CN111414832A

CN111414832A - Real-time online recognition and classification system based on whale dolphin low-frequency underwater acoustic signals

Info

Publication number: CN111414832A
Application number: CN202010180238.2A
Authority: CN
Inventors: 王志陶; 王克雄; 王丁; 段鹏翔
Original assignee: Institute of Hydrobiology of CAS
Current assignee: Institute of Hydrobiology of CAS
Priority date: 2020-03-16
Filing date: 2020-03-16
Publication date: 2020-07-14
Anticipated expiration: 2040-03-16
Also published as: CN111414832B

Abstract

The invention discloses a real-time online identification and classification system based on whale dolphin low-frequency underwater acoustic signals. According to the method, the fundamental frequency and the harmonic signal of the independent three-dimensional underwater acoustic time-frequency domain signal are obtained by subjecting an underwater acoustic signal to pulse interference reduction weight function, signal three-dimensional spectrogram conversion, frequency and time dimension iteration noise reduction, three-dimensional spectrogram threshold filtering, fundamental frequency contour construction, fundamental frequency anti-aliasing processing and harmonic signal extraction; and obtaining the frequency, the quantification and the time parameters of the fundamental frequency of each three-dimensional time-frequency domain signal and the corresponding sound spectrum type and harmonic parameters of the harmonic waves by parameter extraction and combination of a signal classification model. And establishing a whale-fish type comparison data set through three types of parameter acquisition modes, comparing the fundamental frequency and harmonic parameters of the three-dimensional underwater sound time-frequency domain signal after the sound spectrum type is determined with the distribution range of the fundamental frequency and harmonic parameters of the comparison data set in the corresponding sound spectrum type, and identifying the target whale-fish type.

Description

Real-time online recognition and classification system based on whale dolphin low-frequency underwater acoustic signals

Technical Field

The invention relates to the technical field of wild animal protection, in particular to a whale dolphin low-frequency underwater acoustic signal-based real-time online identification and classification system.

Background

In China, the protection level of all whales is not lower than the second level of China. Meanwhile, most whales are distributed in the ecological system of a fresh water river or river mouths and shallow water areas near the shore of the ocean. For example, the national second-level protection animals of China, Changjiang porpoise, only live in the dry stream and Poyang lake in the middle and downstream of Changjiang river and the Dongting lake, and the national first-level protection animals of China, China white dolphin, are only distributed in estuary and near-shore water areas in China. Human activities such as fisheries, shipping and wading are very active in fresh water and offshore waters, and the threat of related activities to whales is not optimistic.

Cetaceans have evolved and completed their own sonar systems in the long-term evolution. The sonar signals of whales can be divided into high-frequency echo positioning signals for detection and communication signals mainly used for communication and emotional expression among individuals. The peak frequency of a high-frequency sonar signal of whales usually exceeds 100kHz, and the directivity of the sonar signal is strong, specifically, the sonar signal recorded on the main shaft is usually dozens of decibels higher than the sound signal recorded on the non-main shaft. Meanwhile, the acoustic attenuation of the high-frequency sonar signals in the water body propagation process is very serious, and the monitoring distance of researchers on the high-frequency sonar signals is usually dozens of meters to hundreds of meters. In contrast, the frequency of the low-frequency communication signal is usually between tens of hertz and tens of kilohertz, and in addition, the directivity of the communication signal is low, and meanwhile, the acoustic attenuation suffered by the communication signal when the communication signal propagates in the water body is low, and the monitoring range of researchers on the whale low-frequency communication signal can usually reach thousands of meters, and even the communication signal of whales can be received by the water body beyond tens of kilometers. In addition, due to the high-frequency characteristics of the high-frequency sonar signals, the performance of the acoustic recording equipment for effectively monitoring the high-frequency sonar signals, such as a data processing module, analog-to-digital conversion performance, sampling rate and memory of a monitoring instrument, is required to be higher, and in the monitoring of the low-frequency communication signals, the requirements on the related performance of the acoustic equipment are not so high. It can be seen that, compared with the high-frequency sonar signals of whales, the low-frequency communication signals of whales are more suitable for monitoring the activity rules of the whales in the habitat of specific water areas.

However, whale monitoring results obtained based on traditional acoustic monitoring means have more or less hysteresis in time, for example, in the field, long-term fixed-point acoustic monitoring is performed, a self-contained acoustic recorder is usually deployed in the field for continuous monitoring, after a period of time, a recording device is recovered and data is imported into a computer, then downloaded acoustic data is analyzed, and the sonar activity rule of whales is obtained, even in short-term mobile acoustic investigation, the situation that signal recording and signal analysis are independent usually exists, that is, signal analysis is usually performed after field signal recording is finished and data is brought back to a laboratory. In addition, the traditional monitoring method of whale communication signals is to manually identify the whale communication signals by manually monitoring sound files or visually checking spectrogram of the sound signals. For large data, especially for such continuous uninterrupted monitoring modes, manual selection of signals becomes impractical. The need to develop an efficient automatic identification algorithm for whale communication signals and realize automatic identification of whale sonar signals becomes urgent.

The protection degree of most whale species is not optimistic, and how to timely and effectively monitor the whale species in real time and timely adjust the range and degree of human activities in corresponding water areas according to returned real-time early warning information to reduce the potential influence of the human activities on the whales becomes very important.

Disclosure of Invention

The invention aims to provide a real-time online identification and classification system based on whale dolphin low-frequency underwater acoustic signals, so as to solve the problems in the background technology.

In order to achieve the purpose, the invention provides the following technical scheme:

the technical scheme of the device is a real-time online identification and classification device based on whale dolphin low-frequency underwater acoustic signals, and is characterized by comprising the following steps: the system comprises a hydrophone, a filter, an amplifier, a signal acquisition card, a microprocessor, a wireless transmission module and a terminal display module;

the hydrophone, the filter, the amplifier, the signal acquisition card, the microprocessor and the wireless transmission module are sequentially connected in series in a wired mode; the wireless transmission module is connected with the terminal display module in a wireless mode;

the hydrophone is used for acquiring underwater acoustic signals; the filter is used for filtering the underwater sound signal to obtain a filtered underwater sound signal; the amplifier is used for amplifying the filtered underwater sound signal to obtain an amplified underwater sound signal; the signal acquisition card is used for carrying out analog-to-digital conversion sampling on the amplified underwater sound signal to obtain an underwater sound digital signal and transmitting the underwater sound digital signal to the microprocessor; the microprocessor obtains a whale fish monitoring result through the real-time online monitoring and early warning method based on the whale fish low-frequency communication signal according to the underwater sound digital signal, and the whale fish monitoring result is wirelessly transmitted to the terminal display module through the wireless transmission module to be displayed;

the terminal display module comprises a mobile client and a laboratory client;

the mobile client can obtain real-time monitoring information of a target whale and realize a real-time early warning function through installing a developed whale real-time online monitoring system APP;

the laboratory client can be used for scientific research personnel of professional institutions to perform deep analysis and related research and protection work on whale communication signals.

The technical scheme of the method is a real-time online identification and classification method based on whale dolphin low-frequency underwater acoustic signals, and the method specifically comprises the following steps:

step 1: denoising the underwater sound digital signal through a pulse interference reduction weight function to obtain a denoised underwater sound digital signal, converting a signal spectrogram to obtain an underwater sound three-dimensional time-frequency domain signal, iteratively denoising through frequency dimension and time dimension to obtain the denoised underwater sound time-frequency domain signal, filtering through spectrogram threshold to obtain a filtered underwater sound time-frequency domain signal, constructing a fundamental frequency profile, performing fundamental frequency anti-aliasing processing and extracting harmonic signals to realize independent spectrum extraction, and obtaining the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal after the anti-aliasing processing and the harmonic signals corresponding to the fundamental frequency;

step 2: extracting the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing treatment through parameters to obtain corresponding frequency parameters, quantitative parameters and time parameters, and harmonic parameters of harmonic signals corresponding to the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing treatment, and further combining a time-frequency domain model to obtain the sound spectrum type of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing treatment through signal classification;

and step 3: establishing a whale-fish type comparison data set through a first parameter acquisition mode, a second parameter acquisition mode and a third parameter acquisition mode, respectively establishing a frequency parameter distribution range of a fundamental frequency, a quantitative parameter distribution range of the fundamental frequency, a time parameter distribution range of the fundamental frequency and a harmonic parameter distribution range of a harmonic signal corresponding to the fundamental frequency according to the whale-fish type comparison data set, and identifying a target whale-fish type by combining the frequency parameter, the quantitative parameter and the time parameter of the fundamental frequency after anti-aliasing processing of a three-dimensional underwater sound time-frequency domain signal, the harmonic parameter and the sound spectrum type of the harmonic signal corresponding to the fundamental frequency after anti-aliasing processing of the three-dimensional underwater sound time-frequency domain signal;

preferably, the underwater sound digital signal in step 1 is:

Sig(j,i)j∈[1,M],i∈[1,N]

N＝fs×t

wherein M is the number of sampling periods, N is the number of sampling points in a sampling period, Sig (j, i) is the sampling value of the ith underwater acoustic digital signal in the jth sampling period, fs is the sampling rate of the signal, and t is the duration of the sampling period;

step 1, the function of reducing the pulse interference weight is as follows:

W(j,i)＝1+[(Sig(j,i)-media(Sig(j.))/(QR(Sig(j.))×θ)]⁶(i∈[1,N])

media(Sig(j.)＝P[Sig(j.),50]

QR(Sig(j.)＝(P[Sig(j.),75]-P[Sig(j.),25])/2

Sigw(j,i)＝Sig(j,i)/W(j,i)

wherein, W (j, i) is the weight value of the ith sampling value of the underwater acoustic digital signal in the jth sampling period, Sig (j.) is all signals in the jth sampling period, media (Sig (j.)) is the median of Sig (j.), and P [ Sig (j.),50] is the sum of the weight values of the ith sampling value of the underwater acoustic digital signal in the jth sampling period

Ordering the signals in Sig (j.) according to the sequence from small to large, ordering the signal values on 50% of sites, if the length N of the signals is a base number, then ordering the values of the N/2+1 th points, if the length N of the signals is an even number, then ordering the average values of the N/2 th and N/2+1 th points, QR (Sig (j.)) is the quartile interval of Sig (j.), theta is a threshold value of a set point, N is the number of sampling points in a sampling period, P [ Sig (j.),25] is the signal value ordering all the signals in Sig (j.) according to the sequence from small to large on 25% of sites, if the length N of the signals is a base number, then ordering the value of the 0.25N +1 th point, if the length N of the signals is an even number, then ordering the average values of the 0.25N and 0.25N +1 th points, p [ Sig (j.),75] is the signal values of all signals in Sig (j.) sorted from small to large on 75% of sites, if the length N of the signals is a base number, the signal values are the values of 0.75N +1 points after sorting, if the length N of the signals is an even number, the signal values are the average values of 0.75N and 0.75N +1 points after sorting, Sigw (j, i) is the result of weighting the ith underwater acoustic digital signal sampling value in the jth sampling period, and P represents the sorted percentage position of the signals;

step 1, obtaining an underwater acoustic time-frequency domain signal through signal three-dimensional spectrogram conversion:

step 1, after noise reduction, obtaining an underwater sound frequency domain signal by the underwater sound digital signal through Fourier window function transformation:

S0(j,p,q)＝S0[j(p),q,Fsigw[j(p),q]]

Δf＝Δq＝fs/nfft

Δt＝Δj(p)＝nfft×K×(1-r)/fs

wherein j (p) is the p Fourier window transform in the j sampling period and is also the time dimension parameter of the three-dimensional time-frequency domain signal, q is the frequency dimension parameter of the three-dimensional time-frequency domain signal, Fsigw [ j (p), q ] is Sigw (j (p), i (S)) the result after Fourier transform, i is the result of Fourier transform with time and frequency respectively at j (p) and q, i 'represents the i' point in the p Fourier window transform in the j sampling period, K represents the total number of points of the signal subjected to Fourier window transform each time, r is the percentage of the overlapping part of the signal when every two adjacent Fourier window functions are subjected to sliding transform, and S0 is the three-dimensional underwater acoustic time-frequency domain signal obtained by Fourier window function transform, which comprises the time dimension j (p), the frequency dimension q and the result of Fourier transform in the time dimension and the frequency dimension Fsigw [ j (p), q ], fs is the sampling frequency of the signal, nfft is the length of the fourier window function, Δ f is the frequency resolution of the time-frequency domain signal, Δ t is the time resolution of the time-frequency domain signal, N is the number of sampling points in each sampling period, Δ q represents the rate of change of the parameter q, and Δ j (p) represents the rate of change of the parameter j (p);

step 1, the frequency dimension iterative noise reduction model is as follows:

S1(j，p，q)＝S1[j(p)，q，log P sigw l[j(p)，q]]

wherein logPsigw [ j (P), q ] is the energy level of the Fourier transform result (Fsigw [ j (P), q ]) after logarithmic conversion, logPsigw1[ j (P), q ] represents the energy level signal after frequency dimension noise reduction, j (P) is the P-th Fourier window transform in the j-th sampling period, q is the frequency dimension parameter of the three-dimensional time-frequency domain signal, K is the signal length of each Fourier window transform, r is the percentage of the signal overlapping part when every two adjacent Fourier window functions are subjected to sliding conversion, nfft is the length of the Fourier window function, S1(j, P, q) is the three-dimensional underwater sound time-frequency domain signal after frequency dimension iterative noise reduction, and comprises a time dimension signal (j (P), a frequency dimension signal (q), an energy level parameter (j (P) and a frequency dimension noise-reduced energy level signal (log sigw1[ j (P), q, j (P) is the energy level signal after frequency dimension noise reduction, j (P) and q is the energy level signal after frequency dimension noise reduction (log, q ];

step 1, the time dimension iterative noise reduction model is as follows:

S2(j，p，q)＝S2[j(p)，q，log P sigw2[j(p)，q]]

wherein logPsigw2[ j (p), q ] represents an energy distribution signal of the underwater acoustic time-frequency domain signal subjected to time dimension noise reduction, phi represents a noise reduction coefficient of a time dimension noise reduction function, and usually takes a value of 0.05, j (p) is the p-th Fourier window transform in the j-th sampling period, q is a frequency dimension parameter of the three-dimensional time-frequency domain signal, K is a signal length subjected to Fourier window transform every time, r is a percentage of a signal overlapping part in every two adjacent Fourier window function sliding transformations, nfft is a length of the Fourier window function, and S2(j, p, q) is the three-dimensional underwater acoustic time-frequency domain signal subjected to frequency dimension iterative noise reduction, and the method comprises the following steps: the underwater acoustic time-frequency domain signal processing method comprises the steps that time dimension signals (j) (P), frequency dimension signals (q) and energy level parameters (j (P) and q) are respectively the energy distribution signals of the underwater acoustic time-frequency domain signals subjected to time dimension noise reduction treatment (log P sigw1[ j (P), q ];

step 1, obtaining an underwater sound time-frequency domain signal through spectrogram threshold filtering:

S3(j，p，q)＝S3[j(p)，q，log P sigw3[j(p)，q]]

wherein logPsigw3[ j (P), q ] is an energy distribution signal of the underwater sound time-frequency domain signal after threshold filtering, α is a set energy threshold of the time-frequency domain signal, j (P) is the P-th Fourier window transformation in the j-th sampling period, q is a frequency dimension parameter of the three-dimensional time-frequency domain signal, and S3(j, P, q) is the three-dimensional underwater sound time-frequency domain signal obtained after spectrogram threshold filtering, and comprises a time dimension signal (j (P), a frequency dimension signal (q), and energy level parameters (j (P) and q corresponding to the energy distribution signal (log P sigw1[ j (P), q) of the underwater sound time-frequency domain signal after threshold filtering;

step 1 the fundamental frequency profile is:

C[j(k′),j(k′)(m)]＝S3[j(p),q,logPsigw3[j(p),q]](k′∈[1,N′],j(k′)∈[j(p),j(p)+N′(j(k′)))],j(k′)(m)∈[1,N′(j(k′))])

C₃≥a

ΔC₂≤Δq

where j (k ') represents the spectrum of the kth' fundamental frequency in the jth sampling period, j (k ') (m) represents the mth point of the spectrum of the kth' fundamental frequency in the jth sampling period, C [ j (k '), j (k') (m)]Representing the m point of the frequency spectrum of the kth fundamental frequency of the three-dimensional underwater sound time-frequency domain signal in the jth sampling period, N ' representing the number of all fundamental frequency spectra of the three-dimensional underwater sound time-frequency domain signal in the jth sampling period, j (p) representing the p Fourier window transformation in the jth sampling period, N ' (j (k ')) representing the maximum number of sites contained in the frequency spectrum of the kth fundamental frequency in the jth sampling period, C₃Values representing the third dimension of the three-dimensional underwater acoustic time-frequency domain signal, i.e. the energy value, C, of the three-dimensional underwater acoustic time-frequency domain signal₂Values representing the second dimension of the three-dimensional underwater acoustic time-frequency domain signal, i.e. frequency dimension values, ac₂Representing the frequency change rate of the three-dimensional underwater sound time-frequency domain signal, α being the energy threshold of the set time-frequency domain signal;

the anti-aliasing processing of the fundamental frequency in the step 1 is to process the crossing or overlapping of the sound spectrum contour lines, and specifically comprises the following steps:

C₁[j(k″),j(k″)(n′)]＝C[j(k′),j(k′)(m)](n′∈[1,N″(j(k″))])

Tan(y1)≈Tan(y2)

wherein j (k ') represents the frequency spectrum of the k' fundamental frequency in the j sample period, j (k ') (n') represents the n 'point of the frequency spectrum of the k' fundamental frequency in the j sample period after anti-aliasing treatment, and C₁[j(k″),j(k″)(n′)]Representing the nth ' point of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the k ' fundamental frequency in the j sampling period is subjected to anti-aliasing treatment, wherein N ' (j (k ')) represents the total number of the sites contained in the frequency spectrum after the k ' fundamental frequency in the j sampling period is subjected to anti-aliasing treatment, and C_1,1The value of the first dimension of the three-dimensional underwater acoustic time-frequency domain signal representing the fundamental frequency of the anti-aliasing process, i.e. the time-dimension value, C, of the three-dimensional underwater acoustic time-frequency domain signal_1,2A value of a second dimension, i.e. a frequency dimension value, of the three-dimensional underwater acoustic time-frequency domain signal representing the fundamental frequency of the anti-aliasing process, Y represents a branch structure site of the fundamental frequency, and Tan (Y1) and Tan (Y2) represent a slope angle on the left side of the branch structure and a slope angle on the right side of the branch structure respectively;

the harmonic signal extraction in the step 1 is to extract signals at integral multiple frequency sites of a fundamental frequency, and is specifically defined as:

HC[j(k″′)，j(k″′)(o)，l]＝[S3[C_1，1[j(k″)，j(k")(n′)]，C_1，2[j(k″)，j(k″)(n′]×l，log P sigw3[C_1，1[j(k″)，j(k″)(n′)]，C_1，2[j(k″)，j(k″)(n′)]]×l]，l](l∈[2，L]，o∈[1，O(j(k″′))])

wherein, C_1,1[j(k″),j(k″)(n′)]Representing the information on the first dimension of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the anti-aliasing processing on the k' fundamental frequency in the j sampling period, namely time dimension information C_1,2[j(k″),j(k″)(n′)]Representing information in a second dimension of the spectrum of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing processing on a k 'fundamental frequency in a j sampling period, namely frequency dimension information, L representing the maximum harmonic number, and O representing the k' fundamental frequency in the j sampling periodThe maximum point number of the first harmonic of the frequency spectrum after anti-aliasing treatment;

HC [ j (k '), j (k ') (o), l ] represents the signal of the o-th site of the l-order harmonic signal corresponding to the frequency spectrum of the k ' th fundamental frequency of the three-dimensional underwater sound time-frequency domain signal in the j-th sampling period after anti-aliasing processing, and HC [ j (k '), j (k ') (o), l ] is a 4-dimensional signal and comprises a time dimension, a frequency dimension, an energy dimension and an order dimension;

preferably, the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing processing in step 2 is: 1, the frequency spectrum of the k' fundamental frequency in the j sampling period after anti-aliasing processing

C₁[j(k″),j(k″)(n′)](n′∈[1,N″(j(k″))])；

Step 2, harmonic signals corresponding to the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing processing are harmonic signals HC [ j (k '), j (k ') (o), l ], l ∈ [2, L ] corresponding to the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal in the jth sampling period in the step 1 after anti-aliasing processing on the kth ';

the specific method for obtaining the corresponding frequency parameter, quantitative parameter, time parameter and harmonic parameter of the harmonic signal corresponding to the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing treatment through parameter extraction in the step 2 comprises the following steps:

step 2, the frequency parameters comprise: start frequency, fundamental frequency of the signal at 0.25 duration point, fundamental frequency of the signal at 0.5 duration point, fundamental frequency of the signal at 0.75 duration point, end frequency, minimum frequency, maximum frequency, frequency variation range, average frequency;

the starting frequency is a second dimension value of the 1 st point of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the k' fundamental frequency in the j sampling period is subjected to anti-aliasing treatment, namely the frequency dimension C_1,2[j(k″),1]；

The fundamental frequency of the signal at the time point of 0.25 duration is C_1,2[j(k″),0.25×N″(j(k″))]The fundamental frequency of the signal at 0.5 time duration is C_1,2[j(k″),0.5×N″(j(k″))]N "(j (k")) represents the total number of loci contained in the spectrum after anti-aliasing processing at the k 'th fundamental frequency in the j' th sampling period;

the fundamental frequency of the signal at the time point of 0.75 duration is C_1,2[j(k″),0.75×N″(j(k″))]；

The end frequency is C_1,2[j(k″),end]；

The minimum frequency is min (C)_1,2[j(k″),u′](u′∈[1,N″(j(k″))]))；

The maximum frequency is max (C)_1,2[j(k″),u′](u′∈[1,N″(j(k″))]))；

The frequency variation range is max (C)_1,2[j(k″),u′])-min(C_1,2[j(k″),u′](u′∈[1,N″(j(k″))]))；

The average frequency is:

[C_1，2[j(k″)，1]+C_1，2[j(k″)，end]+min(C_1，2[j(k″)，u′])+max(C_1，2[j(k″)，u′])]/4(u′∈[1，N″(j(k″))]))

step 2, the quantitative parameters comprise: starting scanning direction and ending scanning direction of the fundamental frequency of the frequency spectrum, the number of inflection points of the fundamental frequency of the spectrogram, the number of fracture points of the fundamental frequency of the spectrogram and the number of step structures of the fundamental frequency of the spectrogram;

the starting sweep direction of the frequency spectrum fundamental frequency is as follows:

C_1,2[j(k″),2]/C_1,2[j(k″),1]in which C is_1,2[j(k″),1]And C_1,2[j(k″),2]Respectively representing second dimension values, namely frequency dimensions, of a 1 st point and a 2 nd point of a frequency spectrum of the k' th fundamental frequency of the three-dimensional underwater sound time-frequency domain signal in a j sampling period after anti-aliasing treatment;

the end sweep direction is as follows:

C_1,2[j(k″),end]/C_1,2[j(k″),end-1]；

the inflection point of the fundamental frequency of the spectrogram is a zero boundary point at which the frequency change rate changes from a positive value to a negative value or from a negative value to a positive value:

is fundamental frequency C of spectrogram₁[j(k″)]The number of inflection points, i', of the fundamental frequency of the spectrogram is the coordinate position of the inflection point.

The step structure of the fundamental frequency of the spectrogram refers to regions with abrupt frequency changes in a continuous spectrogram, and the frequency variation range of the regions is more than 500Hz:

C_1,2[j(k″),i″+1]-C_1,2[j(k″),i″]≥500

wherein the content of the first and second substances,

representing fundamental frequency C of spectrogram₁[j(k″)]I' represents the coordinate position of the cascade structure of the fundamental frequency of the spectrogram;

the fracture point number of the fundamental frequency of the spectrogram refers to the total number of regions with discontinuous frequencies in the spectrogram:

C_1,2[j(k″),i″′]＝0

wherein the content of the first and second substances,

representing fundamental frequency C of spectrogram₁[j(k″)]I' represents the coordinate position of the fracture point of the fundamental frequency of the spectrogram;

step 2, the time parameters comprise: duration, defined as:

wherein dt is the time interval;

step 2, the harmonic parameters comprise: maximum harmonic number, maximum harmonic frequency;

the maximum harmonic number is max (HC)_1,4[j(k″′),j(k″′)(o),l]) Wherein HC_1,4[j(k″′),j(k″′)(o),l]Is HC [ j (k '), j (k') (o), l]A fourth dimension, the order dimension;

the maximum harmonic frequency is max (HC)_1,2[j(k″′),j(k″′)(o),l]) Wherein HC_1,2[j(k″′),j(k″′)(o),l]Is HC [ j (k '), j (k') (o), l]A second dimension, i.e., the frequency dimension; (ii) a

Step 2, further combining the time-frequency domain model to obtain the sound spectrum type of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing treatment through signal classification, specifically comprising the following steps:

the time-frequency domain model is divided into: smooth, sweep down, sweep up, U-shaped, convex, chordal;

the smooth type judging method comprises the following steps:

in the duration of the whole communication signal, the frequency variation amplitude is less than 1kHz within more than 90% of the duration span;

FC₁[j(k″₁),j(k″₁)(n′₁)]＝C₁[j(k″),j(k″)(n′)](n′₁∈[1,N″₁(j(k″₁))])

wherein FC₁[j(k″₁),j(k″₁)(n′₁)]Representing the kth ″' of smooth three-dimensional underwater sound time-frequency domain signal in the jth sampling period₁N 'of the anti-aliased spectrum of the fundamental frequency'₁Dot, N₁(j(k″₁) Representing a smooth three-dimensional underwater acoustic time-frequency domain signal in the first placeKth of j sampling periods ″₁Total number of sites, FC, of spectrum after antialiasing of fundamental frequencies_1,2[j(k″₁),j(k″₁)(n′₁)]Represents the k < th > of the smooth three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₁N 'of the anti-aliased spectrum of the fundamental frequency'₁A second dimension value, i.e. frequency value, of a point;

the judgment method of the downward scanning type comprises the following steps:

the frequency variation trend of the communication signal is mainly reduced, and even if a frequency rising part exists, the frequency variation range is smaller than 1 kHz;

DC₁[j(k″₂),j(k″₂)(n′₂)]＝C₁[j(k″),j(k″)(n′)](n′₂∈[1,N″₂(j(k″₂))])

DC_1,2[j(k″₂),j(k″₁)(n″′₂)]-DC_1,2[j(k″₂),j(k″₂)(n″′₂-1)]＞0(n″′₂∈[2,N″′₂])

max(DC_1,2[j(k″₂),j(k″₂)(n″′₂)])-min(DC_1,2[j(k″₂),j(k″₂)(n″′₂)])≤1000(n″′₂∈[1,N″′₂])

wherein DC₁[j(k″₂),j(k″₂)(n′₂)]Represents the k & ltth & gt & lt & gt of the down-scan three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₂N 'of the anti-aliased spectrum of the fundamental frequency'₂Dot, N₂(j(k″₂) Represents the kth ″' of the down-scan three-dimensional underwater sound time-frequency domain signal in the jth sampling period₂Total number of sites, DC, of spectrum after anti-aliasing processing of fundamental frequencies_1,2[j(k″₂),j(k″₂)(n″′₂)]Represents the k < th > of the down-scanning type three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₂N 'of the anti-aliased spectrum of the fundamental frequency'₂Second dimension value of a point, i.e. frequency value, N'₂Represents the kth ″' in the jth sampling period₁A sweep-down signal DC_1,2[j(k″₁)]Total number of points of increasing medium frequency.

The method for judging the upper scanning type comprises the following steps:

the frequency variation trend of the communication signal is mainly rising, and even if the frequency is reduced, the frequency variation range is less than 1 kHz;

UC₁[j(k″₃),j(k″₃)(n′₃)]＝C₁[j(k″),j(k″)(n′)](n′₃∈[1,N″₃(j(k″₃))])

UC_1,2[j(k″₃),j(k″₃)(n″′₃)]-UC_1,2[j(k″₃),j(k″₃)(n″′₃-1)]＜0(n″′₃∈[2,N″′₃])

max(UC_1,2[j(k″₃),j(k″₃)(n″′₃)])-min(UC_1,2[j(k″₃),j(k″₃)(n″′₃)])≤1000(n″′₃∈[1,N″′₃])

wherein UC₁[j(k″₃),j(k″₃)(n′₃)]Represents the k & ltth & gt & lt & gt of the up-scanning three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₃N 'of the anti-aliased spectrum of the fundamental frequency'₃Dot, N₃(j(k″₃) Represents the k & ltth & gt & lt & gt of the up-scanning three-dimensional underwater sound time-frequency domain signal in the jth sampling period₃Total number of sites of spectrum with fundamental frequency subjected to anti-aliasing processing, UC_1,2[j(k″₃),j(k″₃)(n″′₃)]Represents the k < th > of the up-scanning type three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₃N 'of the anti-aliased spectrum of the fundamental frequency'₃Second dimension value of a point, i.e. frequency value, N'₃Represents the kth ″' in the jth sampling period₃Individual upper scanning type signal UC₁[j(k″₃)]Total number of points of increasing medium frequency.

The U-shaped judgment method comprises the following steps:

the frequency variation of the communication signal is mainly descending at the beginning and then mainly ascending, and the frequency span of each ascending branch or each descending branch exceeds 1kHz and at least one inflection point;

ConcC₁[j(k″₄),j(k″₄)(n′₄)]＝C₁[j(k″),j(k″)(n′)](n′₄∈[1,N″₄(j(k″₄))])

wherein ConcC₁[j(k″₄),j(k″₄)(n′₄)]Represents the k & ltth & gt & lt & gt of the U-shaped three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₄N 'of the anti-aliased spectrum of the fundamental frequency'₄Dot, N₄(j(k″₄) Represents the k & ltth & gt & lt & gt of the U-shaped three-dimensional underwater sound time-frequency domain signal in the jth sampling period₄The total number of sites of the spectrum after anti-aliasing processing of the fundamental frequency, ConcC_1,2[j(k″₄),j(k″₄)(n″′₄)]Represents the k < th > of the U-shaped three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₄N 'of the anti-aliased spectrum of the fundamental frequency'₄Second dimension value of a point, i.e. frequency value, N ″₄(j(k″₄) ) represents ConcC₁[j(k″₄)]Total number of points of increasing medium frequency.

Represents ConcC₁[j(k″₄)]The location of the inflection point of (a),

represents the j (k') th sample period in the j sample period₄) A U-shaped signal ConcC₁[j(k″₄)]Total number of inflection points of;

the convex judging method comprises the following steps:

the frequency variation condition of the whistle of the communication signal is that the whistle begins to mainly rise and then mainly fall, and the frequency span of each rising branch or falling branch exceeds 1kHz and at least one inflection point;

ConvC₁[j(k″₅),j(k″₅)(n′₅)]＝C₁[j(k″),j(k″)(n′)](n′₅∈[1,N″₅(j(k″₅))])

wherein ConvC₁[j(k″₅),j(k″₅)(n′₅)]Representing the k & ltth & gt & lt & gt of the convex three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₅N 'of the anti-aliased spectrum of the fundamental frequency'₅Dot, N₅(j(k″₅) Represents the kth' of the convex three-dimensional underwater sound time-frequency domain signal in the jth sampling period₅Total number of sites of spectrum after anti-aliasing processing of fundamental frequency, ConvC_1,2[j(k″₅),j(k″₅)(n″′₅)]Represents the k < th > of the convex three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₅N 'of the anti-aliased spectrum of the fundamental frequency'₅Second dimension value of a point, i.e. frequency value, N ″₅(j(k″₅) ) represents ConvC₁[j(k″₅)]Total number of points of increasing medium frequency.

Represents ConvC₁[j(k″₅)]The location of the inflection point of (a),

represents the kth in the jth sampling period ″)₅A convex signal ConvC₁[j(k″₅)]Total number of inflection points.

The method for judging the string shape comprises the following steps:

the frequency of the communication signal changes in a trend that the communication signal rises first and then falls or falls first and then rises and returns circularly, and at least two inflection points exist.

SC₁[j(k″₆),j(k″₆)(n′₆)]＝C₁[j(k″),j(k″)(n′)](n′₆∈[1,N″₆(j(k″₆))])

Wherein SC₁[j(k″₆),j(k″₆)(n′₆)]Representing the kth' of the chord type three-dimensional underwater sound time-frequency domain signal in the jth sampling period₆N 'of the anti-aliased spectrum of the fundamental frequency'₆Dot, N₆(j(k″₆) Represents the kth' of the string-shaped three-dimensional underwater sound time-frequency domain signal in the jth sampling period₆The total number of sites of the spectrum after anti-aliasing processing of the fundamental frequencies,

represents the kth in the jth sampling period ″)₆String type signal

Total number of inflection points of;

sequentially judging the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing processing by the smooth judging method, the downward scanning judging method, the upward scanning judging method, the U-shaped judging method, the convex judging method and the chordal judging method to obtain a time-frequency domain model corresponding to the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing processing as a sound spectrum type;

preferably, in step 3, the first parameter obtaining manner is: obtaining a first whale-dolphin-type alignment data set from published literature data through literature search, which is defined as:

RSig_1,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

wherein, RSig_1,x,y,zIs the z-th fundamental frequency parameter of the y-th sound spectrum type under the x-th whale dolphin type in the first parameter acquisition mode, K is the number of whale dolphin types, y ∈ [1,6]]The fundamental frequency sound spectrum type is represented as a smooth type, a downward-sweeping type, an upward-sweeping type, a U-shaped type, a convex type and a chord type in sequence, z ∈ [1, 4]]Representing that the type of the fundamental frequency parameter is sequentially a frequency parameter of the fundamental frequency, a quantitative parameter of the fundamental frequency, a time parameter of the fundamental frequency and a harmonic parameter of a harmonic signal corresponding to the fundamental frequency;

RSig_1,x,y,1＝(RSig_1,x,y,1,1,RSig_1,x,y,1,2,...,RSig_1,x,y,1,9)

wherein (RSig)_1,x,y,1,1,RSig_1,x,y,1,2,...,RSig_1,x,y,1,9) Sequentially representing the starting frequency, the fundamental frequency of a signal at a duration point of 0.25, the fundamental frequency of a signal at a duration point of 0.5, the fundamental frequency of a signal at a duration point of 0.75, the ending frequency, the minimum frequency, the maximum frequency, the frequency variation range and the average frequency in the frequency parameters of the 1 st fundamental frequency of the y-th sound spectrum type under the x-th whale dolphin type in the first parameter acquisition mode;

RSig_1,x,y,2＝(RSig_1,x,y,2,1,RSig_1,x,y,2,2,...,RSig_1,x,y,2,5)

wherein (RSig)_1,x,y,2,1,RSig_1,x,y,2,2,...,RSig_1,x,y,2,5) Sequentially representing the starting direction and the ending direction of the spectrum fundamental frequency and the turning of the spectrogram fundamental frequency in quantitative parameters of the 2 nd fundamental frequency of the y th sound spectrum type under the x th dolphin type in the first parameter acquisition modeThe number of points, the number of fracture points of the fundamental frequency of the spectrogram and the number of cascade structures of the fundamental frequency of the spectrogram;

RSig_1,x,y,3＝RSig_1,x,y,3,1

wherein, RSig_1,x,y,3,1The duration time of the fundamental frequency in the 3 rd fundamental frequency time parameters of the yth sound spectrum type under the xth whale dolphin type in the first parameter acquisition mode is represented;

RSig_1,x,y,4＝(RSig_1,x,y,4,1,RSig_1,x,y,4,2)

wherein (RSig)_1,x,y,4,1,RSig_1,x,y,4,2) Sequentially representing the maximum harmonic number and the maximum harmonic frequency in harmonic parameters of harmonic signals corresponding to the fundamental frequency of the 4 th fundamental frequency of the y sound spectrum type under the x type of whale dolphin in the first parameter acquisition mode;

and 3, the second parameter acquisition mode is as follows: obtaining original audio files of a plurality of sound signals from an openly available whale sound library, extracting fundamental frequencies of the original audio file signals and harmonic signals corresponding to the fundamental frequencies by referring to the step 1 for the plurality of original audio file signals, obtaining frequency parameters, quantitative parameters, time parameters, harmonic parameters of the harmonic signals corresponding to the fundamental frequencies of the original audio file signals and sound spectrum types of the original audio file signals by referring to the step 2, and constructing a second whale dolphin type comparison data set defined as:

RSig_2,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

wherein, RSig_2,x,y,zIs the z-th fundamental frequency parameter of the y-th sound spectrum type under the x-th whale dolphin type in the second parameter acquisition mode, K is the number of whale dolphin types, y ∈ [1,6]]The fundamental frequency sound spectrum type is represented as a smooth type, a downward-sweeping type, an upward-sweeping type, a U-shaped type, a convex type and a chord type in sequence, z ∈ [1, 4]]Representing that the type of the fundamental frequency parameter is sequentially a frequency parameter of the fundamental frequency, a quantitative parameter of the fundamental frequency, a time parameter of the fundamental frequency and a harmonic parameter of a harmonic signal corresponding to the fundamental frequency;

RSig_2,x,y,1＝(RSig_2,x,y,1,1,RSig_2,x,y,1,2,...,RSig_2,x,y,1,9)

wherein (RSig)_2,x,y,1,1,RSig_2,x,y,1,2,...,RSig_2,x,y,1,9) Sequentially representing the starting frequency, the fundamental frequency of the signal at the duration point of 0.25, the fundamental frequency of the signal at the duration point of 0.5, the fundamental frequency of the signal at the duration point of 0.75, the ending frequency, the minimum frequency, the maximum frequency, the frequency variation range and the average frequency in the frequency parameters of the 1 st fundamental frequency of the y-th sound spectrum type under the x-th whale dolphin type in the second parameter acquisition mode;

RSig_2,x,y,2＝(RSig_2,x,y,2,1,RSig_2,x,y,2,2,...,RSig_2,x,y,2,5)

wherein (RSig)_2,x,y,2,1,RSig_2,x,y,2,2,...,RSig_2,x,y,2,5) Sequentially representing the starting sweeping direction and the ending sweeping direction of the spectrum fundamental frequency, the number of turning points of the spectrogram fundamental frequency, the number of fracture points of the spectrogram fundamental frequency and the number of step structures of the spectrogram fundamental frequency in quantitative parameters of the 2 nd fundamental frequency of the y th sound spectrum type under the x th whale dolphin type in the second parameter acquisition mode;

RSig_2,x,y,3＝RSig_2,x,y,3,1

wherein RSig_2,x,y,3,1The duration time of the fundamental frequency in the 3 rd fundamental frequency time parameters of the yth sound spectrum type under the xth whale dolphin type in the first parameter acquisition mode is represented;

RSig_2,x,y,4＝(RSig_2,x,y,4,1,RSig_2,x,y,4,2)

wherein (RSig)_2,x,y,4,1,RSig_2,x,y,4,2) Sequentially representing the maximum harmonic number and the maximum harmonic frequency in the harmonic parameters of the harmonic signals corresponding to the fundamental frequency of the 4 th fundamental frequency of the y sound spectrum type under the x whale dolphin type in the second parameter acquisition mode;

step 3, the third parameter obtaining mode is as follows: obtaining audio signals through a field recording mode, extracting fundamental frequency of the audio signals and harmonic signals corresponding to the fundamental frequency according to the step 1, obtaining frequency parameters, quantitative parameters and time parameters of the fundamental frequency of the audio signals and harmonic parameter audio signal sound spectrum types of the harmonic signals corresponding to the fundamental frequency of the audio signals according to the step 2, and constructing a third whale dolphin type comparison data set, wherein the third whale dolphin type comparison data set is defined as:

RSig_3,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

wherein, RSig_3,x,y,zIs the z-th fundamental frequency parameter of the y-th sound spectrum type under the x-th whale dolphin type in the third parameter acquisition mode, K is the number of whale dolphin types, y ∈ [1,6]]The fundamental frequency sound spectrum type is represented as a smooth type, a downward-sweeping type, an upward-sweeping type, a U-shaped type, a convex type and a chord type in sequence, z ∈ [1, 4]]Representing that the type of the fundamental frequency parameter is sequentially a frequency parameter of the fundamental frequency, a quantitative parameter of the fundamental frequency, a time parameter of the fundamental frequency and a harmonic parameter of a harmonic signal corresponding to the fundamental frequency;

RSig_3,x,y,1＝(RSig_3,x,y,1,1,RSig_3,x,y,1,2,...,RSig_3,x,y,1,9)

wherein (RSig)_3,x,y,1,1,RSig_3,x,y,1,2,...,RSig_3,x,y,1,9) Sequentially representing the starting frequency, the fundamental frequency of the signal at the duration point of 0.25, the fundamental frequency of the signal at the duration point of 0.5, the fundamental frequency of the signal at the duration point of 0.75, the ending frequency, the minimum frequency, the maximum frequency, the frequency variation range and the average frequency in the frequency parameters of the 1 st fundamental frequency of the y-th sound spectrum type under the x-th whale dolphin type in a third parameter acquisition mode;

RSig_3,x,y,2＝(RSig_3,x,y,2,1,RSig_3,x,y,2,2,...,RSig_3,x,y,2,5)

wherein (RSig)_3,x,y,2,1,RSig_3,x,y,2,2,...,RSig_3,x,y,2,5) Sequentially representing the starting sweeping direction and the ending sweeping direction of the spectrum fundamental frequency, the number of turning points of the spectrogram fundamental frequency, the number of fracture points of the spectrogram fundamental frequency and the number of step structures of the spectrogram fundamental frequency in quantitative parameters of the 2 nd fundamental frequency of the y th sound spectrum type under the x th whale dolphin type in a third parameter acquisition mode;

RSig_3,x,y,3＝RSig_3,x,y,3,1

wherein RSig_3,x,y,3,1The duration time of the fundamental frequency in the 3 rd fundamental frequency time parameters of the yth sound spectrum type under the xth whale dolphin type in the first parameter acquisition mode is represented;

RSig_3,x,y,4＝(RSig_3,x,y,4,1,RSig_3,x,y,4,2)

wherein (RSig)_3,x,y,4,1,RSig_3,x,y,4,2) Sequentially representing the maximum harmonic number and the maximum harmonic frequency in harmonic parameters of harmonic signals corresponding to the fundamental frequency of the 4 th fundamental frequency of the y sound spectrum type under the x type whale dolphin type in a third parameter acquisition mode;

step 3, the whale fish type comparison data set is as follows:

RSig_ID,x,y,z

ID∈[1,3],x∈[1,K],y∈[1,6],z∈[1,4]

wherein, RSig_ID,x,y,zIs the z-th fundamental frequency parameter of the y-th sound spectrum type under the x-th whale dolphin type in the ID parameter acquisition mode, K is the number of whale dolphin types, y ∈ [1,6]]The fundamental frequency sound spectrum type is represented as a smooth type, a downward-sweeping type, an upward-sweeping type, a U-shaped type, a convex type and a chord type in sequence, z ∈ [1, 4]]A frequency parameter representing the type of the fundamental frequency parameter as the fundamental frequency, a quantitative parameter of the fundamental frequency,

The time parameter of the fundamental frequency, the harmonic parameter of the harmonic signal corresponding to the fundamental frequency,

step 3, the statistic distribution parameters of the whale fish type comparison data set are as follows:

Srsig_ID,x,y,z,p

ID∈[1,3],x∈[1,K],y∈[1,6],z∈[1,4],p∈[1,4]

wherein, Srsig_ID,x,y,z,pThe result of the p statistical variables of the z fundamental frequency parameters of the y sound spectrum type under the x type of whale in the ID parameter acquisition mode, K is the number of whale types, y ∈ [1,6]]The fundamental frequency sound spectrum type is represented as a smooth type, a downward-sweeping type, an upward-sweeping type, a U-shaped type, a convex type and a chord type in sequence, z ∈ [1, 4]]Frequency parameter representing the type of the fundamental frequency parameter as the fundamental frequency, quantitative parameter of the fundamental frequency, time parameter of the fundamental frequency, harmonic parameter of the harmonic signal corresponding to the fundamental frequency, p ∈ [1, 4]]Representing that the statistical distribution parameters are mean (mean), variance (SD), median (medium) and quartile interval (QD) in sequence;

Srsig_ID,x,y,z,p＝(Srsig_ID,x,y,z,1,Srsig_ID,x,y,z,2,...,Srsig_ID,x,y,z,4)

wherein (Srsig)_ID,x,y,z,1,Srsig_ID,x,y,z,2,...,Srsig_ID,x,y,z,4) Sequentially representing the average value, the variance, the median and the interquartile distance of the z fundamental frequency parameters of the y type of sound spectrum under the x type of whale in the ID parameter acquisition mode;

for any one of the total number of data contained is

Of the data set

In terms of average value thereof

Variance (variance)

Median number

And the interquartile range

Comprises the following steps:

wherein.

To be composed of

After all signals in the sequence from small to big, the signal values on 50% of the sites are sorted,

to be composed of

After all signals in the sequence from small to large, the signal values at 75% of the sites are sorted,

to be composed of

After all signals in the sequence are arranged from small to large, the signal values on 25% of sites are sequenced;

and 3, identifying the type of the target whale dolphin, wherein the specific method comprises the following steps:

if the distribution range judgment model is satisfied, the following steps are carried out:

judging that the corresponding whale fish species type is x, namely the whale fish species type monitoring result of the underwater sound digital signal.

Wherein Dresult_U,P(U∈[1,6],P∈[1,17]) Representing the P type parameter under the U type fundamental frequency sound spectrum type;

u ∈ [1,6] indicates that the sound spectrum types of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing treatment in the step 2 are smooth, downward-sweeping, upward-sweeping, U-shaped, convex and chordal in sequence;

wherein, P ∈ [1,9] represents the frequency parameters of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing processing in step 2, and the frequency parameters sequentially comprise a start frequency, a fundamental frequency of the signal at a duration point of 0.25, a fundamental frequency of the signal at a duration point of 0.5, a fundamental frequency of the signal at a duration point of 0.75, an end frequency, a minimum frequency, a maximum frequency, a frequency variation range and an average frequency;

p ∈ [10,14] represents the quantitative parameters of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing treatment in the step 2, and sequentially comprises the number of turning points of the initial sweeping direction, the ending sweeping direction, the fundamental frequency of the spectrogram, the number of fracture points of the fundamental frequency of the spectrogram and the number of step structures of the fundamental frequency of the spectrogram;

p15 represents the duration of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing processing in the step 2;

p ∈ [16,17] represents the harmonic parameters of the harmonic signals of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signals after anti-aliasing processing in step 2, and the harmonic parameters are the maximum harmonic number and the maximum harmonic frequency in sequence.

Compared with the prior art, the invention has the beneficial effects that:

according to the invention, through developing and integrating the whale low-frequency underwater acoustic communication signal automatic identification and extraction module, high-efficiency automatic identification and extraction of whale low-frequency communication signals are realized, the reliability and stability of monitoring results are improved, and the subjectivity of manual selection is reduced. Species homing identification is carried out on the monitored whale fish sound type through the communication signal parameter extraction and signal classification module and the comparison with the constructed whale fish type comparison data set. The all-weather real-time online monitoring of 24/7 is carried out on the sonar signals of whale dolphin and the identification result of whale dolphin species by combining a wireless transmission module, the acoustic monitoring of whale is changed from a traditional 'recording playback' mode to a 'live broadcast' mode, and more effective basic data are provided for monitoring the activity of whale in a target water area in real time, timely reducing the influence of human activity on animals by regulating and controlling the human activity of corresponding water areas in real time, carrying out more effective whale dolphin protection and making effective whale protection measures.

Drawings

FIG. 1: the invention is a system block diagram.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention are described below clearly and completely, and it is obvious that the described embodiments are some, not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

the terminal display module comprises a mobile client and a laboratory client;

The hydrophone model is as follows; the type of the filter is as follows; the type of the amplifier is as follows; the signal acquisition card is of the type; the type of the microprocessor is as follows; the model of the wireless transmission module is as follows; the model of the terminal display module is as follows;

the specific implementation mode of the invention comprises the following steps:

in the step 1, the underwater sound digital signal is:

Sig(j,i)j∈[1,M],i∈[1,N]

N＝fs×t

where M is the number of sampling periods, which may be defined as 8640(8640 ═ 24 h/day x60 min/h x60 s/min/10 s/sampling period), N is the number of sampling points in a sampling period, 960000 (sampling rate x employs a period duration of 96000x10 s), Sig (j, i) is the ith sampling value of the underwater acoustic digital signal in the jth sampling period, fs is the sampling rate of the signal, e.g., 96kHz, and t is the duration of the sampling period, e.g., 10 s;

step 1, the function of reducing the pulse interference weight is as follows:

W(j,i)＝1+[(Sig(j,i)-media(Sig(j.))/(QR(Sig(j.))×θ)]⁶(i∈[1,N])

media(Sig(j.)＝P[Sig(j.),50]

QR(Sig(j.)＝(P[Sig(j.),75]-P[Sig(j.),25])/2

Sigw(j,i)＝Sig(j,i)/W(j,i)

wherein, W (j, i) is a weight value of the sampling value of the ith underwater acoustic digital signal in the jth sampling period, Sig (j.) is all signals in the jth sampling period, media (Sig (j.)) is a median of Sig (j.), P [ Sig (j.),50] is a signal value which is obtained by arranging all signals in Sig (j.) in a sequence from small to large and is sequenced on 50% of the sites, if the length N of the signal is a base number, the value is the value of N/2+1 points after sequencing, if the length N of the signal is an even number, the average value is the N/2 and N/2+1 points after sequencing, QR (Sig (j.)) is the interquartile distance of Sig (j.), and theta is a threshold of the set point, for example, the signal-to-noise ratio can be 10db, and N is the number of sampling points in the sampling period, p [ Sig (j.),25] is the signal value sequenced on 25% of sites after all signals in Sig (j.) are sequenced from small to large, if the length N of the signals is a base number, the value is the value of 0.25N +1 point after sequencing, if the length N of the signals is an even number, the value is the average value of 0.25N and 0.25N +1 point after sequencing, P [ Sig (j.),75] is the signal value sequenced on 75% of sites after all the signals in Sig (j.) are sequenced from small to large, if the length N of the signals is a base number, the value is the value of 0.75N +1 point after sequencing, if the length N of the signals is an even number, the average value is the value of 0.75N and 0.75N +1 point after sequencing, Sigw (j, i) is the sampled value after the acoustic treatment of the ith water weight digital signal in the jth sampling period, p represents the ordered percentile position of the signal;

the pulse interference reduction weight function is used for reducing interference of pulse signals to a subsequent whale communication signal identification process, for high-energy pulse signals, the value of W (j, i) is very large, the energy value of Sig (j, i) is greatly attenuated, and the sound intensity of the pulse signals is weakened; for signals with weak energy, the value of W (j, i) is close to 1, and the attenuation effect of the weight function on the original signals is negligible;

S0(j,p,q)＝S0[j(p),q,Fsigw[j(p),q]]

Δf＝Δq＝fs/nfft

Δt＝Δj(p)＝nfft×K×(1-r)/fs

wherein j (p) is the p Fourier window transform in the j sampling period and is also the time dimension parameter of the three-dimensional time-frequency domain signal, q is the frequency dimension parameter of the three-dimensional time-frequency domain signal, Fsigw [ j (p), q ] is Sigw (j (p), i (S)) the result after Fourier transform, i is the result of Fourier transform with time and frequency respectively at j (p) and q, i 'represents the i' th point in the p Fourier window transform in the j sampling period, K represents the total number of points of the signal subjected to Fourier window transform each time, for example 1024, r is the percentage of the overlapped part of the signal in each two adjacent sliding transform of the Fourier window function, which can be generally 50%, S0 is the three-dimensional underwater acoustic time-frequency domain signal obtained by Fourier window function transform, and comprises the time dimension j (p), the frequency dimension q and the result of Fourier transform in the time dimension and the frequency dimension Fsigw [ j (p), (p) and f) are the time dimension and the frequency dimension and the result of Fourier transform in the time dimension, q ], fs is the sampling frequency of the signal, nfft is the length of the fourier window function, Δ f is the frequency resolution of the time-frequency domain signal, here 94Hz, Δ t is the time resolution of the time-frequency domain signal, here 5ms, N is the number of sampling points in each sampling period, Δ q represents the rate of change of the parameter q, Δ j (p) represents the rate of change of the parameter j (p);

the frequency dimension iterative noise reduction is to reduce the noise of the energy of the signal converted by the same Fourier window function in different frequency intervals, the frequency dimension noise reduction has the function of further filtering pulse signals, and meanwhile, the preparation is made for better extracting communication signals subsequently;

step 1, the frequency dimension iterative noise reduction model is as follows:

S1(j,p,q)＝S1[j(p),q,log P sigw1[j(p),q]]

wherein logPsigw [ j (P), q ] is the energy level of the Fourier transform result (Fsigw [ j (P), q ]) after logarithmic conversion, logPsigw1[ j (P), q ] represents the energy level signal after frequency dimension noise reduction, j (P) is the P-th Fourier window transform in the j-th sampling period, q is the frequency dimension parameter of the three-dimensional time-frequency domain signal, K is the signal length of each Fourier window transform, r is the percentage of the signal overlapping part when every two adjacent Fourier window functions are subjected to sliding conversion, nfft is the length of the Fourier window function, S1(j, P, q) is the three-dimensional underwater sound time-frequency domain signal after frequency dimension iterative noise reduction, and comprises a time dimension signal (j (P), a frequency dimension signal (q) and an energy level dimension (j (P), and a frequency dimension noise-reduced energy level signal (log sigp 1[ j (P), (j) (P, q) is the energy level signal after frequency dimension noise reduction corresponding to j (log (j (P), (P, g) 1 j (j), (j, q ];

the time dimension sliding noise reduction is to reduce the noise of the energy of the signal in the same frequency interval after the signal is converted by different Fourier window functions, and the noise reduction effect of the time dimension sliding noise reduction on the constant background noise which exists continuously, such as the electronic noise of the recording equipment, is obvious;

step 1, the time dimension iterative noise reduction model is as follows:

S2(j,p,q)＝S2[j(p),q,log P sigw2[j(p),q]]

wherein logPsigw2[ j (p), q ] represents an energy distribution signal of the underwater acoustic time-frequency domain signal subjected to time dimension noise reduction, phi represents a noise reduction coefficient of a time dimension noise reduction function, and usually takes a value of 0.05, j (p) is the p-th Fourier window transform in the j-th sampling period, q is a frequency dimension parameter of the three-dimensional time-frequency domain signal, K is a signal length subjected to Fourier window transform every time, r is a percentage of a signal overlapping part, such as 50%, nfft is a length of the Fourier window function, and S2(j, p, q) is the three-dimensional underwater acoustic time signal subjected to frequency dimension iterative noise reduction, and includes: the underwater acoustic time-frequency domain signal processing method comprises the steps that time dimension signals (j) (P), frequency dimension signals (q) and energy level dimensions (j (P) and q) correspond, and energy distribution signals (log P sigw1[ j (P), q ] of the underwater acoustic time-frequency domain signals subjected to time dimension noise reduction processing are log P sigw 1;

the noise reduction in the time dimension has obvious noise reduction effect on the constant background noise which exists continuously, such as electronic noise of a recording device.

S3(j,p,q)＝S3[j(p),q,log P sigw3[j(p),q]]

wherein logPsigw3[ j (P), q ] is an energy distribution signal of the underwater sound time-frequency domain signal after threshold filtering, α is a set energy threshold of the time-frequency domain signal, for example, the signal-to-noise ratio is 10dB, j (P) is the P-th Fourier window transformation in the j-th sampling period, q is a frequency dimension parameter of the three-dimensional time-frequency domain signal, and S3(j, P, q) is the three-dimensional underwater sound time-frequency domain signal obtained after spectrogram threshold filtering, and comprises a time dimension signal (j (P), a frequency dimension signal (q) and an energy level dimension (log P sigw1[ j (P), q ] which are corresponding to the time dimension signal (j) (P) and q) and are energy distribution signals of the underwater sound time-frequency domain signal after threshold filtering;

step 1 the spectrum extraction comprises: constructing a fundamental frequency outline, performing fundamental frequency anti-aliasing processing and extracting harmonic signals;

temporally adjacent and spatially adjacent time-frequency domain signal points are connected, and simultaneously, spatially acoustic related signals are merged to construct a sound spectrum profile of the signal. The method mainly comprises the three steps of constructing a fundamental frequency contour of a signal, performing fundamental frequency anti-aliasing processing and extracting a harmonic signal.

Step 1 the fundamental frequency profile is:

C₃≥a

ΔC₂≤Δq

where j (k ') represents the spectrum of the kth' fundamental frequency in the jth sampling period, j (k ') (m) represents the mth point of the spectrum of the kth' fundamental frequency in the jth sampling period, C [ j (k '), j (k') (m)]Representing the m point of the frequency spectrum of the kth fundamental frequency of the three-dimensional underwater sound time-frequency domain signal in the jth sampling period, N ' representing the number of all fundamental frequency spectra of the three-dimensional underwater sound time-frequency domain signal in the jth sampling period, j (p) representing the p Fourier window transformation in the jth sampling period, N ' (j (k ')) representing the maximum number of sites contained in the frequency spectrum of the kth fundamental frequency in the jth sampling period, C₃Values representing the third dimension of the three-dimensional underwater acoustic time-frequency domain signal, i.e. the energy value, C, of the three-dimensional underwater acoustic time-frequency domain signal₂Values representing the second dimension of the three-dimensional underwater acoustic time-frequency domain signal, i.e. frequency dimension values, ac₂Representing the frequency change rate of the three-dimensional underwater sound time-frequency domain signal, α is the energy threshold of the set time-frequency domain signal, for example, the signal-to-noise ratio can be 10 dB;

C₁[j(k″),j(k″)(n′)]＝C[j(k′),j(k′)(m)](n′∈[1,N″(j(k″))])

Tan(y1)≈Tan(y2)

HC[j(k″′)，j(k″′)(o)，l]＝[S3[C_1，1[j(k′)，j(k″)(n′)]，C_1，2[j(k″)，j(k″)(n′)]×l，logPsigw3[C_1，1[j(k″)，j(k）″（n′)]，C_1，2[j(k″)，j(k）″（n′)]]×l]，l](l∈[2，L]，o∈[1，O(j(k″′))])

wherein, C_1,1[j(k″),j(k″)(n′)]Representing the information on the first dimension of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the anti-aliasing processing on the k' fundamental frequency in the j sampling period, namely time dimension information C_1,2[j(k″),j(k″)(n′)]Representing information in a second dimension of a frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing processing on a k' fundamental frequency in a j sampling period, namely frequency dimension information, L represents the maximum harmonic numberFor example, 20, O represents the maximum number of points of the l harmonic of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the k' th fundamental frequency in the j-th sampling period is subjected to anti-aliasing processing;

HC [ j (k '), j (k ') (o), l ] represents an l-order harmonic signal corresponding to a frequency spectrum of the k ' th fundamental frequency of the three-dimensional underwater sound time-frequency domain signal in the j sampling period after anti-aliasing processing, and HC [ j (k '), j (k ') (o), l ] is a 4-dimensional signal and comprises a time dimension, a frequency dimension, an energy dimension and an order dimension;

2, the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing treatment is as follows: 1, the frequency spectrum C of the k' th fundamental frequency in the j sampling period after anti-aliasing processing₁[j(k″),j(k″)(n′)](n′∈[1,N″(j(k″))])；

The fundamental frequency of the signal at the time point of 0.25 duration is C_1,2[j(k″),0.25×N″(j(k″))]And N "(j (k")) represents the total number of sites contained in the spectrum after anti-aliasing processing at the k "th fundamental frequency in the j sample period.

The fundamental frequency of the signal at the 0.5 duration point is C_1,2[j(k″),0.5×N″(j(k″))]；

The end frequency is C_1,2[j(k″),end]；

The average frequency is

[C_1，2[j(k″)，1]+C_1，2[j(k″)，end]+min(C_1，2[j(k″)，u′])+max(C_1，2[j(k″)，u′])]/4(u′∈[j，N″(j(k″))]))

C_1,2[j(k″),2]/C_1,2[j(k″),1]in which C is_1,2[j(k″),2]And C_1,2[j(k″),1]Respectively representing the 1 st frequency spectrum of the k' fundamental frequency of the three-dimensional underwater sound time-frequency domain signal in the j sampling period after anti-aliasing treatmentA second dimension value, i.e. frequency dimension, of the point and the 2 nd point;

the end sweep direction is as follows:

C_1,2[j(k″),end]/C_1,2[j(k″),end-1]；

C_1,2[j(k″),i″+1]-C_1,2[j(k″),i″]≥500

wherein the content of the first and second substances,

C_1,2[j(k″),i″′]＝0

wherein the content of the first and second substances,

step 2, the time parameters comprise: duration, defined as:

wherein dt is the time interval;

the smooth type judging method comprises the following steps:

wherein FC₁[j(k″₁),j(k″₁)(n′₁)]Representing the kth ″' of smooth three-dimensional underwater sound time-frequency domain signal in the jth sampling period₁N 'of the anti-aliased spectrum of the fundamental frequency'₁Dot, N₁(j(k″₁) Represents the kth ″' of the smooth three-dimensional underwater sound time-frequency domain signal in the jth sampling period₁Total number of sites, FC, of spectrum after antialiasing of fundamental frequencies_1,2[j(k″₁),j(k″₁)(n′₁)]Represents the k < th > of the smooth three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₁N 'of the anti-aliased spectrum of the fundamental frequency'₁A second dimension value, i.e. frequency value, of a point;

The method for judging the upper scanning type comprises the following steps:

The U-shaped judgment method comprises the following steps:

Represents ConcC₁[j(k″₄)]The location of the inflection point of (a),

the convex judging method comprises the following steps:

wherein ConvC₁[j(k″₅),j(k″₅)(n′₅)]Representing the k & ltth & gt & lt & gt of the convex three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₅N 'of the anti-aliased spectrum of the fundamental frequency'₅Dot, N₅(j(k″₅) Represents the kth' of the convex three-dimensional underwater sound time-frequency domain signal in the jth sampling period₅Fundamental frequency channel reactanceConvc, the total number of sites of the aliased spectrum_1,2[j(k″₅),j(k″₅)(n″′₅)]Represents the k < th > of the convex three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₅N 'of the anti-aliased spectrum of the fundamental frequency'₅Second dimension value of a point, i.e. frequency value, N ″₅(j(k″₅) ) represents ConvC₁[j(k″₅)]Total number of points of increasing medium frequency.

Represents ConvC₁[j(k″₅)]The location of the inflection point of (a),

The method for judging the string shape comprises the following steps:

represents the kth in the jth sampling period ″)₆Individual chord type signal SC₁[j(k″₆)]Total number of inflection points of;

step 3, the first parameter obtaining mode is as follows: obtaining a first whale-dolphin-type alignment data set from published literature data through literature search, which is defined as:

RSig_1,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

RSig_1,x,y,1＝(RSig_1,x,y,1,1,RSig_1,x,y,1,2,...,RSig_1,x,y,1,9)

RSig_1,x,y,2＝(RSig_1,x,y,2,1,RSig_1,x,y,2,2,...,RSig_1,x,y,2,5)

wherein (RSig)_1,x,y,2,1,RSig_1,x,y,2,2,...,RSig_1,x,y,2,5) Sequentially representing the starting sweeping direction and the ending sweeping direction of the spectrum fundamental frequency, the number of turning points of the spectrogram fundamental frequency, the number of fracture points of the spectrogram fundamental frequency and the number of step structures of the spectrogram fundamental frequency in quantitative parameters of the 2 nd fundamental frequency of the y th sound spectrum type under the x th dolphin type in the first parameter acquisition mode;

the time parameters are as follows: RSig_1,x,y,3＝RSig_1,x,y,3,1

Wherein, RSig_1,x,y,3,1The duration time of the fundamental frequency in the 3 rd fundamental frequency time parameters of the yth sound spectrum type under the xth whale dolphin type in the first parameter acquisition mode is represented; harmonic parameters RSig of harmonic signals corresponding to the fundamental frequency_1,x,y,4＝(RSig_1,x,y,4,1,RSig_1,x,y,4,2)

RSig_2,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

RSig_2,x,y,1＝(RSig_2,x,y,1,1,RSig_2,x,y,1,2,...,RSig_2,x,y,1,9)

RSig_2,x,y,2＝(RSig_2,x,y,2,1,RSig_2,x,y,2,2,...,RSig_2,x,y,2,5)

RSig_2,x,y,3＝RSig_2,x,y,3,1

RSig_2,x,y,4＝(RSig_2,x,y,4,1,RSig_2,x,y,4,2)

RSig_3,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

RSig_3,x,y,1＝(RSig_3,x,y,1,1,RSig_3,x,y,1,2,...,RSig_3,x,y,1,9)

wherein (RSig)_3,x,y,1,1,RSig_3,x,y,1,2,...,RSig_3,x,y,1,9) Sequentially represents the starting frequency of the frequency parameter of the 1 st fundamental frequency of the y sound spectrum type under the x whale dolphin type in the third parameter acquisition mode, the fundamental frequency of the signal at the 0.25 duration point,The fundamental frequency of the signal at the 0.5 duration point, the fundamental frequency of the signal at the 0.75 duration point, the end frequency, the minimum frequency, the maximum frequency, the frequency variation range, the average frequency;

RSig_3,x,y,2＝(RSig_3,x,y,2,1,RSig_3,x,y,2,2,...,RSig_3,x,y,2,5)

RSig_3,x,y,3＝RSig_3,x,y,3,1

RSig_3,x,y,4＝(RSig_3,x,y,4,1,RSig_3,x,y,4,2)

step 3, the whale fish type comparison data set is as follows:

RSig_ID,x,y,z

ID∈[1,3],x∈[1,K],y∈[1,6],z∈[1,4]

wherein, RSig_ID,x,y,zIs the z-th fundamental frequency parameter of the y-th sound spectrum type under the x-th whale dolphin type in the ID parameter acquisition mode, K is the number of whale dolphin types, y ∈ [1,6]]The fundamental frequency sound spectrum type is represented as a smooth type, a downward-sweeping type, an upward-sweeping type, a U-shaped type, a convex type and a chord type in sequence, z ∈ [1, 4]]The parameter type of the fundamental frequency is sequentially the frequency parameter of the fundamental frequency, the quantitative parameter of the fundamental frequency, the time parameter of the fundamental frequency and the harmonic parameter of the harmonic signal corresponding to the fundamental frequency,

Srsig_ID,x,y,z,p

ID∈[1,3],x∈[1,K],y∈[1,6],z∈[1,4],p∈[1,4]

Srsig_ID,x,y,z,p＝(Srsig_ID,x,y,z,1,Srsig_ID,x,y,z,2,...,Srsig_ID,x,y,z,4)

for any one of the total number of data contained is

Of the data set

In terms of average value thereof

Variance (variance)

Median number

And the interquartile range

Comprises the following steps:

wherein.

To be composed of

to be composed of

to be composed of

p ∈ [1,9] represents the frequency parameters of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing processing in the step 2, and sequentially comprises a starting frequency, a fundamental frequency of the signal at a duration point of 0.25, a fundamental frequency of the signal at a duration point of 0.5, a fundamental frequency of the signal at a duration point of 0.75, an ending frequency, a minimum frequency, a maximum frequency, a frequency variation range and an average frequency;

p ∈ [16,17] represents the harmonic parameters of the harmonic signals of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signals subjected to anti-aliasing processing in the step 2, and the harmonic parameters are the maximum harmonic number and the maximum harmonic frequency in sequence;

if the model accords with the distribution range discrimination model, the corresponding whale dolphin type is judged to be x, namely the whale dolphin type monitoring result of the underwater sound digital signal.

Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims

1. A real-time online identification and classification method based on whale dolphin low-frequency underwater acoustic signals is characterized by comprising the following steps:

and step 3: the whale-fish type comparison data set is established through a first parameter acquisition mode, a second parameter acquisition mode and a third parameter acquisition mode, a frequency parameter distribution range of a fundamental frequency, a quantitative parameter distribution range of the fundamental frequency, a time parameter distribution range of the fundamental frequency and a harmonic parameter distribution range of a harmonic signal corresponding to the fundamental frequency are respectively established according to the whale-fish type comparison data set, and the target whale-fish type is identified by combining the frequency parameter, the quantitative parameter and the time parameter of the fundamental frequency after anti-aliasing processing of a three-dimensional underwater sound time-frequency domain signal, the harmonic parameter and the sound spectrum type of the harmonic signal corresponding to the fundamental frequency after anti-aliasing processing of the three-dimensional underwater sound time-frequency domain signal.

2. The method for real-time online identification and classification based on whale dolphin low-frequency underwater acoustic signals as claimed in claim 1, wherein: in the step 1, the underwater sound digital signal is:

Sig(j,i)j∈[1,M],i∈[1,N]

N＝fs×t

step 1, the function of reducing the pulse interference weight is as follows:

W(j,i)＝1+[(Sig(j,i)-media(Sig(j.))/(QR(Sig(j.))×θ)]⁶(i∈[1,N])

media(Sig(j.)＝P[Sig(j.),50]

QR(Sig(j.)＝(P[Sig(j.),75]-P[Sig(j.),25])/2

Sigw(j,i)＝Sig(j,i)/W(j,i)

wherein, W (j, i) is the weight value of the sampling value of the ith underwater acoustic digital signal in the jth sampling period, Sig (j.) is all signals in the jth sampling period, media (Sig (j.)) is the median of Sig (j.), P [ Sig (j.),50] is the signal value which is sequenced on 50% of the sites after all the signals in Sig (j.) are sequenced from small to large, if the length N of the signals is a base number, the value is the value of the sequenced N/2+1 points, if the length N of the signals is an even number, the average value is the sequenced N/2 and N/2+1 points, QR (Sig (j.)) is the interquartile distance of Sig (j.), theta is the threshold value of the set point, N is the number of the sampling points in the sampling period, P [ Sig (j.)) is the interquartile distance of Sig (j.)) according to the sequence from small to large, the signal values sequenced on 25% of sites are the values of 0.25N +1 point after sequencing if the length N of the signal is an even number, the values of the 0.25N and 0.25N +1 points after sequencing are the average values of the 0.25N and 0.25N +1 points after sequencing, P [ Sig (j.),75] is the value of the 0.75N +1 point after all the signals in the Sig (j.) are arranged in the order from small to large, if the length N of the signal is a base number, the value of the 0.75N +1 point after sequencing is the average value of the 0.75N and 0.75N +1 points after sequencing, if the length N of the signal is an even number, Sigw (j, i) is the result after the ith sampling value of the underwater acoustic digital signal in the jth sampling period is subjected to weight processing, and P represents the sequencing percentage position of the signal;

S0(j,p,q)＝S0[j(p),q,Fsigw[j(p),q]]

Δf＝Δq＝fs/nfft

Δt＝Δj(p)＝nfft×K×(1-r)/fs

step 1, the frequency dimension iterative noise reduction model is as follows:

S1(j，p，q)＝S1[j(p)，q，log P sigw1[j(p)，q]]

step 1, the time dimension iterative noise reduction model is as follows:

S2(j，p，q)＝S2[j(p)，q，log P sigw2[j(p)，q]]

S3(j，p，q)＝S3[j(p)，q，log P sigw3[j(p)，q]]

step 1 the fundamental frequency profile is:

C₃≥a

ΔC₂≤Δq

C₁[j(k″)，j(k″)(n′)]＝C[j(k′)，j(k′)(m)](n′∈[1，N″(j(k″))])

Tan(y1)≈Tan(y2)

wherein j (k ') represents the frequency spectrum of the k' fundamental frequency in the j sample period, j (k ') (n') represents the n 'point of the frequency spectrum of the k' fundamental frequency in the j sample period after anti-aliasing treatment, and C₁[j(k″)，j(k″)(n′)]Representing the nth ' point of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the k ' fundamental frequency in the j sampling period is subjected to anti-aliasing processing, wherein N ' (j (k ')) represents the position where the k ' fundamental frequency in the j sampling period is subjected to anti-aliasing processingTotal number of sites contained in the processed spectrum, C_1，1The value of the first dimension of the three-dimensional underwater acoustic time-frequency domain signal representing the fundamental frequency of the anti-aliasing process, i.e. the time-dimension value, C, of the three-dimensional underwater acoustic time-frequency domain signal_1，2A value of a second dimension, i.e. a frequency dimension value, of the three-dimensional underwater acoustic time-frequency domain signal representing the fundamental frequency of the anti-aliasing process, Y represents a branch structure site of the fundamental frequency, and Tan (Y1) and Tan (Y2) represent a slope angle on the left side of the branch structure and a slope angle on the right side of the branch structure respectively;

HC[j(k″′)，j(k″′)(o)，l]＝[S3[C_1，1[j(k″)，j(k″)(n′)]，C_1，2[j(k″)，j(k″)(n′)]×l，logPsigw3[C_1，1[j(k″)，j(k″)(n′)]，C_1，2[j(k″)，j(k″)(n′)]]×l]，l]

(l∈[2，L]，o∈[1，O(j(k″′))])

wherein, C_1，1[j(k″)，j(k″)(n′)]Representing the information on the first dimension of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the anti-aliasing processing on the k' fundamental frequency in the j sampling period, namely time dimension information C_1，2[j(k″)，j(k″)(n′)]Representing information on a second dimension of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the k 'fundamental frequency in the j sampling period is subjected to anti-aliasing processing, namely frequency dimension information, L represents the maximum harmonic number, O represents the maximum point number of the 1 st harmonic of the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after the k' fundamental frequency in the j sampling period is subjected to anti-aliasing processing, HC [ j (k '), j (k') (O), l]Representing the signal of the o position of the l order harmonic signal corresponding to the frequency spectrum of the k ' fundamental frequency of the three-dimensional underwater sound time-frequency domain signal in the j sampling period after anti-aliasing treatment, HC [ j (k '), j (k ') (o), l]Is a 4-dimensional signal including a time dimension, a frequency dimension, an energy dimension and an order dimension.

3. Real-time online identification classification based on whale dolphin low-frequency underwater acoustic signals according to claim 1The method is characterized in that: 2, the frequency spectrum of the three-dimensional underwater sound time-frequency domain signal after anti-aliasing treatment is as follows: 1, the frequency spectrum C of the k' th fundamental frequency in the j sampling period after anti-aliasing processing₁[j(k″),j(k″)(n′)](n′∈[1,N″(j(k″))])；

The end frequency is C_1,2[j(k″),end]；

The average frequency is:

the end sweep direction is as follows:

C_1,2[j(k″),end]/C_1,2[j(k″),end-1]；

is fundamental frequency C of spectrogram₁[j(k″)]The number of inflection points of, i' is soundThe coordinate position of the inflection point of the spectrogram fundamental frequency;

C_1,2[j(k″),i″+1]-C_1,2[j(k″),i″]≥500

wherein the content of the first and second substances,

C_1,2[j(k″),i″′]＝0

wherein the content of the first and second substances,

step 2, the time parameters comprise: duration, defined as:

wherein dt is the time interval;

the smooth type judging method comprises the following steps:

wherein DC₁[j(k″₂),j(k″₂)(n′₂)]Represents the k & ltth & gt & lt & gt of the down-scan three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₂N 'of the anti-aliased spectrum of the fundamental frequency'₂Dot, N₂(j(k″₂) Represents the kth ″' of the down-scan three-dimensional underwater sound time-frequency domain signal in the jth sampling period₂Total number of sites, DC, of spectrum after anti-aliasing processing of fundamental frequencies_1,2[j(k″₂),j(k″₂)(n″′₂)]Represents the k < th > of the down-scanning type three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₂N 'of the anti-aliased spectrum of the fundamental frequency'₂Second dimension value of a point, i.e. frequency value, N'₂Represents the kth ″' in the jth sampling period₁A sweep-down signal DC_1,2[j(k″₁)]The total number of points with increasing medium frequency;

the method for judging the upper scanning type comprises the following steps:

wherein UC₁[j(k″₃),j(k″₃)(n′₃)]Represents the k & ltth & gt & lt & gt of the up-scanning three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₃N 'of the anti-aliased spectrum of the fundamental frequency'₃Dot, N₃(j(k″₃) Represents the k & ltth & gt & lt & gt of the up-scanning three-dimensional underwater sound time-frequency domain signal in the jth sampling period₃Total number of sites of spectrum with fundamental frequency subjected to anti-aliasing processing, UC_1,2[j(k″₃),j(k″₃)(n″′₃)]Represents the k < th > of the up-scanning type three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₃N 'of the anti-aliased spectrum of the fundamental frequency'₃Second dimension value of a point, i.e. frequency value, N'₃Represents the kth ″' in the jth sampling period₃Individual upper scanning type signal UC₁[j(k″₃)]The total number of points with increasing medium frequency;

the U-shaped judgment method comprises the following steps:

wherein ConcC₁[j(k″₄),j(k″₄)(n′₄)]Represents the k & ltth & gt & lt & gt of the U-shaped three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₄N 'of the anti-aliased spectrum of the fundamental frequency'₄Dot, N₄(j(k″₄) Represents the k & ltth & gt & lt & gt of the U-shaped three-dimensional underwater sound time-frequency domain signal in the jth sampling period₄The total number of sites of the spectrum after anti-aliasing processing of the fundamental frequency, ConcC_1,2[j(k″₄),j(k″₄)(n″′₄)]Represents the k < th > of the U-shaped three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₄N 'of the anti-aliased spectrum of the fundamental frequency'₄Second dimension value of a point, i.e. frequency value, N ″₄(j(k″₄) ) represents ConcC₁[j(k″₄)]The total number of points where the medium frequency increases,

represents ConcC₁[j(k″₄)]The location of the inflection point of (a),

the convex judging method comprises the following steps:

wherein ConvC₁[j(k″₅),j(k″₅)(n′₅)]Representing the k & ltth & gt & lt & gt of the convex three-dimensional underwater sound time-frequency domain signal in the j & ltth & gt sampling period₅N 'of the anti-aliased spectrum of the fundamental frequency'₅Dot, N₅(j(k″₅) Represents the kth' of the convex three-dimensional underwater sound time-frequency domain signal in the jth sampling period₅Total number of sites of spectrum after anti-aliasing processing of fundamental frequency, ConvC_1,2[j(k″₅),j(k″₅)(n″′₅)]Represents the k < th > of the convex three-dimensional underwater sound time-frequency domain signal in the j < th > sampling period₅N 'of the anti-aliased spectrum of the fundamental frequency'₅Second dimension value of a point, i.e. frequency value, N ″₅(j(k″₅) ) represents ConvC₁[j(k″₅)]The total number of points where the medium frequency increases,

represents ConvC₁[j(k″₅)]The location of the inflection point of (a),

represents the kth in the jth sampling period ″)₅Convex typeSignal ConvC₁[j(k″₅)]Total number of inflection points of;

the method for judging the string shape comprises the following steps:

the frequency variation trend of the communication signal is that the communication signal rises firstly and then falls or falls firstly and then rises circularly, and at least two inflection points are arranged at the same time;

SC₁[j(k″₆),j(k″₆)(n′₆)]＝C₁[j(k″),j(k″)(n′)](n′₆∈[1,N₆″(j(k″₆))])

and sequentially judging the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing processing by the smooth judging method, the downward scanning judging method, the upward scanning judging method, the U-shaped judging method, the convex judging method and the chordal judging method to obtain a time-frequency domain model corresponding to the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing processing as the sound spectrum type.

4. The method for real-time online identification and classification based on whale dolphin low-frequency underwater acoustic signals as claimed in claim 1, wherein:

RSig_1,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

RSig_1,x,y,1＝(RSig_1,x,y,1,1,RSig_1,x,y,1,2,...,RSig_1,x,y,1,9)

RSig_1,x,y,2＝(RSig_1,x,y,2,1,RSig_1,x,y,2,2,...,RSig_1,x,y,2,5)

RSig_1,x,y,3＝RSig_1,x,y,3,1

wherein, RSig_1,x,y,3,1Represents the y sound under the x whale dolphin type in the first parameter acquisition modeThe duration of a fundamental frequency in the time parameters of the 3 rd fundamental frequency of the spectral type;

RSig_1,x,y,4＝(RSig_1,x,y,4,1,RSig_1,x,y,4,2)

RSig_2,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

RSig_2,x,y,1＝(RSig_2,x,y,1,1,RSig_2,x,y,1,2,...,RSig_2,x,y,1,9)

wherein (RSig)_2,x,y,1,1,RSig_2,x,y,1,2,...,RSig_2,x,y,1,9) Sequentially represents the starting frequency of the frequency parameter of the 1 st fundamental frequency of the y sound spectrum type under the x whale dolphin type, the fundamental frequency of the signal at the duration point of 0.25 and the fundamental frequency of the signal at the duration point of 0.5 in the second parameter acquisition modeRate, fundamental frequency of the signal at 0.75 duration point, end frequency, minimum frequency, maximum frequency, frequency variation range, average frequency;

RSig_2,x,y,2＝(RSig_2,x,y,2,1,RSig_2,x,y,2,2,...,RSig_2,x,y,2,5)

RSig_2,x,y,3＝RSig_2,x,y,3,1

RSig_2,x,y,4＝(RSig_2,x,y,4,1,RSig_2,x,y,4,2)

RSig_3,x,y,z

x∈[1,K],y∈[1,6],z∈[1,4]

wherein, RSig_3,x,y,zIs the z-th fundamental frequency parameter of the y-th sound spectrum type under the x-th whale dolphin type in the third parameter acquisition mode, K is the number of whale dolphin types,y∈[1,6]the fundamental frequency sound spectrum type is represented as a smooth type, a downward-sweeping type, an upward-sweeping type, a U-shaped type, a convex type and a chord type in sequence, z ∈ [1, 4]]Representing that the type of the fundamental frequency parameter is sequentially a frequency parameter of the fundamental frequency, a quantitative parameter of the fundamental frequency, a time parameter of the fundamental frequency and a harmonic parameter of a harmonic signal corresponding to the fundamental frequency;

RSig_3,x,y,1＝(RSig_3,x,y,1,1,RSig_3,x,y,1,2,...,RSig_3,x,y,1,9)

RSig_3,x,y,2＝(RSig_3,x,y,2,1,RSig_3,x,y,2,2,...,RSig_3,x,y,2,5)

RSig_3,x,y,3＝RSig_3,x,y,3,1

RSig_3,x,y,4＝(RSig_3,x,y,4,1,RSig_3,x,y,4,2)

wherein (RSig)_3,x,y,4,1,RSig_3,x,y,4,2) Sequentially represents the maximum harmonic number and the maximum harmonic number in the harmonic parameters of the harmonic signals corresponding to the fundamental frequency of the 4 th fundamental frequency of the y sound spectrum type under the x whale dolphin type in the third parameter acquisition modeA harmonic frequency;

step 3, the whale fish type comparison data set is as follows:

RSig_ID,x,y,z

ID∈[1,3],x∈[1,K],y∈[1,6],z∈[1,4]

Srsig_ID,x,y,z,p

ID∈[1,3],x∈[1,K],y∈[1,6],z∈[1,4],p∈[1,4]

Srsig_ID,x,y,z,p＝(Srsig_ID,x,y,z,1,Srsig_ID,x,y,z,2,...,Srsig_ID,x,y,z,4)

for any one of the total number of data contained is

Of the data set

In terms of average value thereof

Variance (variance)

Median number

And the interquartile range

Comprises the following steps:

wherein.

To be composed of

All signals in the sequence from small to largeAfter alignment, signal values at 50% of the sites were ranked,

to be composed of

to be composed of

judging that the corresponding whale fish species type is x, namely judging the whale fish species type monitoring result of the underwater sound digital signal;

and P ∈ [16,17] represents harmonic parameters of a harmonic signal of the fundamental frequency of the three-dimensional underwater sound time-frequency domain signal subjected to anti-aliasing processing in the step 2, wherein the harmonic parameters are the maximum harmonic number and the maximum harmonic frequency in sequence.

5. A whale dolphin low-frequency underwater sound signal-based real-time online identification and classification device applied to the whale dolphin low-frequency underwater sound signal-based real-time online identification and classification method as claimed in any one of claims 1-4, characterized by comprising:

the system comprises a hydrophone, a filter, an amplifier, a signal acquisition card, a microprocessor, a wireless transmission module and a terminal display module;

the terminal display module comprises a mobile client and a laboratory client;