to obtain normalization results of all elements in the ecological population evaluation set, and taking the average value of the normalization results of all the elements in the ecological population evaluation set as an ecological population evaluation coefficient;
And taking a sequence formed by all the ecological population evaluation coefficients corresponding to the ecological population monitoring area in a preset time period according to the time sequence as an ecological population evaluation coefficient sequence, taking the average value of the ecological population evaluation coefficient sequence as an ecological population evaluation characteristic value, and taking the judgment result of the ecological population evaluation characteristic value in the preset interval as an evaluation result of the ecological population.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410179279.8A CN117727332B (en) | 2024-02-18 | 2024-02-18 | Ecological population assessment method based on language spectrum feature analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410179279.8A CN117727332B (en) | 2024-02-18 | 2024-02-18 | Ecological population assessment method based on language spectrum feature analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117727332A CN117727332A (en) | 2024-03-19 |
| CN117727332B true CN117727332B (en) | 2024-04-26 |
Family
ID=90203858
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410179279.8A Active CN117727332B (en) | 2024-02-18 | 2024-02-18 | Ecological population assessment method based on language spectrum feature analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117727332B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104658538A (en) * | 2013-11-18 | 2015-05-27 | 中国计量学院 | Mobile bird recognition method based on birdsong |
| CN109886214A (en) * | 2019-02-26 | 2019-06-14 | 中南民族大学 | A kind of chirm characteristic strengthening method based on image procossing |
| CN112908343A (en) * | 2019-11-19 | 2021-06-04 | 中国科学院声学研究所 | Acquisition method and system for bird species number based on cepstrum spectrogram |
| CN116524939A (en) * | 2023-04-23 | 2023-08-01 | 南京理工大学 | ECAPA-TDNN-based automatic identification method for bird song species |
| CN117095694A (en) * | 2023-10-18 | 2023-11-21 | 中国科学技术大学 | Bird song recognition method based on tag hierarchical structure attribute relationship |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10672414B2 (en) * | 2018-04-13 | 2020-06-02 | Microsoft Technology Licensing, Llc | Systems, methods, and computer-readable media for improved real-time audio processing |
-
2024
- 2024-02-18 CN CN202410179279.8A patent/CN117727332B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104658538A (en) * | 2013-11-18 | 2015-05-27 | 中国计量学院 | Mobile bird recognition method based on birdsong |
| CN109886214A (en) * | 2019-02-26 | 2019-06-14 | 中南民族大学 | A kind of chirm characteristic strengthening method based on image procossing |
| CN112908343A (en) * | 2019-11-19 | 2021-06-04 | 中国科学院声学研究所 | Acquisition method and system for bird species number based on cepstrum spectrogram |
| CN116524939A (en) * | 2023-04-23 | 2023-08-01 | 南京理工大学 | ECAPA-TDNN-based automatic identification method for bird song species |
| CN117095694A (en) * | 2023-10-18 | 2023-11-21 | 中国科学技术大学 | Bird song recognition method based on tag hierarchical structure attribute relationship |
Non-Patent Citations (2)
| Title |
|---|
| 基于改进MFCC的鸟鸣声识别方法研究;程龙;张华清;;中国传媒大学学报(自然科学版);20170625(第03期);全文 * |
| 基于语谱图特征信息分割提取的声景观中鸟类生物多样性分析;蒋锦刚;邵小云;万海波;齐家国;荆长伟;程天佑;;生态学报;20161208(第23期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117727332A (en) | 2024-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Priyadarshani et al. | Automated birdsong recognition in complex acoustic environments: a review | |
| CN106653032B (en) | Based on the animal sounds detection method of multiband Energy distribution under low signal-to-noise ratio environment | |
| CN111261189B (en) | Vehicle sound signal feature extraction method | |
| Jiang et al. | Clicks classification of sperm whale and long-finned pilot whale based on continuous wavelet transform and artificial neural network | |
| Ranjard et al. | Unsupervised bird song syllable classification using evolving neural networks | |
| Kershenbaum et al. | An image processing based paradigm for the extraction of tonal sounds in cetacean communications | |
| CN113642484B (en) | Magnetotelluric signal noise suppression method and system based on BP neural network | |
| Ibrahim et al. | Transfer learning for efficient classification of grouper sound | |
| Song et al. | A machine learning-based underwater noise classification method | |
| Wang et al. | Convolutional neural network-based moving ground target classification using raw seismic waveforms as input | |
| Liu et al. | Classification of cetacean whistles based on convolutional neural network | |
| Permana et al. | Implementation of constant-q transform (CQT) and mel spectrogram to converting bird’s sound | |
| Rademan et al. | Soft-output signal detection for cetacean vocalizations using spectral entropy, k-means clustering and the continuous wavelet transform | |
| Xie et al. | Detecting frog calling activity based on acoustic event detection and multi-label learning | |
| Adam | The use of the Hilbert-Huang transform to analyze transient signals emitted by sperm whales | |
| Mercado et al. | Classification of humpback whale vocalizations using a self-organizing neural network | |
| CN117727332B (en) | Ecological population assessment method based on language spectrum feature analysis | |
| Mellinger et al. | A method for filtering bioacoustic transients by spectrogram image convolution | |
| CN113345443A (en) | Marine mammal vocalization detection and identification method based on mel-frequency cepstrum coefficient | |
| Dong et al. | Advances in automatic bird species recognition from environmental audio | |
| CN111091816B (en) | Data processing system and method based on voice evaluation | |
| CN114692693A (en) | Distributed optical fiber signal identification method, device and storage medium based on fractal theory | |
| Liu et al. | Classification method of birdsong based on Gabor_WT feature image and convolutional neural network | |
| Malfante et al. | Machine learning for automatic classification of volcano-seismic signatures | |
| Karam | Various speech processing techniques for speech compression and recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |