CN115691549A - Mechanical voiceprint recognition method and system for convolutional neural network construction based on contact pickup sensor - Google Patents
Mechanical voiceprint recognition method and system for convolutional neural network construction based on contact pickup sensor Download PDFInfo
- Publication number
- CN115691549A CN115691549A CN202211313204.1A CN202211313204A CN115691549A CN 115691549 A CN115691549 A CN 115691549A CN 202211313204 A CN202211313204 A CN 202211313204A CN 115691549 A CN115691549 A CN 115691549A
- Authority
- CN
- China
- Prior art keywords
- voiceprint
- data
- neural network
- dfcnn
- map
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of third-party construction machinery for identifying acoustic prints along an oil and gas storage and transportation pipeline, in particular to a mechanical acoustic print identification method and system for convolutional neural network construction based on a contact type pickup sensor. The method comprises the following steps: step S1: acquiring voiceprint data of a mechanical construction site through a contact type pickup sensor, and obtaining a voiceprint map corresponding to the voiceprint data; step S2: preprocessing the vocal print map, removing the noise of the vocal print map and improving the definition; and step S3: extracting characteristic information of a voiceprint atlas through Fourier transform; and step S4: and identifying the characteristic information of the voiceprint atlas through a preset DFCNN neural network model. The invention can filter out other noises of the environment by adopting the contact sensor to pick up the sound in the soil, thereby better identifying the construction machinery.
Description
Technical Field
The invention relates to the technical field of third-party construction machinery for identifying acoustic prints along an oil and gas storage and transportation pipeline, in particular to a mechanical acoustic print identification method and system for convolutional neural network construction based on a contact type pickup sensor.
Background
In the prior art, a video monitoring technology is adopted, a plurality of point positions are set, a camera is rotated to stay for a certain time for monitoring, meanwhile, a dispatcher is required to perform polling in a monitoring room for a long time to stare at a screen for monitoring, and when an abnormal phenomenon is found, the dispatcher is arranged to deal with the abnormal phenomenon. However, this method may result in missed reports, where firstly the back of the monitoring node of the camera cannot be monitored effectively, secondly the personnel cannot monitor the screen for 24h, and then the number of polling cameras is large, and thus each camera cannot be monitored in real time all day long. Therefore, how to provide a mechanical voiceprint recognition method for construction is a technical problem which needs to be solved urgently by a person skilled in the art.
The voiceprint recognition technology based on deep learning makes great progress through development of many years. Voiceprints are the most important communication mode in daily life, and things, events and the like can be identified through voiceprints at ordinary times. The other side of the phone can be known by voice, and the vehicle is perceived by voiceprints to be approaching gradually behind. And the voiceprint signals are easy to collect, and the propagation distance of the voiceprint signals is long. In the national defense field such as military, voiceprint recognition is used for investigation and recognition of the front enemy and aerial fighters, and is used for detection of abnormal operation state of mechanical equipment in industry, so that the voiceprint is picked up by using a contact sensor, the recognition technology has a certain application prospect in the aspect of recognizing construction machinery, and the sound of construction equipment can be well recognized for safety monitoring.
Disclosure of Invention
The invention aims to provide a method and a system for identifying a mechanical voiceprint based on convolutional neural network construction of a contact pickup sensor.
The invention improves the problem that the prior art fails to effectively monitor due to the fact that a plurality of point rotating cameras stay for a certain time to monitor, and the invention identifies the characteristic information of the voiceprint map based on the preset DFCNN neural network model by the voiceprint identification technology based on deep learning, thereby greatly improving the accuracy of the identification result.
The invention improves the problems that in the prior art, the acoustic lines of the construction machinery are not propagated far in the soil, so that the monitoring and extraction are inaccurate, and the identification result is influenced.
In order to achieve the purpose, the invention provides the following technical scheme:
a mechanical voiceprint recognition method based on convolutional neural network construction of a contact pickup sensor comprises the following steps:
step S1: acquiring voiceprint data of a mechanical construction site through the contact type pickup sensor, and obtaining a voiceprint map corresponding to the voiceprint data;
step S2: preprocessing the voiceprint map to remove noise of the voiceprint map;
and step S3: extracting characteristic information of the voiceprint atlas through Fourier transform;
and step S4: and identifying the characteristic information of the voiceprint atlas through a preset DFCNN neural network model.
In some embodiments of the present application, the preset DFCNN neural network model is obtained by a method including:
step S1: acquiring voiceprint data of a mechanical construction site through the contact type pickup sensor;
step S2: carrying out data division on the acquired voiceprint data to obtain a divided data set, and converting the data set into a voiceprint characteristic map to label a target frame;
and step S3: preprocessing the voiceprint characteristic map, and removing noise of the voiceprint characteristic map so as to improve definition;
and step S4: extracting characteristic information of the voiceprint characteristic map through Fourier transform;
step S5: performing model training on the feature information of the voiceprint feature map through a DFCNN model, and obtaining the preset DFCNN neural network model;
step S6: and performing model evaluation on the preset DFCNN neural network model through the data set in the step S2.
In some embodiments of the present application, the DFCNN model includes 4 convolutional layers, 4 pooling layers, 1 fully-connected layer, the middle layer activation function uses ReLU, the last layer uses softmax, and batch normalization acceleration training is used after each convolutional layer, the DFCNN model outputs 32 features by fourier transforming each frame of data information of the collected voiceprint data and converting the voiceprint data into image data as input, taking time and frequency as two dimensions of the image data, and then convolving the image data using a convolution kernel of 3*3, extracts maximum parameters using max pooling, the pooling layer outputs 64 features, and halves the dimension of the image data by one cnn _ cell so that the feature value becomes 128 and is incorporated into the fully-connected layer.
In some embodiments of the present application, the convolution kernel of the convolutional layer is set to have a size of (3,3), 32 features are extracted and padding is extended to be 1, the convolution step is set to be 1, and the size of the image data after one convolutional layer is (n +2p-f + 1) = (n, n);
the pooling layer sets the size of pool _ size to (2,2) by using a maxpolining method, extracts the largest number in the pool _ size, and after the image data passes through the pooling layer, the dimension of the image data is halved;
connecting each neuron of the fully-connected layer with all neurons of a previous pooling layer connected with the fully-connected layer, wherein an excitation function of each neuron uses a ReLU function, and an output value of a last layer is transferred to sotfmax logistic regression for classification.
In some embodiments of the present application, the step S4 further includes:
when the characteristic information of the voiceprint atlas is identified through the preset DFCNN neural network model, when a voiceprint of the construction machinery is identified, an alarm is given immediately;
the sound pattern of the construction machine comprises sound pattern data of an excavator, a rammer and a traversing machine.
In order to achieve the above object, the present invention also provides a mechanical voiceprint recognition system based on convolutional neural network construction of a contact pickup sensor, which includes:
the contact type pickup sensor is used for acquiring voiceprint data of a mechanical construction site;
the voice print data processing unit is used for processing the voice print data and obtaining a voice print atlas corresponding to the voice print data;
the voiceprint map preprocessing unit is used for preprocessing the voiceprint map and removing noise of the voiceprint map;
the voiceprint map processing unit is used for extracting the characteristic information of the voiceprint map through Fourier transform;
and the data information identification unit is used for identifying the characteristic information of the voiceprint map through a preset DFCNN neural network model.
In some embodiments of the present application, the preset DFCNN neural network model includes:
the acquisition module is used for acquiring voiceprint data of a mechanical construction site through the contact pickup sensor;
the classification module is used for carrying out data division on the acquired voiceprint data and obtaining a divided data set, and the classification module is also used for converting the data set into a voiceprint characteristic map to label a target frame;
the preprocessing module is used for preprocessing the voiceprint characteristic map and removing noise of the voiceprint characteristic map so as to improve definition;
the extraction module is used for extracting the characteristic information of the voiceprint characteristic map through Fourier transform;
the training module is used for carrying out model training on the feature information of the voiceprint feature map through a DFCNN model and obtaining the preset DFCNN neural network model;
an evaluation module for performing model evaluation on the preset DFCNN neural network model through the data set.
In some embodiments of the present application, the DFCNN model includes 4 convolutional layers, 4 pooling layers, 1 fully-connected layer, the middle layer activation function uses ReLU, the last layer uses softmax, and batch normalization acceleration training is used after each convolutional layer, the DFCNN model is used to convolve the image data by fourier transforming each frame of data information of the collected voiceprint data and converting the voiceprint data into image data as input, with time and frequency as two dimensions of the image data, and then using the convolution kernel of 3*3 to convolve the image data, the convolutional layers output 32 features, extract the maximum parameters using max pooling, the pooling layers output 64 features, and halve the dimensions of the image data by cnn _ cell once, so that the feature value becomes 128 and enters the fully-connected layer.
In some embodiments of the present application, the convolution kernel of the convolutional layer has a size of (3,3), 32 features are extracted and padding is extended to 1, the convolution step is set to 1, and the size of the image data after one convolutional layer is (n +2p-f + 1) = (n, n);
the pooling layer is used for setting the size of pool _ size to be (2,2) by using a Maxpooling method, extracting the largest number in the pool _ size, and after the image data passes through the pooling layer, reducing the dimensionality of the image data by half;
each neuron of the fully-connected layer is connected with all neurons of a previous pooling layer connected with the fully-connected layer, wherein the excitation function of each neuron uses a ReLU function, and the output value of the last layer is transferred to sotfmax logistic regression for classification.
In some embodiments of the present application, the data information identification unit further includes:
the alarm unit is used for immediately giving an alarm when the data information identification unit identifies the characteristic information of the voiceprint atlas through the preset DFCNN neural network model and identifies the voiceprint of the construction machine;
the sound pattern of the construction machine comprises sound pattern data of an excavator, a rammer and a traversing machine.
The invention provides a mechanical voiceprint recognition method and a system for convolutional neural network construction based on a contact pickup sensor, and compared with the prior art, the method and the system have the beneficial effects that:
according to the construction machinery voiceprint filtering method, a contact type pickup sensor is used for collecting a construction machinery voiceprint signal sample, the frequency spectrum characteristic of a construction machinery voiceprint is extracted through Fourier transform, a DFCNN model is trained after marking, a trained multi-classifier is formed, then the voiceprint is intelligently detected in real time, based on an algorithm for intelligently identifying the construction machinery voiceprint through the DFCNN, filtering of other voiceprints can be effectively achieved, when the construction machinery voiceprint is identified, real-time alarming is carried out, and therefore the safety risk of damage of a third party of a pipeline is reduced. After the intelligent construction machine recognition system is implemented, intelligent recognition of construction machines can be realized on oil and gas pipeline sites of the whole country, automatic recognition alarm is carried out on third-party construction machines, and intelligent safety control is realized.
Drawings
FIG. 1 is a flow chart of a mechanical voiceprint recognition method of the present invention based on a contact pickup sensor convolutional neural network construction;
FIG. 2 is a flow chart of a method for obtaining a neural network model of the DFCNN default of the present invention;
FIG. 3 is a functional block diagram of a mechanical voiceprint recognition system of the present invention based on a contact pickup sensor convolutional neural network construction.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
In the description of the present application, it is to be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", and the like indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience in describing the present application and simplifying the description, but do not indicate or imply that the referred device or element must have a particular orientation, be constructed in a particular orientation, and be operated, and thus should not be construed as limiting the present application.
The terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present application, "a plurality" means two or more unless otherwise specified.
In the description of the present application, it is to be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be directly connected or indirectly connected through an intermediate member, or they may be connected to each other through an intermediate member. The specific meaning of the above terms in the present application can be understood in a specific case by those of ordinary skill in the art.
In the prior art, the mechanical voiceprint recognition of construction is realized by a video monitoring technology, a plurality of point positions are set, a camera is rotated to stay for a certain time to monitor, meanwhile, a dispatcher needs to watch on a screen for a long time in a monitoring room for monitoring, and the dispatcher arranges the dispatcher to deal with abnormal phenomena. However, this method may result in missed reports, where firstly the back of the monitoring node of the camera cannot be monitored effectively, secondly the personnel cannot monitor the screen for 24h, and then the number of polling cameras is large, and thus each camera cannot be monitored in real time all day long. Therefore, how to provide a mechanical voiceprint recognition method for construction is a technical problem which needs to be solved urgently by the technical personnel in the field.
Therefore, the invention provides a method and a system for identifying the mechanical voiceprint of the convolutional neural network construction based on the contact pickup sensor.
Referring to fig. 1, a disclosed embodiment of the present invention provides a mechanical voiceprint recognition method for convolutional neural network construction based on a contact pickup sensor, including:
step S1: acquiring voiceprint data of a mechanical construction site through a contact type pickup sensor, and obtaining a voiceprint map corresponding to the voiceprint data;
step S2: preprocessing the voiceprint map to remove the noise of the voiceprint map;
and step S3: extracting characteristic information of a voiceprint atlas through Fourier transform;
and step S4: and identifying the characteristic information of the voiceprint atlas through a preset DFCNN neural network model.
It can be understood that the contact type pickup sensor can be buried in soil, and can also be fixedly arranged at a fixed position of a construction site, and the invention can well pick up voiceprints of construction machinery through the contact type pickup sensor to realize real-time monitoring without specific limitation.
In an embodiment of the present application, referring to fig. 2, the predetermined DFCNN neural network model is obtained by the following method, including:
step S1: acquiring voiceprint data of a mechanical construction site through a contact type pickup sensor;
step S2: carrying out data division on the acquired voiceprint data to obtain a divided data set, and converting the data set into a voiceprint characteristic map to label a target frame;
and step S3: preprocessing the voiceprint characteristic map, and removing noise of the voiceprint characteristic map to improve definition;
and step S4: extracting characteristic information of a voiceprint characteristic map through Fourier transform;
step S5: performing model training on the feature information of the voiceprint feature map through a DFCNN model, and obtaining a preset DFCNN neural network model;
step S6: and performing model evaluation on the preset DFCNN neural network model through the data set in the step S2.
It can be understood that the construction machinery voiceprint real-time detection method and the construction machinery voiceprint real-time detection system can acquire a construction machinery voice signal sample through the contact type pickup sensor, adopt Fourier transform to extract frequency spectrum characteristics of the construction machinery voiceprint, train a DFCNN model after marking to form a trained multi-classifier, then intelligently detect the voiceprint in real time, effectively filter other voiceprints, give an alarm in real time after the construction machinery voiceprint is identified, and reduce the safety risk of pipeline third party damage.
In a specific embodiment of the present application, the DFCNN model includes 4 convolutional layers, 4 pooling layers, and 1 fully-connected layer, the middle layer activation function uses ReLU, the last layer uses softmax, a batch normalization acceleration training is used after each convolutional layer, the DFCNN model performs fourier transform on each frame of data information of the collected voiceprint data, converts the voiceprint data into image data as input, uses time and frequency as two dimensions of the image data, then convolutes the image data using the convolution kernel of 3*3, the convolutional layers output 32 features, extracts a maximum parameter using maximum pooling maxpopooling, the pooling layers output 64 features, and halves the dimension of the image data by once cnn _ cell, so that a feature value becomes 128, and the fully-connected layer is accessed.
In one embodiment of the present application, the size of the convolution kernel of the convolutional layer is set to (3,3), 32 features are extracted and padding is extended to 1, the convolution step is set to 1, and after one convolutional layer, the size of the image data is (n +2p-f + 1) = (n, n);
the pooling layer sets the size of pool _ size to (2,2) by using a Maxpooling method, extracts the largest number in the pool _ size, and reduces the dimensionality of the image data by half after the image data passes through the pooling layer;
connecting each neuron of the fully-connected layer with all neurons of a previous pooling layer connected by the fully-connected layer, wherein the excitation function of each neuron uses the ReLU function, and the output value of the last layer is passed to sotfmax logistic regression for classification.
In a specific embodiment of the present application, step S4 further includes:
when the characteristic information of the voiceprint atlas is identified through a preset DFCNN neural network model, when the voiceprint of the construction machinery is identified, an alarm is given immediately;
the sound pattern of the construction machine comprises sound pattern data of an excavator, a rammer and a traversing machine.
Based on the same technical concept, referring to fig. 3, the present invention also provides a mechanical voiceprint recognition system based on convolutional neural network construction of a contact pickup sensor, which includes:
the contact type pickup sensor is used for acquiring voiceprint data of a mechanical construction site;
the data processing unit is used for processing the voiceprint data and obtaining a voiceprint map corresponding to the voiceprint data;
the voiceprint atlas preprocessing unit is used for preprocessing the voiceprint atlas and removing the noise of the voiceprint atlas;
the voiceprint map processing unit is used for extracting the characteristic information of the voiceprint map through Fourier transform;
and the data information identification unit is used for identifying the characteristic information of the voiceprint map through a preset DFCNN neural network model.
It can be understood that the contact type pickup sensor can be buried in soil, and can also be fixedly arranged at a fixed position of a construction site, and the invention can well pick up voiceprints of construction machinery through the contact type pickup sensor to realize real-time monitoring without specific limitation.
In a specific embodiment of the present application, the preset DFCNN neural network model includes:
the acquisition module is used for acquiring voiceprint data of a mechanical construction site through the contact type pickup sensor;
the classification module is used for carrying out data division on the acquired voiceprint data and obtaining a divided data set, and the classification module is also used for converting the data set into a voiceprint characteristic map to label the target frame;
the preprocessing module is used for preprocessing the vocal print characteristic map and removing noise of the vocal print characteristic map so as to improve definition;
the extraction module is used for extracting the characteristic information of the voiceprint characteristic map through Fourier transform;
the training module is used for carrying out model training on the feature information of the voiceprint feature map through the DFCNN model and obtaining a preset DFCNN neural network model;
and the evaluation module is used for carrying out model evaluation on the preset DFCNN neural network model through the data set.
The method has the advantages that the construction machinery voice signal samples are collected through the contact type pickup sensor, the Fourier transform is adopted, the frequency spectrum characteristics of the construction machinery voiceprints are extracted, the DFCNN model is trained after the marks are marked, the trained multi-classifier is formed, then the voiceprints are intelligently detected in real time, other voiceprints can be effectively filtered, the alarm is given in real time after the construction machinery voiceprints are identified, and the safety risk of pipeline third party damage is reduced.
In a specific embodiment of the present application, the DFCNN model includes 4 convolutional layers, 4 pooling layers, and 1 fully-connected layer, the middle layer activation function uses ReLU, the last layer uses softmax, and a batch normalization acceleration training is used after each convolutional layer, the DFCNN model is configured to perform fourier transform on each frame of data information of the collected voiceprint data, convert the voiceprint data into image data as input, use time and frequency as two dimensions of the image data, then perform convolution on the image data using 3*3 convolution, output 32 features by the convolutional layer, extract the maximum parameter using max pooling, output 64 features by the pooling layer, and halve the dimension of the image data through cnn _ cell once, so that the feature value becomes 128, and access to the fully-connected layer.
In one embodiment of the present application, the convolution kernel of the convolutional layer has a size of (3,3), 32 features are extracted and padding is extended to 1, the convolution step is set to 1, and after one convolutional layer, the size of the image data is (n +2p-f + 1) = (n, n);
the pooling layer is used for setting the size of pool _ size to be (2,2) by using a Maxpooling method, extracting the largest number in the pool _ size, and after the image data passes through the pooling layer, reducing the dimensionality of the image data by half;
each neuron of the fully-connected layer is connected to all neurons of the previous pooling layer connected to the fully-connected layer, wherein the excitation function of each neuron uses the ReLU function, and the output values of the last layer are passed to the sotfmax logistic regression for classification.
In a specific embodiment of the present application, the data information identification unit further includes:
the alarm unit is used for immediately alarming when the data information identification unit identifies the characteristic information of the voiceprint map through the preset DFCNN neural network model and identifies the voiceprint of the construction machinery;
the sound pattern of the construction machine comprises sound pattern data of an excavator, a rammer and a traversing machine.
According to the first concept of the invention, the voiceprint recognition technology based on deep learning is adopted, the characteristic information of the voiceprint map is recognized based on the preset DFCNN neural network model, the accuracy of the recognition result is greatly improved, and in addition, the system is automatically controlled to operate, so that a large amount of manual labor force can be saved.
According to the second concept of the present invention, the sound in the soil is picked up by using the touch sensor, so that other noises in the environment can be filtered, and the construction machine can be better identified. .
The voiceprint recognition technology based on deep learning makes great progress through development for many years, and is industrially used for detecting the abnormal operation state of mechanical equipment, so that the voiceprint is picked up by using the contact type sensor, the recognition technology has a certain application prospect in the aspect of recognizing construction machinery, and the sound of construction machines and tools can be well recognized for safety monitoring. In summary, according to the mechanical voiceprint recognition method and system for convolutional neural network construction based on the contact pickup sensor, the contact pickup sensor is used for collecting a sound signal sample of construction machinery, fourier transform is used for extracting the frequency spectrum characteristic of the voiceprint of the construction machinery, a DFCNN model is trained after marking, a trained multi-classifier is formed, then the voiceprint is intelligently detected in real time, an algorithm for intelligently recognizing the voiceprint of the construction machinery based on the DFCNN can effectively filter other voiceprints, and when the voiceprint of the construction machinery is recognized, real-time alarm is given, so that the safety risk of third-party damage of a pipeline is reduced.
The above description is only an embodiment of the present invention, but not intended to limit the scope of the present invention, and any structural changes made according to the present invention should be considered as being limited within the scope of the present invention without departing from the spirit of the present invention.
It can be clearly understood by those skilled in the art that, for convenience and simplicity of description, the specific working process and related description of the system described above may refer to the corresponding process in the foregoing method embodiment, and details are not described herein again.
It should be noted that, the system provided in the foregoing embodiment is only illustrated by dividing the functional modules, and in practical applications, the functions may be distributed by different functional modules according to needs, that is, the modules or steps in the embodiment of the present invention are further decomposed or combined, for example, the modules in the foregoing embodiment may be combined into one module, or may be further split into multiple sub-modules, so as to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the modules or steps, and are not to be construed as unduly limiting the present invention.
Those of skill in the art would appreciate that the various illustrative modules, method steps, and modules described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that programs corresponding to the software modules, method steps may be located in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. To clearly illustrate this interchangeability of electronic hardware and software, various illustrative components and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as electronic hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The terms "comprises," "comprising," or any other similar term are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
So far, the technical solutions of the present invention have been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of the present invention is obviously not limited to these specific embodiments. Equivalent changes or substitutions of related technical features can be made by those skilled in the art without departing from the principle of the invention, and the technical scheme after the changes or substitutions can fall into the protection scope of the invention.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention.
Claims (10)
1. A mechanical voiceprint recognition method for convolutional neural network construction based on a contact pickup sensor is characterized by comprising the following steps:
step S1: acquiring voiceprint data of a mechanical construction site through the contact type pickup sensor, and obtaining a voiceprint map corresponding to the voiceprint data;
step S2: preprocessing the voiceprint map to remove noise of the voiceprint map;
and step S3: extracting characteristic information of the voiceprint atlas through Fourier transform;
and step S4: and identifying the characteristic information of the voiceprint atlas through a preset DFCNN neural network model.
2. The mechanical voiceprint recognition method for convolutional neural network construction based on the contact pickup sensor, as claimed in claim 1, wherein the preset DFCNN neural network model is obtained by the following method, including:
step S1: acquiring voiceprint data of a mechanical construction site through the contact type pickup sensor;
step S2: carrying out data division on the acquired voiceprint data to obtain a divided data set, and converting the data set into a voiceprint characteristic map to label a target frame;
and step S3: preprocessing the voiceprint characteristic map, and removing noise of the voiceprint characteristic map so as to improve definition;
and step S4: extracting characteristic information of the voiceprint characteristic map through Fourier transform;
step S5: performing model training on the feature information of the voiceprint feature map through a DFCNN model, and obtaining the preset DFCNN neural network model;
step S6: and performing model evaluation on the preset DFCNN neural network model through the data set in the step S2.
3. The mechanical voiceprint recognition method based on the convolutional neural network construction of the contact pickup sensor as claimed in claim 2,
the DFCNN model comprises 4 convolutional layers, 4 pooling layers and 1 fully-connected layer, wherein a ReLU is used as an intermediate layer activation function, softmax is used as a last layer, batch normalization acceleration training is used after each convolutional layer, the DFCNN model performs Fourier transform on each frame of data information of the collected voiceprint data, converts the voiceprint data into image data as input, takes time and frequency as two dimensions of the image data, then convolutes the image data by using a convolution kernel of 3*3, outputs 32 features, extracts maximum parameters by using maximum pooling Maxpooling, outputs 64 features by the pooling layers, and halves the dimensions of the image data by one time cnn _ cell so that a feature value becomes 128 and is connected into the fully-connected layer.
4. The mechanical voiceprint recognition method based on the convolutional neural network construction of the contact pickup sensor as claimed in claim 3,
setting the size of a convolution kernel of the convolution layer to be (3,3), extracting 32 features and expanding padding to be 1, setting the convolution step size to be 1, and setting the size of the image data to be (n +2p-f + 1) × (n +2p-f + 1) = (n, n) after one convolution layer;
the pooling layer sets the size of pool _ size to (2,2) by using a maxpoloring method, extracts the largest number in the pool _ size, and after the image data passes through the pooling layer, the dimension of the image data is halved;
connecting each neuron of the fully-connected layer with all neurons of a previous pooling layer connected with the fully-connected layer, wherein an excitation function of each neuron uses a ReLU function, and an output value of a last layer is transferred to sotfmax logistic regression for classification.
5. The mechanical voiceprint recognition method based on convolutional neural network construction of a contact pickup sensor according to claim 1, wherein the step S4 further comprises:
when the characteristic information of the voiceprint atlas is identified through the preset DFCNN neural network model, when a voiceprint of the construction machinery is identified, an alarm is given immediately;
the sound pattern of the construction machine comprises sound pattern data of an excavator, a rammer and a traversing machine.
6. The utility model provides a machinery voiceprint identification system of convolution neural network construction based on contact pickup sensor which characterized in that includes:
the contact type pickup sensor is used for acquiring voiceprint data of a mechanical construction site;
the voice print data processing unit is used for processing the voice print data and obtaining a voice print atlas corresponding to the voice print data;
the voiceprint map preprocessing unit is used for preprocessing the voiceprint map and removing noise of the voiceprint map;
the voiceprint map processing unit is used for extracting the characteristic information of the voiceprint map through Fourier transform;
and the data information identification unit is used for identifying the characteristic information of the voiceprint map through a preset DFCNN neural network model.
7. The mechanical voiceprint recognition system based on the convolutional neural network construction of the contact pickup sensor as claimed in claim 6, wherein the preset DFCNN neural network model comprises:
the acquisition module is used for acquiring voiceprint data of a mechanical construction site through the contact pickup sensor;
the classification module is used for carrying out data division on the acquired voiceprint data and obtaining a divided data set, and the classification module is also used for converting the data set into a voiceprint characteristic map to label a target frame;
the preprocessing module is used for preprocessing the voiceprint characteristic map and removing noise of the voiceprint characteristic map so as to improve definition;
the extraction module is used for extracting the characteristic information of the voiceprint characteristic map through Fourier transform;
the training module is used for carrying out model training on the feature information of the voiceprint feature map through a DFCNN model and obtaining the preset DFCNN neural network model;
an evaluation module for performing model evaluation on the preset DFCNN neural network model through the data set.
8. The system of claim 7, wherein the convolutional neural network based on contact pickup sensor is constructed by mechanical voiceprint recognition,
the DFCNN model comprises 4 convolutional layers, 4 pooling layers and 1 fully-connected layer, wherein a ReLU is used as an intermediate layer activation function, softmax is used as a last layer, batch normalization accelerated training is used after each convolutional layer, the DFCNN model is used for performing Fourier transform on each frame of data information of the collected voiceprint data, converting the voiceprint data into image data as input, taking time and frequency as two dimensions of the image data, then convolving the image data by using a convolution kernel of 3*3, outputting 32 features by the convolutional layers, extracting maximum parameters by using maximum pooling Maxpoiling, outputting 64 features by the pooling layers, halving the dimensions of the image data through one cnn _ cell, and enabling a feature value to be changed into 128 and be connected into the fully-connected layer.
9. The system of claim 8, wherein the convolutional neural network based on touch pick-up sensor is constructed by a mechanical voiceprint recognition system,
the convolution kernel of the convolution layer has the size of (3,3), 32 features are extracted and padding is extended to be 1, the convolution step is set to be 1, and after one convolution layer, the size of the image data is (n +2p-f + 1) × (n +2p-f + 1) = (n, n);
the pooling layer is used for setting the size of pool _ size to be (2,2) by using a Maxpooling method, extracting the largest number in the pool _ size, and after the image data passes through the pooling layer, reducing the dimensionality of the image data by half;
each neuron of the fully-connected layer is connected with all neurons of a previous pooling layer connected with the fully-connected layer, wherein the excitation function of each neuron uses a ReLU function, and the output value of the last layer is transferred to sotfmax logistic regression for classification.
10. The system of claim 6, wherein the data information recognition unit further comprises:
the alarm unit is used for immediately giving an alarm when the data information identification unit identifies the characteristic information of the voiceprint atlas through the preset DFCNN neural network model and identifies the voiceprint of the construction machine;
the sound pattern of the construction machine comprises sound pattern data of an excavator, a rammer and a traversing machine.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211313204.1A CN115691549A (en) | 2022-10-25 | 2022-10-25 | Mechanical voiceprint recognition method and system for convolutional neural network construction based on contact pickup sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211313204.1A CN115691549A (en) | 2022-10-25 | 2022-10-25 | Mechanical voiceprint recognition method and system for convolutional neural network construction based on contact pickup sensor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115691549A true CN115691549A (en) | 2023-02-03 |
Family
ID=85098950
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211313204.1A Pending CN115691549A (en) | 2022-10-25 | 2022-10-25 | Mechanical voiceprint recognition method and system for convolutional neural network construction based on contact pickup sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115691549A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117235583A (en) * | 2023-11-13 | 2023-12-15 | 国网浙江省电力有限公司双创中心 | Monitoring method and system for GIS breaker actuating mechanism |
-
2022
- 2022-10-25 CN CN202211313204.1A patent/CN115691549A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117235583A (en) * | 2023-11-13 | 2023-12-15 | 国网浙江省电力有限公司双创中心 | Monitoring method and system for GIS breaker actuating mechanism |
| CN117235583B (en) * | 2023-11-13 | 2024-01-30 | 国网浙江省电力有限公司双创中心 | Monitoring method and system for GIS breaker actuating mechanism |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111157099B (en) | Distributed optical fiber sensor vibration signal classification method and identification classification system | |
| CN110570613A (en) | Fence vibration intrusion positioning and mode identification method based on distributed optical fiber system | |
| CN109616140B (en) | Abnormal sound analysis system | |
| CN105841961A (en) | Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network | |
| CN109816987B (en) | Electronic police law enforcement snapshot system for automobile whistling and snapshot method thereof | |
| CN110398647B (en) | Transformer state monitoring method | |
| CN106407993A (en) | Intelligent voice robot system based on image recognition technology and method thereof | |
| CN115691549A (en) | Mechanical voiceprint recognition method and system for convolutional neural network construction based on contact pickup sensor | |
| CN109060371A (en) | A kind of auto parts and components abnormal sound detection device | |
| CN115859078A (en) | Millimeter wave radar fall detection method based on improved Transformer | |
| CN113990303B (en) | Environmental sound identification method based on multi-resolution cavity depth separable convolution network | |
| CN109447199A (en) | A kind of multi-modal criminal's recognition methods and system based on step information | |
| WO2023102527A1 (en) | System and method for gas detection at a field site using multiple sensors | |
| CN116092119A (en) | Human behavior recognition system based on multidimensional feature fusion and working method thereof | |
| CN115170942A (en) | Fish behavior identification method with multilevel fusion of sound and vision | |
| CN113484489A (en) | Remote monitoring and early warning method for water eutrophication | |
| CN114420137A (en) | Wild animal detection method and equipment | |
| CN114692687A (en) | Underwater sound signal identification method and training method of underwater sound signal identification model | |
| CN114093106B (en) | Intrusion signal alarm method and system based on EfficientNET classification network | |
| CN117390413B (en) | Recognition method for distributed power optical fiber vibration signal noise reduction and time sequence feature extraction | |
| CN108615018A (en) | Object state identification method based on the extraction of time domain histogram feature | |
| CN109686032B (en) | Aquaculture organism anti-theft monitoring method and system | |
| CN113792774B (en) | Intelligent fusion sensing method for underwater targets | |
| CN109389994A (en) | Identification of sound source method and device for intelligent transportation system | |
| CN114925722A (en) | Perimeter security intrusion signal detection method based on generalized S transformation and transfer learning |
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 |