CN115910097A - Audible signal identification method and system for latent fault of high-voltage circuit breaker - Google Patents
Audible signal identification method and system for latent fault of high-voltage circuit breaker Download PDFInfo
- Publication number
- CN115910097A CN115910097A CN202210100166.5A CN202210100166A CN115910097A CN 115910097 A CN115910097 A CN 115910097A CN 202210100166 A CN202210100166 A CN 202210100166A CN 115910097 A CN115910097 A CN 115910097A
- Authority
- CN
- China
- Prior art keywords
- cepstrum
- spectrum
- gamma
- frequency
- time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000001228 spectrum Methods 0.000 claims abstract description 169
- 238000001914 filtration Methods 0.000 claims abstract description 68
- 238000010606 normalization Methods 0.000 claims abstract description 65
- 239000011159 matrix material Substances 0.000 claims abstract description 56
- 230000005236 sound signal Effects 0.000 claims abstract description 41
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 35
- 230000009471 action Effects 0.000 claims abstract description 29
- 238000000605 extraction Methods 0.000 claims abstract description 25
- 238000004364 calculation method Methods 0.000 claims abstract description 22
- 230000009467 reduction Effects 0.000 claims abstract description 21
- 230000006870 function Effects 0.000 claims description 21
- 238000012546 transfer Methods 0.000 claims description 16
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000009499 grossing Methods 0.000 claims description 8
- 230000001629 suppression Effects 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 5
- 230000007704 transition Effects 0.000 claims 1
- 238000004590 computer program Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 238000007689 inspection Methods 0.000 description 7
- 230000007246 mechanism Effects 0.000 description 6
- 238000007781 pre-processing Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 238000003745 diagnosis Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000007906 compression Methods 0.000 description 3
- 230000006835 compression Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000007599 discharging Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000013021 overheating Methods 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 210000000860 cochlear nerve Anatomy 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000009432 framing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 210000003454 tympanic membrane Anatomy 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a method and a system for recognizing audible signals of latent faults of a high-voltage circuit breaker. Wherein, the method comprises the following steps: the method comprises the following steps: converting a breaker action sound signal collected by a microphone into a two-dimensional time frequency spectrum with two dimensions of a time domain and a frequency domain; performing feature extraction and dimension reduction on the two-dimensional time spectrum through a Mel cepstrum coefficient, a gamma-pass filtering cepstrum coefficient and a power law normalization cepstrum coefficient, determining a Mel cepstrum feature, a gamma-pass filtering cepstrum feature and a power law normalization cepstrum feature, and forming a cepstrum feature matrix according to the Mel cepstrum feature, the gamma-pass filtering cepstrum feature and the power law normalization cepstrum feature; and taking the convolutional neural network as a classifier, and identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the cepstrum feature matrix. The accuracy is improved, and meanwhile, the calculation speed is not obviously reduced, so that the calculation efficiency is optimized.
Description
Technical Field
The invention relates to the technical field of equipment high-voltage electrical equipment abnormity identification, in particular to a method and a system for identifying audible sound signals of latent faults of a high-voltage circuit breaker.
Background
The high-voltage circuit breaker is used as an important power transmission and transformation device, and the operation mechanism of the high-voltage circuit breaker can act correctly or not, so that the safe operation of a system is directly related. According to statistics, the mechanical faults of the operating mechanism of the high-voltage circuit breaker account for 70% -80% of all faults. The technology for evaluating the mechanical state of the high-voltage circuit breaker operating mechanism and diagnosing the fault and the detection equipment have great significance.
Various common faults (such as a failure fault, a malfunction fault, a jamming fault, a fracture fault and the like) in the practical application of the high-voltage circuit breaker are mostly closely related to an operating mechanism of the high-voltage circuit breaker, the faults are often caused by gradual development and accumulation of latent mechanical faults (oil leakage of an oil buffer, fatigue of a spring, abrasion of a transmission pin and the like), early characteristics of the latent faults are not obvious enough, and if the faults are not discovered and processed in time, serious faults are easy to develop to cause more serious economic loss. The current research on latent faults mainly focuses on overheating and discharging, while the research on mechanical latent faults is less, and therefore further research on feature extraction and early identification of the latent faults needs to be carried out.
The mechanical fault recognition of the circuit breaker mainly analyzes and processes vibration or voiceprint signals in action, the vibration and the voiceprint are different expression forms of mechanical waves in different media, in the actual use process, the vibration sensing needs to be punched and deployed on the circuit breaker, the deployment positions are different, the monitoring result is extremely strong in influence, the voiceprint sensor does not need to be in direct contact with a circuit breaker body, the response difference of small deviation of measuring point positions to the voiceprint signals is small, and the method is more convenient and fast to use on site.
Disclosure of Invention
The invention provides an audible sound signal identification method and system for latent faults of a high-voltage circuit breaker, and aims to solve the technical problems that the existing research on latent faults mainly focuses on overheating and discharging, and the mechanical latent faults are less researched.
According to a first aspect of the present invention, there is provided a method for audible acoustic signal identification of a latent fault of a high voltage circuit breaker, comprising:
converting a breaker action sound signal collected by a microphone into a two-dimensional time frequency spectrum with two dimensions of a time domain and a frequency domain;
performing feature extraction and dimension reduction on the two-dimensional time spectrum through a Mel cepstrum coefficient, a gamma-pass filtering cepstrum coefficient and a power law normalization cepstrum coefficient, determining a Mel cepstrum feature, a gamma-pass filtering cepstrum feature and a power law normalization cepstrum feature, and forming a cepstrum feature matrix according to the Mel cepstrum feature, the gamma-pass filtering cepstrum feature and the power law normalization cepstrum feature;
and taking the convolutional neural network as a classifier, and identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the cepstrum feature matrix.
Optionally, converting the breaker action acoustic signal collected by the microphone into a two-dimensional time frequency spectrum having two dimensions of a time domain and a frequency domain includes:
the hamming window is determined according to the following formula:
wherein w (N) is a Hamming window, and N is a length;
performing short-time discrete Fourier transform on the breaker action sound signal, and determining a two-dimensional time-frequency spectrum matrix:
Wherein, X (k) is each frame frequency spectrum of a two-dimensional time-frequency spectrum matrix, k is a frequency point serial number, and X (n) is a circuit breaker action sound signal.
Optionally, the performing feature extraction and dimension reduction on the two-dimensional time spectrum through a mel-frequency cepstrum coefficient to determine a mel-frequency cepstrum feature includes:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
enabling the power spectrum to pass through a Mel scale filter bank with a triangular frequency domain, and determining Mel cepstrum parameters;
obtaining M Weimeier cepstrum by discrete cosine transform after taking natural logarithm of the Meier cepstrum parameter, and carrying out mean value normalization on the M Weimeier cepstrum to determine the Meier cepstrum characteristic;
where M is the number of Mel filters.
Optionally, the determining the characteristics of the gamma-pass filtered cepstrum by performing feature extraction and dimension reduction on the two-dimensional time spectrum through the gamma-pass filtered cepstrum coefficient includes:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
enabling the power spectrum to pass through a Gamma-tone filter bank consisting of M gamma filters with different scale parameters and shape parameters, and determining gamma-pass filtering cepstrum parameters;
and taking natural logarithm of the gamma-pass filtering cepstrum parameters, obtaining M-dimensional gamma-pass filtering cepstrum through discrete cosine transformation, and performing mean normalization on the gamma-pass filtering cepstrum to determine the characteristics of the gamma-pass filtering cepstrum.
Optionally, the determining the power law normalized cepstrum feature by performing feature extraction and dimension reduction on the two-dimensional time spectrum through the power law normalized cepstrum coefficient includes:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
passing the power spectrum through a Gamma filter group consisting of M gamma filters with different scale parameters and shape parameters to obtain a power spectrum after passing through the filter;
according to the power spectrum after passing through the filter, smoothing each frame and the two frames before and after the frame, filtering out low-frequency parts, carrying out asymmetric noise suppression, and calculating the average power in the middle time;
determining the time average and frequency average transfer functions according to the medium-time average power;
normalizing the time-frequency domain of the original short-time energy spectrum according to the time-average and frequency-average transfer functions, and determining the energy spectrum after time-frequency normalization;
and performing average power normalization on the time-frequency normalized energy spectrum according to the average power estimation value, performing discrete cosine transform and mean value normalization, and determining power law normalization cepstrum characteristics.
Optionally, the method for identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the cepstrum feature matrix by using the convolutional neural network as a classifier includes:
taking the cepstrum feature matrix as an input layer, and constructing a mixed cepstrum coefficient-convolution neural network identification model;
and according to the mixed cepstrum coefficient-convolutional neural network identification model, fault type identification is carried out on the audible sound signal of the latent fault of the high-voltage circuit breaker.
According to another aspect of the present invention, there is also provided a high voltage circuit breaker latent fault audible signal identification system comprising:
the action signal conversion module is used for converting the breaker action sound signals collected by the microphone into a two-dimensional time frequency spectrum with two dimensions of a time domain and a frequency domain;
a feature matrix forming module, configured to perform feature extraction and dimension reduction on the two-dimensional time spectrum through a mel cepstrum coefficient, a gamma-pass filtering cepstrum coefficient and a power law normalized cepstrum coefficient, determine a mel cepstrum feature, a gamma-pass filtering cepstrum feature and a power law normalized cepstrum feature, and form a cepstrum feature matrix according to the mel cepstrum feature, the gamma-pass filtering cepstrum feature and the power law normalized cepstrum feature;
and the fault signal identification module is used for identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker by taking the convolutional neural network as a classifier according to the cepstrum feature matrix.
Optionally, the motion signal conversion module includes:
a determine Hamming Window submodule for determining a Hamming Window according to the following formula:
wherein w (N) is a Hamming window and N is a length;
and determining a frequency spectrum matrix submodule for carrying out short-time discrete Fourier transform on the circuit breaker action sound signal and determining a two-dimensional time frequency spectrum matrix:
Wherein, X (k) is each frame frequency spectrum of a two-dimensional time-frequency spectrum matrix, k is a frequency point serial number, and X (n) is a circuit breaker action sound signal.
Optionally, a feature matrix module is configured, including:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time spectrum after modulus taking;
a Merr cepstrum parameter determining submodule used for determining the Merr cepstrum parameters by enabling the power spectrum to pass through a Mel scale filter bank with a triangular frequency domain;
determining a Mel cepstrum feature submodule, which is used for obtaining an M Viumei cepstrum through discrete cosine transform after taking the natural logarithm of the Mel cepstrum parameters, and performing mean normalization on the M Viumei cepstrum to determine the Mel cepstrum feature;
where M is the number of Mel filters.
Optionally, a feature matrix module is configured, including:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time spectrum after modulus taking;
the gamma-pass filtering cepstrum parameter determining submodule is used for enabling the power spectrum to pass through a gamma filter group consisting of M gamma filters with different scale parameters and shape parameters, and determining gamma-pass filtering cepstrum parameters;
and the gamma pass filtering cepstrum characteristic determining sub-module is used for obtaining an M-dimensional gamma pass filtering cepstrum through discrete cosine transformation after natural logarithm is taken on the gamma pass filtering cepstrum parameters, and performing mean normalization on the gamma pass filtering cepstrum to determine the gamma pass filtering cepstrum characteristics.
Optionally, a feature matrix module is configured, including:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time spectrum after modulus taking;
the power spectrum filtering submodule is used for enabling the power spectrum to pass through a Gamma filter bank consisting of M gamma filters with different scale parameters and shape parameters to obtain a power spectrum after the power spectrum passes through the filter;
the calculation middle-time average power submodule is used for smoothing each frame and the front and rear frames according to the power spectrum after passing through the filter, filtering out low-frequency parts, carrying out asymmetric noise suppression and calculating middle-time average power;
the average transfer function determining submodule is used for determining the time average and frequency average transfer functions according to the medium-time average power;
the energy spectrum submodule after time-frequency normalization is determined, and is used for normalizing the time-frequency domain of the original short-time energy spectrum according to the time-average and frequency-average transfer functions and determining the energy spectrum after time-frequency normalization;
and the power law normalization cepstrum feature determining submodule is used for performing average power normalization on the time-frequency normalized energy spectrum according to the average power estimation value, performing discrete cosine transform and mean value normalization, and determining the power law normalization cepstrum feature.
Optionally, the fault signal identifying module comprises:
a neural network identification model submodule is constructed, and is used for constructing a mixed cepstrum coefficient-convolution neural network identification model by taking the cepstrum feature matrix as an input layer;
and the fault type identification submodule is used for identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the mixed cepstrum coefficient-convolutional neural network identification model.
Therefore, starting from the operation and inspection requirements of power grid equipment, the diagnosis method is researched by taking the latent mechanical fault voiceprint of the circuit breaker which is more consistent with the field application scene as an object by combining the professional characteristics of company operation and inspection and the development trend of constructing an intelligent operation and inspection system. The intelligent identification of the abnormal working condition of the high-voltage circuit breaker operating mechanism is realized through the fusion of an artificial intelligence technology and the traditional operation and inspection service. The management and control force of the equipment state and the transportation and inspection management penetrating power are effectively improved, the innovation development and the efficiency improvement of data-driven transportation and inspection business are realized, and the innovation of a transportation and inspection working mode and a production management mode is comprehensively promoted. The accuracy is improved, and meanwhile, the calculation speed is not obviously reduced, so that the calculation efficiency is optimized.
Drawings
A more complete understanding of exemplary embodiments of the present invention may be had by reference to the following drawings in which:
fig. 1 is a schematic flow chart of an audible signal identification method for a latent fault of a high-voltage circuit breaker according to the embodiment;
fig. 2 is a schematic diagram of an audible signal recognition system for a latent fault of a high-voltage circuit breaker according to the embodiment.
Detailed Description
The exemplary embodiments of the present invention will now be described with reference to the accompanying drawings, however, the present invention may be embodied in many different forms and is not limited to the embodiments described herein, which are provided for complete and complete disclosure of the present invention and to fully convey the scope of the present invention to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the accompanying drawings is not intended to be limiting of the invention. In the drawings, the same unit/element is denoted by the same reference numeral.
Unless otherwise defined, terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Further, it will be understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense.
According to a first aspect of the present invention, there is provided a method 100 for audible acoustic signal identification of a latent fault of a high voltage circuit breaker, the method 100 comprising, with reference to fig. 1:
s101, converting a breaker action sound signal acquired by a microphone into a two-dimensional time frequency spectrum with two dimensions of a time domain and a frequency domain;
s102, performing feature extraction and dimension reduction on the two-dimensional time spectrum through a Mel cepstrum coefficient, a gamma-pass filtering cepstrum coefficient and a power law normalization cepstrum coefficient, determining a Mel cepstrum feature, a gamma-pass filtering cepstrum feature and a power law normalization cepstrum feature, and forming a cepstrum feature matrix according to the Mel cepstrum feature, the gamma-pass filtering cepstrum feature and the power law normalization cepstrum feature;
and S103, taking the convolutional neural network as a classifier, and identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the cepstrum feature matrix.
Specifically, the method comprises the steps of building a circuit breaker test platform, collecting voiceprint signals of the circuit breaker in different latent fault operation states, and distinguishing the switching-on and switching-off sound signals of the high-voltage circuit breaker in different fault states by using a mixed cepstrum calculation method and a CNN convolutional neural network structure. Firstly, forming an acoustic signal sample library in a field acquisition mode; then, respectively adopting Mel Frequency Cepstral Coefficient (MFCC), gamma-pass Filter Cepstral Coefficient (GFCC) and Power-law Normalized Cepstral Coefficient (PNCC) to perform dimensionality reduction and primary feature extraction on the original signal; and finally, introducing a Convolutional Neural Network (CNN) as a classifier, forming a classification model of six operating states by using a Mixed-Cepstral coeffient (Mixed-Cepstral coeffient) formed by the three characteristics of the MFCC, the GFCC and the PNCC, and verifying the effectiveness of the model through a data set.
The circuit breaker acoustic signal identification method includes the steps of firstly converting circuit breaker action acoustic signals collected by a microphone into a time-frequency spectrogram with two dimensions of a time domain and a frequency domain, then performing feature extraction and dimension reduction on an original time spectrum by calculating 3 cepstrums such as MFCCs, GFCCs and PNCCs, and finally using a convolutional neural network as a classifier to perform fault type identification. The whole process can be roughly divided into three parts of sound signal acquisition, preprocessing and pattern recognition, wherein the preprocessing method of the sound signals is most important, and the preprocessing method is mainly used for carrying out feature extraction and data compression on the original time domain signals of the circuit breaker, so that the calculation amount of a subsequent recognition model is reduced, and the recognition effect is improved.
1 Circuit breaker action Acoustic Signal time Spectrum calculation
The time domain signal of the acoustic signal of the circuit breaker is generally a one-dimensional pulse signal, the characteristic information of the signal is not obvious enough, and the signal can be converted into a two-dimensional time frequency spectrum by using a short-time discrete Fourier transform mode, so that the identification speed and the identification accuracy of a deep learning model can be improved. In the short-time Fourier transform process, framing, windowing and discrete Fourier transform are required. Wherein, the window function can generally select a hamming window to reduce the spectrum leakage caused by fourier transform, and the hamming window w (N) with length N is as follows:
and (3) carrying out short-time discrete Fourier transform on the discrete time domain frame to obtain a time-frequency spectrum matrix:
wherein k is a frequency point serial number, and x (n) is an original discrete time domain signal.
2 quasi-steady state process analysis
The circuit breaker latent fault general characteristics are not obvious, so that when the circuit breaker voiceprint diagnosis is carried out, the voiceprint characteristics of the circuit breaker latent fault are extracted on the premise of ensuring the identification speed of an acoustic signal, and the identification accuracy is improved. The cepstrum coefficient calculation method widely used in the field of voice recognition can compress data of a sample and simultaneously reserve closing key voiceprint information, so that compression and feature extraction of a breaker acoustic signal are achieved, and the diagnosis speed and the diagnosis accuracy of classifiers such as a subsequently-connected convolutional neural network and the like are improved.
In various cepstrum coefficients, the construction basis of the MFCC is an auditory model, the construction basis of the GFCC is an eardrum model, the PNCC has more advantages in the aspect of sound feature extraction under a noise background, and the cepstrum is applied to a certain degree in the field of voice recognition, so that the MFCC, the GFCC and the PNCC are selected as basic cepstrum features to form a cepstrum feature matrix for subsequent feature fusion and sound signal recognition.
(1) MFCC calculation
MFCC is a kind of cepstrum parameter based on human auditory perception characteristic, and in frequency domain, the sound level heard by human ear is not linear with frequency, and in Mel domain, human auditory perception is proportional to Mel frequency. The relationship can be expressed by the following formula:
Mel(f)＝2595lg(1+f/700) (3)
the calculation of the mel-frequency cepstrum coefficient is carried out by taking a frame as a unit, and the specific calculation steps of the mel-frequency cepstrum coefficient are as follows:
firstly, calculating according to the formula (1) to obtain a frequency spectrum X (k) of each frame, and obtaining a power spectrum by squaring after taking a module for each frame X (k). Passing the power spectrum through a Mel scale filter bank with triangular frequency domain to obtain a new parameter R (k), wherein the lower limit of the filter bank frequency is f min Upper limit of f max And M is the number of Mel filters.
Then taking the natural logarithm of R (k):
the M-dimensional MFCC is then obtained by Discrete Cosine Transform (DCT):
finally mean normalization is required. The MFCC with M dimension can be obtained through the steps.
(2) GFCC calculation
The extraction process of GFCC is almost the same as the extraction process of MFCC, and the difference is that the filter bank through which the power spectrum passesIs a gamma filter bank consisting of M gamma filters of different scale parameters and shape parameters, rather than a Mel scale filter. The upper and lower limits of the filter bank frequency are f max And f min The subsequent calculation steps are also the same.
(3) PNCC computation
The first two steps of the PNCC extraction process are the same as GFCC, and P [ m, l ] is obtained after the power spectrum passes through a Gamma filter, wherein l represents the channel number. Smoothing each frame and the two frames before and after, and calculating the average power:
the obtained medium-time average power is used for the estimation and compensation of the background noise of the following environment, and the low-frequency part is filtered by using spectral subtraction to achieve the purpose of suppressing the noise, namely, the asymmetric noise suppression is carried out to obtain
Then smoothing is carried out on different channels:
l 1 ＝max(l-p,1) (8)
l 2 ＝min(l+p,M) (9)
where M represents the number of channels and p is typically set to 4.
By using
For P [ m, l]Time-frequency domain normalization:
the mean power estimate μm can be used to normalize T m, l by mean power normalization:
where k is a coefficient that can be set to any constant.
In order to more closely approach the compressive sensing characteristics of the auditory nerve of the human ear, unlike the logarithmic nonlinearity adopted by MFCC, PNCC adopts power-law nonlinear compression:
V[m,l]＝U[m,l] 1/15 (13)
and finally, carrying out discrete cosine transform and mean value normalization to obtain the PNCC.
After the calculation of MFCC, GFCC and PNCC is respectively finished, the three are combined into a [ Z multiplied by M multiplied by 3] cepstrum characteristic matrix, wherein Z is a time domain component and depends on the time frame number of an original time frequency spectrum; m is a frequency domain component, and is equal to the number of filters used in the calculation of each cepstral coefficient, and the larger the number of filters, the more information is, but the larger the data amount is, the generally set interval is 40 to 48.
5.3 convolutional neural network computation
Because the cepstrum feature matrix is a three-dimensional map formed by overlapping a plurality of two-dimensional maps with the same size, a Convolutional Neural Network (CNN) which is representative in the field of image recognition can be introduced and can be used as a classifier of the cepstrum feature matrix of the acoustic signal. In the field of image recognition, CNN generally splits a color image into three color layers of red, green and blue (RGB) as input layers of a network, so as to learn and perceive characteristics of different color changes. Similarly, the study takes a cepstrum feature matrix formed by three cepstrums of the acoustic signal as an input layer, and a Mixed cepstrum Coefficient-Convolutional Neural Network (MCC-CNN) recognition model is constructed, so that the voice classification recognition is carried out. Compared with a manually designed cepstrum mixing method, the method has the advantage that the three cepstrums are fused through a learning mechanism of a deep neural network, so that the fusion mode of the mixed cepstrum has self-adaptability.
Because the dimension reduction compression is carried out on the circuit breaker closing voiceprint data, the circuit breaker closing voiceprint data can be realized through a VGG-like lightweight CNN network, and the whole network comprises 3 convolution-pooling layers and 4 full connection layers. In the network, dropout and batch normalization are added to prevent overfitting and gradient disappearance, and the detailed structure is shown in table 1.
Table 1 type VGG lightweight CNN network structure
Tab，1 VGG-like lightweight CNN Structure
In order to verify the effectiveness of the recognition model of the method, the recognition success rates and the operation times of the MCC-CNN, the MFCC-CNN, the GFCC-CNN, the PNCC-CNN and the conventional CNN model are compared, and the result is shown in Table 2, so that the MCC-CNN model is best in recognition success rate, and the superiority of the MCC-CNN acoustic recognition model provided by the invention is proved.
Compared with the method for training and identifying by directly inputting the time-frequency spectrogram into the CNN network, the MCC-CNN is subjected to preprocessing of mixed cepstrum calculation, the calculated amount of data is greatly reduced, and the reduction of the data of the sample also means that the identification difficulty of the deep neural network is reduced, so that the identification rate can be improved and the identification time can be reduced. Compared with a method using a single cepstrum coefficient preprocessing method, the MCC-CNN can adapt to various latent mechanical fault sound signals by using more types of cepstrum characteristics, and from the result, the trained method for calculating the MCC-CNN has no calculation speed obviously inferior to a method for calculating a single cepstrum coefficient with smaller data volume, improves the accuracy, and simultaneously has no obvious reduction of the calculation speed, because the Dropout operation eliminates a plurality of redundant neuron connections, the calculation efficiency is optimized.
TABLE 2 comparison of the results of different pre-treatment methods
Tab.5 Comparison of different preprocessing methods
Optionally, converting the breaker action acoustic signal collected by the microphone into a two-dimensional time-frequency spectrum having two dimensions of a time domain and a frequency domain includes:
the hamming window is determined according to the following formula:
wherein w (N) is a Hamming window and N is a length;
performing short-time discrete Fourier transform on the breaker action sound signal, and determining a two-dimensional time-frequency spectrum matrix:
Wherein, X (k) is each frame frequency spectrum of a two-dimensional time-frequency spectrum matrix, k is a frequency point serial number, and X (n) is a circuit breaker action sound signal.
Optionally, the performing feature extraction and dimension reduction on the two-dimensional time spectrum through a mel-frequency cepstrum coefficient to determine a mel-frequency cepstrum feature includes:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
enabling the power spectrum to pass through a Mel scale filter bank with a triangular frequency domain, and determining Mel cepstrum parameters;
obtaining M Weimeier cepstrum by discrete cosine transform after taking natural logarithm of the Meier cepstrum parameter, and carrying out mean value normalization on the M Weimeier cepstrum to determine the Meier cepstrum characteristic;
where M is the number of Mel filters.
Optionally, the performing feature extraction and dimension reduction on the two-dimensional time spectrum through a gamma pass filtering cepstrum coefficient to determine a gamma pass filtering cepstrum feature includes:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
enabling the power spectrum to pass through a Gamma-atom filter group consisting of M gamma filters with different scale parameters and shape parameters, and determining gamma-pass filtering cepstrum parameters;
and obtaining an M-dimensional gamma pass filtering cepstrum by discrete cosine transform after taking natural logarithm of the gamma pass filtering cepstrum parameters, and performing mean normalization on the gamma pass filtering cepstrum to determine the characteristics of the gamma pass filtering cepstrum.
Optionally, the determining the power law normalized cepstrum feature by performing feature extraction and dimension reduction on the two-dimensional time spectrum through the power law normalized cepstrum coefficient includes:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
passing the power spectrum through a Gamma filter group consisting of M gamma filters with different scale parameters and shape parameters to obtain a power spectrum after passing through the filter;
according to the power spectrum after passing through the filter, smoothing each frame and the front and rear frames, filtering out low-frequency parts, carrying out asymmetric noise suppression, and calculating the average power in the middle time;
determining the time average and frequency average transfer functions according to the medium-time average power;
normalizing the time-frequency domain of the original short-time energy spectrum according to the time-average transfer function and the frequency-average transfer function, and determining the energy spectrum after time-frequency normalization;
and performing average power normalization on the time-frequency normalized energy spectrum according to the average power estimation value, performing discrete cosine transform and mean value normalization, and determining power law normalization cepstrum characteristics.
Optionally, the method for identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the cepstrum feature matrix by using the convolutional neural network as a classifier comprises the following steps:
taking the cepstrum feature matrix as an input layer, and constructing a mixed cepstrum coefficient-convolution neural network identification model;
and according to the mixed cepstrum coefficient-convolutional neural network identification model, identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker.
According to another aspect of the present invention, there is also provided a high voltage circuit breaker latent fault audible signal identification system comprising:
the action signal conversion module is used for converting the breaker action sound signals collected by the microphone into a two-dimensional time frequency spectrum with two dimensions of a time domain and a frequency domain;
the characteristic matrix forming module is used for performing characteristic extraction and dimension reduction on the two-dimensional time spectrum through a Mel cepstrum coefficient, a gamma-pass filtering cepstrum coefficient and a power law normalization cepstrum coefficient, determining a Mel cepstrum characteristic, a gamma-pass filtering cepstrum characteristic and a power law normalization cepstrum characteristic, and forming a cepstrum characteristic matrix according to the Mel cepstrum characteristic, the gamma-pass filtering cepstrum characteristic and the power law normalization cepstrum characteristic;
and the fault signal identification module is used for identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker by taking the convolutional neural network as a classifier according to the cepstrum feature matrix.
Optionally, the motion signal conversion module includes:
a Hamming window determination submodule for determining a Hamming window according to the following formula:
wherein w (N) is a Hamming window and N is a length;
and determining a frequency spectrum matrix submodule for carrying out short-time discrete Fourier transform on the circuit breaker action sound signal and determining a two-dimensional time frequency spectrum matrix:
Wherein, X (k) is each frame frequency spectrum of a two-dimensional time-frequency spectrum matrix, k is a frequency point serial number, and X (n) is a circuit breaker action sound signal.
Optionally, a feature matrix module is formed, including:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time spectrum after modulus taking;
a Merr cepstrum parameter determining submodule used for determining the Merr cepstrum parameters by enabling the power spectrum to pass through a Mel scale filter bank with a triangular frequency domain;
determining a Mel cepstrum feature submodule, which is used for obtaining an M Viumei cepstrum through discrete cosine transform after taking the natural logarithm of the Mel cepstrum parameters, and performing mean normalization on the M Viumei cepstrum to determine the Mel cepstrum feature;
where M is the number of Mel filters.
Optionally, a feature matrix module is configured, including:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time frequency spectrum after taking a module;
the gamma-pass filtering cepstrum parameter determining submodule is used for enabling the power spectrum to pass through a gamma filter bank consisting of M gamma filters with different scale parameters and shape parameters and determining gamma-pass filtering cepstrum parameters;
and the gamma-pass filtering cepstrum characteristic determining sub-module is used for obtaining M-dimensional gamma-pass filtering cepstrum through discrete cosine transformation after natural logarithm is taken on the gamma-pass filtering cepstrum parameters, and performing mean normalization on the gamma-pass filtering cepstrum to determine the gamma-pass filtering cepstrum characteristics.
Optionally, a feature matrix module is configured, including:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time spectrum after modulus taking;
the power spectrum filtering submodule is used for enabling the power spectrum to pass through a Gamma filter group consisting of M gamma filters with different scale parameters and shape parameters to obtain a power spectrum after the power spectrum passes through the filter;
the middle-time average power calculation submodule is used for smoothing each frame and the front and rear frames according to the power spectrum after the filter, filtering out a low-frequency part, carrying out asymmetric noise suppression and calculating middle-time average power;
the average transfer function determining submodule is used for determining the time average and frequency average transfer functions according to the medium-time average power;
the energy spectrum submodule after time-frequency normalization is determined, and is used for normalizing the time-frequency domain of the original short-time energy spectrum according to the time-average and frequency-average transfer functions and determining the energy spectrum after time-frequency normalization;
and the power law normalization cepstrum feature determining submodule is used for performing average power normalization on the time-frequency normalized energy spectrum according to the average power estimation value, performing discrete cosine transform and mean value normalization, and determining the power law normalization cepstrum feature.
Optionally, the fault signal identifying module comprises:
a neural network identification model submodule is constructed, and is used for constructing a mixed cepstrum coefficient-convolution neural network identification model by taking the cepstrum feature matrix as an input layer;
and the fault type identification submodule is used for identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the mixed cepstrum coefficient-convolutional neural network identification model.
The robust dc voltage control system 500 in the embodiment of the invention when the VSC is connected to the weak grid corresponds to the robust dc voltage control method 300 in the embodiment of the invention when the VSC is connected to the weak grid, and details thereof are not repeated here.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein. The scheme in the embodiment of the application can be implemented by adopting various computer languages, such as object-oriented programming language Java and transliterated scripting language JavaScript.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.
Claims (12)
1. A method for recognizing audible signals of latent faults of a high-voltage circuit breaker is characterized by comprising the following steps:
converting a breaker action sound signal collected by a microphone into a two-dimensional time frequency spectrum with two dimensions of a time domain and a frequency domain;
performing feature extraction and dimension reduction on the two-dimensional time spectrum through a Mel cepstrum coefficient, a gamma-pass filtering cepstrum coefficient and a power law normalization cepstrum coefficient, determining a Mel cepstrum feature, a gamma-pass filtering cepstrum feature and a power law normalization cepstrum feature, and forming a cepstrum feature matrix according to the Mel cepstrum feature, the gamma-pass filtering cepstrum feature and the power law normalization cepstrum feature;
and taking the convolutional neural network as a classifier, and identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the cepstrum feature matrix.
2. The method of claim 1, wherein converting the breaker action acoustic signal collected by the microphone into a two-dimensional time-frequency spectrum having two dimensions, a time domain and a frequency domain, comprises:
the hamming window is determined according to the following formula:
wherein w (N) is a Hamming window and N is a length;
performing short-time discrete Fourier transform on the breaker action sound signal, and determining a two-dimensional time-frequency spectrum matrix:
Wherein, X (k) is each frame frequency spectrum of a two-dimensional time-frequency spectrum matrix, k is a frequency point serial number, and X (n) is a circuit breaker action sound signal.
3. The method of claim 1, wherein the feature extraction and dimension reduction of the two-dimensional time spectrum are performed by using mel-frequency cepstral coefficients, and determining mel-frequency cepstral features comprises:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
enabling the power spectrum to pass through a Mel scale filter bank with a triangular frequency domain, and determining Mel cepstrum parameters;
obtaining M Weimeier cepstrum by discrete cosine transform after taking natural logarithm of the Meier cepstrum parameter, and carrying out mean value normalization on the M Weimeier cepstrum to determine the Meier cepstrum characteristic;
where M is the number of Mel filters.
4. The method of claim 1, wherein the gamma-pass filtered cepstrum features are determined by performing feature extraction and dimensionality reduction on the two-dimensional time spectrum through gamma-pass filtered cepstrum coefficients, and the method comprises the following steps:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
enabling the power spectrum to pass through a Gamma-tone filter bank consisting of M gamma filters with different scale parameters and shape parameters, and determining gamma-pass filtering cepstrum parameters;
and obtaining an M-dimensional gamma pass filtering cepstrum by discrete cosine transform after taking natural logarithm of the gamma pass filtering cepstrum parameters, and performing mean normalization on the gamma pass filtering cepstrum to determine the characteristics of the gamma pass filtering cepstrum.
5. The method of claim 1, wherein determining the power-law normalized cepstrum features by performing feature extraction and dimensionality reduction on the two-dimensional time spectrum through power-law normalized cepstrum coefficients comprises:
performing modulus operation on each frame of the two-dimensional time spectrum and then squaring to obtain a power spectrum;
passing the power spectrum through a Gamma filter group consisting of M gamma filters with different scale parameters and shape parameters to obtain a power spectrum after passing through the filter;
according to the power spectrum after passing through the filter, smoothing each frame and the front and rear frames, filtering out low-frequency parts, carrying out asymmetric noise suppression, and calculating the average power in the middle time;
determining the time average and frequency average transfer functions according to the medium-time average power;
normalizing the time-frequency domain of the original short-time energy spectrum according to the time-average and frequency-average transfer functions, and determining the energy spectrum after time-frequency normalization;
and performing average power normalization on the time-frequency normalized energy spectrum according to the average power estimation value, performing discrete cosine transform and mean value normalization, and determining power law normalization cepstrum characteristics.
6. The method of claim 1, wherein the fault type identification of the audible acoustic signal of the latent fault of the high-voltage circuit breaker according to the cepstrum feature matrix by using a convolutional neural network as a classifier comprises the following steps:
taking the cepstrum feature matrix as an input layer, and constructing a mixed cepstrum coefficient-convolution neural network identification model;
and according to the mixed cepstrum coefficient-convolutional neural network identification model, identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker.
7. A high voltage circuit breaker latent fault audible signal identification system comprising:
the action signal conversion module is used for converting the breaker action sound signals collected by the microphone into a two-dimensional time frequency spectrum with two dimensions of a time domain and a frequency domain;
the characteristic matrix forming module is used for performing characteristic extraction and dimension reduction on the two-dimensional time spectrum through a Mel cepstrum coefficient, a gamma-pass filtering cepstrum coefficient and a power law normalization cepstrum coefficient, determining a Mel cepstrum characteristic, a gamma-pass filtering cepstrum characteristic and a power law normalization cepstrum characteristic, and forming a cepstrum characteristic matrix according to the Mel cepstrum characteristic, the gamma-pass filtering cepstrum characteristic and the power law normalization cepstrum characteristic;
and the fault signal identification module is used for identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the cepstrum characteristic matrix by taking the convolutional neural network as a classifier.
8. The system of claim 7, wherein the transition action signal module comprises:
a Hamming window determination submodule for determining a Hamming window according to the following formula:
wherein w (N) is a Hamming window and N is a length;
and determining a frequency spectrum matrix submodule for carrying out short-time discrete Fourier transform on the circuit breaker action sound signal and determining a two-dimensional time frequency spectrum matrix:
Wherein, X (k) is each frame frequency spectrum of a two-dimensional time-frequency spectrum matrix, k is a frequency point serial number, and X (n) is a circuit breaker action sound signal.
9. The system of claim 7, wherein the feature matrix module is configured to include:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time spectrum after modulus taking;
a Merr cepstrum parameter determining submodule used for determining the Merr cepstrum parameters by the power spectrum through a Mel scale filter bank with a triangular frequency domain;
determining a Mel cepstrum feature submodule, which is used for obtaining an M Viumei cepstrum through discrete cosine transform after taking the natural logarithm of the Mel cepstrum parameters, and performing mean normalization on the M Viumei cepstrum to determine the Mel cepstrum feature;
where M is the number of Mel filters.
10. The system of claim 7, wherein the feature matrix module is configured to include:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time spectrum after modulus taking;
the gamma-pass filtering cepstrum parameter determining submodule is used for enabling the power spectrum to pass through a gamma filter group consisting of M gamma filters with different scale parameters and shape parameters, and determining gamma-pass filtering cepstrum parameters;
and the gamma pass filtering cepstrum characteristic determining sub-module is used for obtaining an M-dimensional gamma pass filtering cepstrum through discrete cosine transformation after natural logarithm is taken on the gamma pass filtering cepstrum parameters, and performing mean normalization on the gamma pass filtering cepstrum to determine the gamma pass filtering cepstrum characteristics.
11. The system of claim 7, wherein the feature matrix module is configured to include:
the power spectrum submodule is used for obtaining a power spectrum by squaring each frame of frequency spectrum of the two-dimensional time spectrum after modulus taking;
the power spectrum filtering submodule is used for enabling the power spectrum to pass through a Gamma filter bank consisting of M gamma filters with different scale parameters and shape parameters to obtain a power spectrum after the power spectrum passes through the filter;
the calculation middle-time average power submodule is used for smoothing each frame and the front and rear frames according to the power spectrum after passing through the filter, filtering out low-frequency parts, carrying out asymmetric noise suppression and calculating middle-time average power;
the average transfer function determining submodule is used for determining the time average and frequency average transfer functions according to the medium-time average power;
the energy spectrum submodule after time-frequency normalization is determined, and is used for normalizing the time-frequency domain of the original short-time energy spectrum according to the time-average and frequency-average transfer functions and determining the energy spectrum after time-frequency normalization;
and the power law normalization cepstrum feature determining submodule is used for performing average power normalization on the time-frequency normalized energy spectrum according to the average power estimation value, performing discrete cosine transform and mean value normalization, and determining the power law normalization cepstrum feature.
12. The system of claim 7, wherein identifying a fault signal module comprises:
a neural network identification model submodule is constructed, and is used for constructing a mixed cepstrum coefficient-convolution neural network identification model by taking the cepstrum feature matrix as an input layer;
and the fault type identification submodule is used for identifying the fault type of the audible sound signal of the latent fault of the high-voltage circuit breaker according to the mixed cepstrum coefficient-convolutional neural network identification model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210100166.5A CN115910097A (en) | 2022-01-27 | 2022-01-27 | Audible signal identification method and system for latent fault of high-voltage circuit breaker |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210100166.5A CN115910097A (en) | 2022-01-27 | 2022-01-27 | Audible signal identification method and system for latent fault of high-voltage circuit breaker |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115910097A true CN115910097A (en) | 2023-04-04 |
Family
ID=86491961
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210100166.5A Pending CN115910097A (en) | 2022-01-27 | 2022-01-27 | Audible signal identification method and system for latent fault of high-voltage circuit breaker |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115910097A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116577651A (en) * | 2023-07-12 | 2023-08-11 | 中国电力科学研究院有限公司 | Sensor position selection method and device for voiceprint monitoring device of high-voltage circuit breaker |
| CN117131366A (en) * | 2023-10-26 | 2023-11-28 | 北京国电通网络技术有限公司 | Transformer maintenance equipment control method and device, electronic equipment and readable medium |
| CN118364271A (en) * | 2024-06-19 | 2024-07-19 | 国网山东省电力公司邹城市供电公司 | GIS breaker operation state monitoring method and system based on sound and image characteristics |
-
2022
- 2022-01-27 CN CN202210100166.5A patent/CN115910097A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116577651A (en) * | 2023-07-12 | 2023-08-11 | 中国电力科学研究院有限公司 | Sensor position selection method and device for voiceprint monitoring device of high-voltage circuit breaker |
| CN117131366A (en) * | 2023-10-26 | 2023-11-28 | 北京国电通网络技术有限公司 | Transformer maintenance equipment control method and device, electronic equipment and readable medium |
| CN117131366B (en) * | 2023-10-26 | 2024-02-06 | 北京国电通网络技术有限公司 | Transformer maintenance equipment control method and device, electronic equipment and readable medium |
| CN118364271A (en) * | 2024-06-19 | 2024-07-19 | 国网山东省电力公司邹城市供电公司 | GIS breaker operation state monitoring method and system based on sound and image characteristics |
| CN118364271B (en) * | 2024-06-19 | 2024-08-30 | 国网山东省电力公司邹城市供电公司 | GIS breaker operation state monitoring method and system based on sound and image characteristics |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11908455B2 (en) | Speech separation model training method and apparatus, storage medium and computer device | |
| CN109036382B (en) | Audio feature extraction method based on KL divergence | |
| CN115910097A (en) | Audible signal identification method and system for latent fault of high-voltage circuit breaker | |
| WO2019232829A1 (en) | Voiceprint recognition method and apparatus, computer device and storage medium | |
| CN112466326B (en) | Voice emotion feature extraction method based on transducer model encoder | |
| CN108597496A (en) | Voice generation method and device based on generation type countermeasure network | |
| Gurugubelli et al. | Perceptually enhanced single frequency filtering for dysarthric speech detection and intelligibility assessment | |
| CN111785285A (en) | Voiceprint recognition method for home multi-feature parameter fusion | |
| CN108281146A (en) | A kind of phrase sound method for distinguishing speek person and device | |
| CN106941005A (en) | A kind of vocal cords method for detecting abnormality based on speech acoustics feature | |
| CN110111769B (en) | Electronic cochlea control method and device, readable storage medium and electronic cochlea | |
| CN111933186B (en) | Method, device and system for fault identification of on-load tap-changer | |
| CN105448291A (en) | Parkinsonism detection method and detection system based on voice | |
| Ilyas | Pseudo-colored rate map representation for speech emotion recognition | |
| CN112735477B (en) | Voice emotion analysis method and device | |
| CN115346561B (en) | Depression emotion assessment and prediction method and system based on voice characteristics | |
| CN112329819A (en) | Underwater target identification method based on multi-network fusion | |
| CN111145726A (en) | Deep learning-based sound scene classification method, system, device and storage medium | |
| Hasan et al. | Preprocessing of continuous bengali speech for feature extraction | |
| Rahman et al. | Dynamic time warping assisted svm classifier for bangla speech recognition | |
| CN108806725A (en) | Speech differentiation method, apparatus, computer equipment and storage medium | |
| CN113571095B (en) | Speech emotion recognition method and system based on nested deep neural network | |
| Thomas et al. | Acoustic and data-driven features for robust speech activity detection | |
| KR20190135916A (en) | Apparatus and method for determining user stress using speech signal | |
| CN113539243A (en) | Training method of voice classification model, voice classification method and related device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |