CN112562739A - Large-scale feeding type broiler respiratory disease sound data analysis system - Google Patents
Large-scale feeding type broiler respiratory disease sound data analysis system Download PDFInfo
- Publication number
- CN112562739A CN112562739A CN202011309454.9A CN202011309454A CN112562739A CN 112562739 A CN112562739 A CN 112562739A CN 202011309454 A CN202011309454 A CN 202011309454A CN 112562739 A CN112562739 A CN 112562739A
- Authority
- CN
- China
- Prior art keywords
- sound
- sound data
- data
- cloud
- audio
- 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
- 241000287828 Gallus gallus Species 0.000 title claims abstract description 36
- 208000023504 respiratory system disease Diseases 0.000 title claims abstract description 13
- 238000007405 data analysis Methods 0.000 title claims abstract description 11
- 238000013528 artificial neural network Methods 0.000 claims abstract description 42
- 238000004458 analytical method Methods 0.000 claims abstract description 21
- 230000002159 abnormal effect Effects 0.000 claims abstract description 12
- 201000010099 disease Diseases 0.000 claims abstract description 11
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims abstract description 11
- 238000005457 optimization Methods 0.000 claims abstract description 6
- 244000144992 flock Species 0.000 claims abstract description 5
- 230000005236 sound signal Effects 0.000 claims description 14
- 238000000034 method Methods 0.000 claims description 11
- 238000001228 spectrum Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 6
- 230000003595 spectral effect Effects 0.000 claims description 5
- 206010011224 Cough Diseases 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 claims description 4
- 206010041232 sneezing Diseases 0.000 claims description 4
- 230000000007 visual effect Effects 0.000 claims description 4
- 230000001360 synchronised effect Effects 0.000 claims description 3
- 235000013330 chicken meat Nutrition 0.000 abstract description 6
- 238000009395 breeding Methods 0.000 abstract description 2
- 230000001488 breeding effect Effects 0.000 abstract description 2
- 238000011176 pooling Methods 0.000 description 12
- 238000013527 convolutional neural network Methods 0.000 description 10
- 230000004913 activation Effects 0.000 description 6
- 230000003190 augmentative effect Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000012800 visualization Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000003014 reinforcing effect Effects 0.000 description 2
- 238000005728 strengthening Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000010827 pathological analysis Methods 0.000 description 1
- 244000144977 poultry Species 0.000 description 1
- 235000013594 poultry meat Nutrition 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
- G10L25/66—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination for extracting parameters related to health condition
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/48—Other medical applications
- A61B5/4803—Speech analysis specially adapted for diagnostic purposes
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/28—Constructional details of speech recognition systems
- G10L15/30—Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L17/00—Speaker identification or verification techniques
- G10L17/26—Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Pathology (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Heart & Thoracic Surgery (AREA)
- Biomedical Technology (AREA)
- Epidemiology (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention belongs to the technical field of electronic information, and particularly relates to a system for analyzing sound data of respiratory diseases of mass-fed broiler chickens, which comprises the following components: collecting sound data; analyzing sound data; the sound data are transmitted from the cloud to the enhanced neural network of the cloud server for analysis, and the analysis result is uploaded to the cloud backup; optimizing the algorithm to optimize the result of the sound data analysis in the step two; constructing a positioning model and measuring the position of a sound source area; and data are collected and analyzed for a long time, algorithm optimization is carried out, and the analysis result is transmitted to a local server, so that early warning and warning of disease attack are realized. The invention can enable the machine to replace manual work to monitor the cry of the chicken flock continuously 24 hours a day, and if abnormal cry is frequently found, the breeding factory can be reminded to pay attention and treat at the early stage of disease occurrence, thereby achieving the purposes of reducing cost and improving profit.
Description
Technical Field
The invention belongs to the technical field of electronic information, and particularly relates to a large-scale feeding type broiler respiratory disease sound data analysis system.
Background
According to the estimation of poultry pathological analysis experts of Shandong agriculture university, the accuracy rate of manually judging the association between the sound and the respiratory diseases is expected to be more than 95%, at present, the culturist with experience in the large-scale cultivation factory can judge whether the broiler chicken is sick or not according to the sound of the broiler chicken, but the manual judgment often has the problems of delay or omission, so that the early treatment can not be started any more, the medicine, material and other costs of the cultivation factory are increased, and the death rate is increased.
Disclosure of Invention
In order to solve the problems mentioned in the background technology, the invention discloses a large-scale feeding type broiler respiratory disease sound data analysis system.
The design purpose of the scheme is to enable the machine to replace manual work to monitor the cry of the chicken flocks continuously for 24 hours a day, and if the abnormal cry is frequently found, the breeding plant can be reminded to pay attention to and treat the disease in the early stage, so that the cost is reduced, and the profit is improved. The technical scheme is as follows:
a system for analyzing respiratory disease sound data of mass-feeding type broiler chickens, which comprises,
the method comprises the following steps: the method comprises the steps that sound data are collected, a plurality of sound collection sensors are arranged in a coop, the sound collection sensors transmit the collected sound data to a cloud end through a network, and the sound data are converted into audio data in an OPUS format at the cloud end to be stored;
step two: analyzing the sound data, wherein the sound data is transmitted from the cloud to an enhanced neural network of a cloud server for analysis, and an analysis result is uploaded to a cloud backup;
step three: performing algorithm optimization, namely transmitting the abnormal sound data obtained by analysis in the step two to a local server from a cloud end, performing expert marking on the local server to generate a local label file, and additionally updating the local label file to a cloud end label file in the cloud end server to optimize the sound data analysis result in the step two;
step four: constructing a positioning model, measuring the direction of a sound source area, setting the relative position between a sound collecting sensor and the broiler flock, and fixing a sound receiving model to position the specific position of the broiler chicken which emits abnormal sound;
step five: and (4) collecting and analyzing data for a long time, optimizing the algorithm in the third step, and transmitting the analysis result in the second step to a local server to realize early warning and warning of the onset of disease.
Further, the neural network for analyzing the sound data includes:
and (3) audio decomposition algorithm: performing Fourier transform on the audio signal in the time domain, and obtaining the frequency spectrum expression of the audio signal in frequency and energy abundance (amplitude and synthetic index of Fourier expansion coefficient) indexes in a minimum time window (1/420 seconds);
for audio data, the original fourier transform equation is as follows:
Gf(∈,u)=∫f(t)g(t-u)ej∈tdt
for short-time fourier transforms there are:
where z (u) is the source signal and g (u-t) is the window function
Each piece of the analyzed audio data produces three graphs, i.e., spectral representations, wherein the spectral representations of the entire audio data are presented on a "FULL speech" graph; at a specific frequency, the audio frequency moves forward along a time axis, and the frequency spectrum with uniform audio abundance is completely presented on a BACKGROUND graph; moving forward along a time axis at a specific frequency, and when a singular point appears in an audio abundance function relative to equal uniformly distributed numerical values before and after the time point at a specific time point, the frequency spectrum expression of the singular point appears on a FOREGROUND graph; therefore, the technology can generate a visual image of sound frequency and detect abnormal points for the examination of experts or chicken farm staff;
and (3) analysis algorithm: after the decomposition algorithm converts and presents the audio data, the expert distinguishes and marks the sound characteristics contained in the audio data according to the audio data and the three images, the sound characteristics comprise coughing, snorting, sneezing and vocalizing sounds, local label files are formed after marking, the local label files are synchronized to the cloud to form cloud label files for training the robust neural network, the robust neural network preliminarily forms parameterized characteristics with identification capability, and then the parameterized characteristics with stable identification capability of the robust neural network are optimized through the algorithm in step three.
The reinforced neural network part introduces a CNN structure which has strong feature extraction capability, and mainly adopts a convolution layer, a pooling layer, an activation function layer, a full-link layer and an output layer. For the convolutional layer, the following function is shown:
whereinIs the weight of the l-th layer,is the bias of the l-th layer.Is the input tile for the (i, j) position of the l-th layer. The weights of the feature maps are shared (weight sharing reduces the amount of parameters). One of the biggest advantages of the CNN structure is the weight sharing mechanism, because it can significantly reduce the computational complexity when the number of parameters is smaller, and also make training and optimization simpler. Pooling layers are an important module in CNN architecture, and the most important goal of such layers is to reduce the size of the feature map by fusing sub-regions using some function, such as averaging or max/min:
wherein y isi,j,kIs the output of the pooling operator at position (i, j) of the kth feature map, ai,j,kIs the pooling region R in the kth feature mapijCharacteristic value at the middle position (m, n). It is particularly noted that this formula corresponds to an average pooling when p is 1, and becomes a maximum pooling as p approaches infinity. Commonly used activation functions are the ReLU activation function, the leakyreu function, which is defined as follows:
ai,j,k=max(zi,j,k,0)
to avoid the gradient vanishing problem, we can choose to use the LeakyReLU function:
wherein λ is in the range of (0, 1). Leaky ReLU does not force the negative part to be zero, but instead it allows for a small non-zero gradient. For a fully connected layer, the following function is expressed:
after the classification output is obtained, a Softmax loss function can be calculated between the predicted value and the true value of the output:
and in this way optimize our entire CNN network.
Further, the system provides a corresponding interface to support other systems to call data, provides a sensor interface, and other systems can control the sensors according to the sensor interface, specifically can control the number of the sensors, and modifies the settings of the sensors, such as the maximum frequency (8192Hz), the minimum frequency (0Hz), and the like of the sound signals collected by the sensors. Other systems can call the sensor interface to obtain the parameter setting of the sensor, so that other systems can know the state of the sensor conveniently; the module also provides calling of the reinforced neural network for other systems, specifically, other systems can set certain parameters of the reinforced neural network through the interface and can also obtain the state of the reinforced neural network so that other systems can know the state of the reinforced neural network conveniently. For example, the confidence level of the augmented neural network can be set, and the recognition of the augmented neural network can be quantitatively accurate by setting different confidence values. The module also provides an acquisition and control interface of cloud data for other systems. Other systems can call the interface to acquire the sound signal data stored on the cloud server, or modify the sound signal data, the tag data and the like stored on the cloud service area, and can also upload the sound data collected by the sensor through the interface and upload the tag data marked manually.
Further, a filter is installed inside the sound collection sensor.
The invention has the beneficial effects that:
1. the sound data analysis system for respiratory diseases of the large-scale raised broiler chickens utilizes the sensors, and efficiently filters sound signals collected by the sensors through the filters, so that accurate collection of the sound signals is realized.
2. Compared with the traditional manual screening, the neural network has the advantages of high identification precision, less occupied resources and the like, and can better help an administrator to manage the broilers. The human expert discriminates and labels the sound characteristics contained in the audio data according to the audio and image data, and particularly pays attention to the discrimination and judgment of coughing, snorting, sneezing and vocalizing sounds to form the label. The CNN structure quoted here carries out feature capture on the audio data containing the label, and the parameterization depth and strength of the response feature are continuously increased, so that the process becomes artificial neural network training, and the process is increased to extract and judge the feature. After a large amount of data training, the CNN structure initially has the ability of identifying whether the audio data contains the sound features of interest and the marking behavior. Finally, the identification efficiency of the reinforced neural network is higher through the reinforcement of the network.
3. The invention can provide real-time monitoring for 24 hours a day all day, when abnormal sound occurs, the sensor can collect the sound in time and transmit the sound to the neural network, and after the neural network identification is strengthened, the result is immediately notified to the administrator, so that the administrator can process and treat the broilers with diseases in a short time.
4. The invention can also store the sick data of the broilers and carry out visual analysis on the data so as to facilitate better management of the broilers by administrators and prevent certain diseases which are easy to appear in advance. The reinforcing neural network can store input data on a cloud end or a local disk, and an administrator can obtain the data at any time for analysis. Meanwhile, the new scheme can also visualize data, so that an administrator can visually know the disease condition of the broiler chicken, and the new scheme is of great help for the administrator to prevent diseases of the broiler chicken.
5. The system for analyzing the respiratory disease sound data of the large-scale raised broiler chickens is high in identification precision and efficiency; the average age of the broiler chicken is small (about 45 days), the difference of physique characteristics related to sound is smaller than that of adult chicken, and the early warning effectiveness can be 1-2 thousandth in the broiler chicken feeding environment by using the developed algorithm module. Through strengthening the neural network, after training, the precision of discernment can reach expert's level, and the discernment time is short, has very high efficiency, satisfies the requirement of real-time. And because manual intervention is not needed, the scheme can save time and labor.
Drawings
Fig. 1 is a schematic view of the flow structure of the present solution.
Detailed Description
In order to make the technical solution of the present invention more clear and definite for those skilled in the art, the technical solution of the present invention is described in detail below with reference to fig. 1, but the embodiment of the present invention is not limited thereto.
A system for analyzing respiratory disease sound data of mass-feeding type broiler chickens, which comprises,
the method comprises the following steps: the method comprises the steps that sound data are collected, a plurality of sound collection sensors are arranged in a coop, the sound collection sensors transmit the collected sound data to a cloud end through a network, and the sound data are converted into audio data in an OPUS format at the cloud end to be stored;
and filtering invalid data collected by the sensor in the background of large-range feeding of the chicken coop. When recording audio data with a sensor, a mobile phone or a recording pen, the effective identification range should be tested first. The sensor collects data in real time, and the data of a certain time length is analyzed at regular intervals. In the method, 30 minutes are used as initial setting at regular intervals; here, "fixed time length" was set to 5 minutes as an initial setting.
The filter is also needed to be installed in the sensor, and the filter is used for filtering invalid data collected by the sensor, namely, the invalid data collected by the sensor is filtered, so that more effective data can be obtained in the process of data conversion later, and the trained reinforced neural network has higher identification precision and better robustness.
Step two: analyzing the sound data, wherein the sound data is transmitted from the cloud to an enhanced neural network of a cloud server for analysis, and an analysis result is uploaded to a cloud backup; the cloud server is a central processing unit, and a computer and other embedded devices can also be used as the central processing unit so as to operate a network strengthening algorithm.
Step three: performing algorithm optimization, namely transmitting the abnormal sound data obtained by analysis in the step two to a local server from a cloud end, performing expert marking on the local server to generate a local label file, and additionally updating the local label file to a cloud end label file in the cloud end server to optimize the sound data analysis result in the step two;
data communication needs to be established between a central processing unit (cloud) and a local terminal (terminal equipment such as a notebook PC or a mobile phone). The communication may be local area network (the augmented neural network is deployed on a local server) or the internet (the augmented neural network is deployed in the cloud). After the data communication is established, the data can be transmitted to the enhanced neural network after the local terminal receives the data collected by the sensor, the network identifies the sound data, and the result is returned to the local terminal equipment.
Step four: constructing a positioning model, measuring the direction of a sound source area, setting the relative position between a sound collecting sensor and the broiler flock, and fixing a sound receiving model to position the specific position of the broiler chicken which emits abnormal sound;
step five: and (4) collecting and analyzing data for a long time, optimizing the algorithm in the third step, and transmitting the analysis result in the second step to a local server to realize early warning and warning of the onset of disease. The local server is provided with a visualization device, the reinforcing neural network needs to display an analysis result on the visualization device after identifying the data transmitted by the sensor, and the visualization device is helpful for a manager to monitor and manage the broilers in real time and analyze diseases.
Further, the neural network for analyzing the sound data includes:
and (3) audio decomposition algorithm: performing Fourier transform on the audio signal in the time domain, and obtaining the frequency spectrum expression of the audio signal in frequency and energy abundance (amplitude and synthetic index of Fourier expansion coefficient) indexes in a minimum time window (1/420 seconds);
for audio data, the original fourier transform equation is as follows:
Gf(∈,u)=∫f(t)g(t-u)ej∈tdt
for short-time fourier transforms there are:
where z (u) is the source signal and g (u-t) is the window function
Each piece of the analyzed audio data produces three graphs, i.e., spectral representations, wherein the spectral representations of the entire audio data are presented on a "FULL speech" graph; at a specific frequency, the audio frequency moves forward along a time axis, and the frequency spectrum with uniform audio abundance is completely presented on a BACKGROUND graph; moving forward along a time axis at a specific frequency, and when a singular point appears in an audio abundance function relative to equal uniformly distributed numerical values before and after the time point at a specific time point, the frequency spectrum expression of the singular point appears on a FOREGROUND graph; therefore, the technology can generate a visual image of sound frequency and detect abnormal points for the examination of experts or chicken farm staff;
and (3) analysis algorithm: after the decomposition algorithm converts and presents the audio data, the expert distinguishes and marks the sound characteristics contained in the audio data according to the audio data and the three images, the sound characteristics comprise coughing, snorting, sneezing and vocalizing sounds, local label files are formed after marking, the local label files are synchronized to the cloud to form cloud label files for training the robust neural network, the robust neural network preliminarily forms parameterized characteristics with identification capability, and then the parameterized characteristics with stable identification capability of the robust neural network are optimized through the algorithm in step three.
The reinforced neural network part introduces a CNN structure which has strong feature extraction capability, and mainly adopts a convolution layer, a pooling layer, an activation function layer, a full-link layer and an output layer. For the convolutional layer, the following function is shown:
whereinIs the weight of the l-th layer,is the bias of the l-th layer.Is the input tile for the (i, j) position of the l-th layer. The weights of the feature maps are shared (weight sharing reduces the number of parameters). One of the biggest advantages of the CNN structure is the weight sharing mechanism, because it can significantly reduce the computational complexity when the number of parameters is smaller, and also make training and optimization simpler. Pooling layers are an important module in CNN architecture, and the most important goal of such layers is to reduce the size of the feature map by fusing sub-regions using some function, such as averaging or max/min:
wherein y isi,j,kIs the output of the pooling operator at position (i, j) of the kth feature map, ai,j,kIs the pooling region R in the kth feature mapijCharacteristic value at the middle position (m, n). It is particularly noted that this formula corresponds to an average pooling when p is 1, and becomes a maximum pooling as p approaches infinity. Commonly used activation functions are the ReLU activation function, the leakyreu function, which is defined as follows:
ai,j,k=max(zi,j,k,0)
to avoid the gradient vanishing problem, we can choose to use the LeakyReLU function:
wherein λ is in the range of (0, 1). Leaky ReLU does not force the negative part to be zero, but instead it allows for a small non-zero gradient. For a fully connected layer, the following function is expressed:
after the classification output is obtained, a Softmax loss function can be calculated between the predicted value and the true value of the output:
and in this way optimize our entire CNN network.
Further, the system provides a corresponding interface to support other systems to call data, provides a sensor interface, and other systems can control the sensors according to the sensor interface, specifically can control the number of the sensors, and modifies the settings of the sensors, such as the maximum frequency (8192Hz), the minimum frequency (0Hz), and the like of the sound signals collected by the sensors. Other systems can call the sensor interface to obtain the parameter setting of the sensor, so that other systems can know the state of the sensor conveniently; the module also provides calling of the reinforced neural network for other systems, specifically, other systems can set certain parameters of the reinforced neural network through the interface and can also obtain the state of the reinforced neural network so that other systems can know the state of the reinforced neural network conveniently. For example, the confidence level of the augmented neural network can be set, and the recognition of the augmented neural network can be quantitatively accurate by setting different confidence values. The module also provides an acquisition and control interface of cloud data for other systems. Other systems can call the interface to acquire the sound signal data stored on the cloud server, or modify the sound signal data, the tag data and the like stored on the cloud service area, and can also upload the sound data collected by the sensor through the interface and upload the tag data marked manually.
While the invention has been described with respect to the preferred embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (3)
1. A large-scale feeding type broiler respiratory disease sound data analysis system which is characterized in that,
the method comprises the following steps: the method comprises the steps that sound data are collected, a plurality of sound collection sensors are arranged in a coop, the sound collection sensors transmit the collected sound data to a cloud end through a network, and the sound data are converted into audio data in an OPUS format at the cloud end to be stored;
step two: analyzing the sound data, wherein the sound data is transmitted from the cloud to an enhanced neural network of a cloud server for analysis, and an analysis result is uploaded to a cloud backup;
step three: performing algorithm optimization, namely transmitting the abnormal sound data obtained by analysis in the step two to a local server from a cloud end, performing expert marking on the local server to generate a local label file, and additionally updating the local label file to a cloud end label file in the cloud end server to optimize the sound data analysis result in the step two;
step four: constructing a positioning model, measuring the direction of a sound source area, setting the relative position between a sound collecting sensor and the broiler flock, and fixing a sound receiving model to position the specific position of the broiler chicken which emits abnormal sound;
step five: and (4) collecting and analyzing data for a long time, optimizing the algorithm in the third step, and transmitting the analysis result in the second step to a local server to realize early warning and warning of the onset of disease.
2. The system for analyzing the sound data of the respiratory diseases of the large-scale raised broiler chicken according to claim 1, wherein the reinforced neural network used for the sound data analysis comprises:
and (3) audio decomposition algorithm: carrying out Fourier transform on the audio signal in the time domain, and obtaining the frequency spectrum representation of the audio signal in frequency and energy abundance indexes in a minimum time window;
generating three graphs of the analyzed audio data of each segment, wherein the spectral representation of the entire audio data is presented on the "FULL speech" graph; at a specific frequency, the audio frequency moves forward along a time axis, and the frequency spectrum with uniform audio abundance is completely presented on a BACKGROUND graph; moving forward along a time axis at a specific frequency, and when a singular point appears in an audio abundance function relative to equal uniformly distributed numerical values before and after the time point at a specific time point, the frequency spectrum expression of the singular point appears on a FOREGROUND graph; therefore, the technology can generate a visual image of sound frequency and detect abnormal points for the examination of experts or chicken farm staff;
and (3) analysis algorithm: after the decomposition algorithm converts and presents the audio data, the expert distinguishes and marks the sound characteristics contained in the audio data according to the audio data and the three images, the sound characteristics comprise coughing, snorting, sneezing and vocalizing sounds, local label files are formed after marking, the local label files are synchronized to the cloud to form cloud label files for training the robust neural network, the robust neural network preliminarily forms parameterized characteristics with identification capability, and then the parameterized characteristics with stable identification capability of the robust neural network are optimized through the algorithm in step three.
3. The system for analyzing the sound data of the respiratory diseases of the large-scale raised broiler chicken according to claim 1, wherein a filter is installed in the sound collection sensor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011309454.9A CN112562739A (en) | 2020-11-20 | 2020-11-20 | Large-scale feeding type broiler respiratory disease sound data analysis system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011309454.9A CN112562739A (en) | 2020-11-20 | 2020-11-20 | Large-scale feeding type broiler respiratory disease sound data analysis system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112562739A true CN112562739A (en) | 2021-03-26 |
Family
ID=75044239
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011309454.9A Pending CN112562739A (en) | 2020-11-20 | 2020-11-20 | Large-scale feeding type broiler respiratory disease sound data analysis system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112562739A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113456055A (en) * | 2021-07-05 | 2021-10-01 | 自牧机器人(青岛)有限公司 | Poultry respiratory tract real-time monitoring system based on artificial intelligence |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109009129A (en) * | 2018-08-20 | 2018-12-18 | 南京农业大学 | Sow respiratory disease method for early warning based on acoustic analysis |
CN109493874A (en) * | 2018-11-23 | 2019-03-19 | 东北农业大学 | A kind of live pig cough sound recognition methods based on convolutional neural networks |
CN110189756A (en) * | 2019-06-28 | 2019-08-30 | 北京派克盛宏电子科技有限公司 | It is a kind of for monitoring the method and system of live pig abnormal sound |
US20200005766A1 (en) * | 2019-08-15 | 2020-01-02 | Lg Electronics Inc. | Deeplearning method for voice recognition model and voice recognition device based on artificial neural network |
CN110782905A (en) * | 2019-11-05 | 2020-02-11 | 秒针信息技术有限公司 | Positioning method, device and system |
-
2020
- 2020-11-20 CN CN202011309454.9A patent/CN112562739A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109009129A (en) * | 2018-08-20 | 2018-12-18 | 南京农业大学 | Sow respiratory disease method for early warning based on acoustic analysis |
CN109493874A (en) * | 2018-11-23 | 2019-03-19 | 东北农业大学 | A kind of live pig cough sound recognition methods based on convolutional neural networks |
CN110189756A (en) * | 2019-06-28 | 2019-08-30 | 北京派克盛宏电子科技有限公司 | It is a kind of for monitoring the method and system of live pig abnormal sound |
US20200005766A1 (en) * | 2019-08-15 | 2020-01-02 | Lg Electronics Inc. | Deeplearning method for voice recognition model and voice recognition device based on artificial neural network |
CN110782905A (en) * | 2019-11-05 | 2020-02-11 | 秒针信息技术有限公司 | Positioning method, device and system |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113456055A (en) * | 2021-07-05 | 2021-10-01 | 自牧机器人(青岛)有限公司 | Poultry respiratory tract real-time monitoring system based on artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109243470B (en) | Broiler cough monitoring method based on audio technology | |
CN107247971B (en) | Intelligent analysis method and system for ultrasonic thyroid nodule risk index | |
CN110580916B (en) | Weight acoustic measurement model creation method and weight measurement method and device | |
CN106847293A (en) | Facility cultivation sheep stress behavior acoustical signal monitoring method | |
CN112164408A (en) | Pig coughing sound monitoring and early warning system based on deep learning | |
CN108281177B (en) | Internet of things intensive care system | |
CN114596448A (en) | Meat duck health management method and management system thereof | |
CN112331231B (en) | Broiler feed intake detection system based on audio technology | |
CN112544503B (en) | Monitoring and early warning system and method for intelligent beehive | |
NO20210919A1 (en) | Systems and methods for predicting growth of a population of organisms | |
CN109063589A (en) | Instrument and equipment on-line monitoring method neural network based and system | |
CN112562739A (en) | Large-scale feeding type broiler respiratory disease sound data analysis system | |
CN108345857A (en) | A kind of region crowd density prediction technique and device based on deep learning | |
CN117994650A (en) | Intelligent agricultural management system based on big data | |
CN113989538A (en) | Depth image-based chicken flock uniformity estimation method, device, system and medium | |
CN109828623B (en) | Production management method and device for greenhouse crop context awareness | |
CN117253192A (en) | Intelligent system and method for silkworm breeding | |
CH719673A2 (en) | AI-BASED REAL-TIME ACOUSTIC WILDLIFE MONITORING SYSTEM | |
CN109242219B (en) | Prediction method and prediction device for layer feeding behavior | |
CN116723614B (en) | AI-based adaptive fish propagation environment illumination control method | |
CN118173102B (en) | Bird voiceprint recognition method in complex scene | |
CN118248337B (en) | AI-based health condition monitoring system for old people | |
CN118097391A (en) | Multi-mode fusion fish swarm ingestion intensity classification method, system, equipment and medium | |
CN117237820B (en) | Method and device for determining pest hazard degree, electronic equipment and storage medium | |
CN116421152B (en) | Sleep stage result determining method, device, equipment and medium |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210326 |