WO2020006934A1 - 猪只疾病识别方法、装置及计算机可读存储介质 - Google Patents

猪只疾病识别方法、装置及计算机可读存储介质 Download PDF

Info

Publication number
WO2020006934A1
WO2020006934A1 PCT/CN2018/111654 CN2018111654W WO2020006934A1 WO 2020006934 A1 WO2020006934 A1 WO 2020006934A1 CN 2018111654 W CN2018111654 W CN 2018111654W WO 2020006934 A1 WO2020006934 A1 WO 2020006934A1
Authority
WO
WIPO (PCT)
Prior art keywords
pig
disease
cough
sample data
data
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.)
Ceased
Application number
PCT/CN2018/111654
Other languages
English (en)
French (fr)
Inventor
王健宗
蔡元哲
程宁
肖京
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2020006934A1 publication Critical patent/WO2020006934A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
    • A01K—ANIMAL HUSBANDRY; AVICULTURE; APICULTURE; PISCICULTURE; FISHING; REARING OR BREEDING ANIMALS, NOT OTHERWISE PROVIDED FOR; NEW BREEDS OF ANIMALS
    • A01K29/00—Other apparatus for animal husbandry
    • A01K29/005—Monitoring or measuring activity
    • 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
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/21—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being power information
    • 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
    • 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
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems

Definitions

  • the present application relates to the technical field of animal disease identification, and in particular, to a method, a device, and a computer non-volatile readable storage medium for identifying pig diseases.
  • Livestock and poultry production is an important source of the human food supply chain.
  • monitoring the health of the pigs in the farm is critical to the sustainable management of the pigs.
  • real-time monitoring of livestock and poultry is possible.
  • Cough as a mechanism to prevent respiratory infections in pigs, may be a symptom of respiratory disturbances or infections.
  • the sound of pigs cough is directly related to the respiratory diseases of pigs. Pigs are usually set in pig farms. The sound monitoring system is used to monitor the cough sound of the pigs, and then to alert the pigs according to the size of the cough sounds of the pigs. Experienced veterinarians can diagnose diseases based on the cough sounds to prevent greater losses on the farm.
  • the prior art pig monitoring system can only provide early warning of the disease, and cannot judge the type of pig disease, so that the breeder cannot determine the type of pig disease at the early stage of the disease, even when the pig is coughed.
  • the veterinarian also needs to check one by one and judge the types of pig diseases one by one, which makes the process of identifying pig diseases cumbersome and results in slower identification of pig diseases.
  • the embodiments of the present application provide a method and device for identifying pig diseases, and a computer non-volatile readable storage medium, which solves the problem of tedious process of identifying pig diseases in the related art, which leads to a slow identification of pig diseases.
  • a method for identifying pig diseases includes:
  • each pig cough sample data carrying a corresponding pig disease type label
  • the cough data of the pig to be identified is input to the pig disease recognition model, and the pig disease type corresponding to the cough data of the pig to be identified is identified.
  • a pig disease identification device includes:
  • An obtaining unit configured to obtain multiple pig cough sample data, and each pig cough sample data carries a corresponding pig disease type label;
  • the recognition unit is configured to input the cough data of the pig to be identified into the pig disease recognition model, and identify a pig disease type corresponding to the cough data of the pig to be identified.
  • a computer non-volatile readable storage medium in which computer readable instructions are stored, and the program implements the following steps when executed by a processor:
  • each pig cough sample data carrying a corresponding pig disease type label
  • the cough data of the pig to be identified is input to the pig disease recognition model, and the pig disease type corresponding to the cough data of the pig to be identified is identified.
  • a computer device including a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor.
  • the processor implements the following steps when the program is executed. :
  • each pig cough sample data carrying a corresponding pig disease type label
  • the cough data of the pig to be identified is input to the pig disease recognition model, and the pig disease type corresponding to the cough data of the pig to be identified is identified.
  • a pig cough sample is input to a recurrent neural network model for training to construct a pig disease recognition model.
  • the pig disease recognition model records the mapping relationship between pig cough data and pig disease types.
  • the identification model can accurately identify pig diseases.
  • the embodiment of the present application recognizes pig diseases by constructing a pig identification model. When abnormalities in the audio data of the pigs are monitored, the veterinarian inspects them one by one and judges the types of pig diseases one by one, which saves manpower time. Colleagues facilitate the pig breeding sites to quickly deploy pig treatments and corresponding treatment plans, reducing the economics caused by the diseases loss.
  • FIG. 1 is a flowchart of a method for identifying pig diseases according to an embodiment of the present application
  • FIG. 2 is a flowchart of another method for identifying pig diseases according to an embodiment of the present application.
  • 3A is a block diagram of a pig disease recognition model training process according to an embodiment of the present application.
  • 3B is a block diagram of a pig disease recognition model recognition process according to an embodiment of the present application.
  • FIG. 4 is a flowchart of another method for identifying pig diseases according to an embodiment of the present application.
  • FIG. 5 is a structural block diagram of a pig disease identification device according to an embodiment of the present application.
  • FIG. 6 is a structural block diagram of another pig disease identification device according to an embodiment of the present application.
  • FIG. 7 is a structural block diagram of another pig disease recognition device according to an embodiment of the present application.
  • FIG. 8 is a block diagram of a pig disease identification device according to an embodiment of the present application.
  • FIG. 1 is a first flowchart according to an embodiment of the present application. As shown in FIG. 1, the process includes the following steps:
  • Step S101 obtaining a plurality of pig cough sample data, and each pig cough sample data carries a corresponding pig disease type label;
  • the pig cough sample data is audio data collected from the selected pig samples, and the pig cough samples carry the types of pig diseases that need to be identified, such as pneumonia, infectious atrophic rhinitis, pleural pneumonia, asthma Disease, pulmonary filariasis, or other diseases. Samples of different types of pigs are selected according to actual needs. Of course, samples of healthy pigs are also included to facilitate the diagnosis of pigs with unknown pig diseases. Identify.
  • the number of selected pig samples is not limited in the embodiments of the present application. Generally, the same number of pig samples can be selected for each disease type. For example, five pig samples are selected for each disease type. Of course, the number of pig samples can also be selected according to the disease symptoms of different types of pig diseases at various stages of the disease. For example, in order to ensure that the collected pig samples cover enough disease types and enough disease symptoms, it is a disease for each stage. Select a large number of pig samples for the pig disease types with obvious symptoms or more disease symptoms, and select a small number of pig samples for other pig disease types.
  • the pig cough sample data can be obtained by installing a collection device on the pig sample, and the collection device can also be installed in the pig breeding site.
  • the embodiment of the present application is not limited.
  • wearable sensors are installed on pigs to obtain the pig's audio data in real time.
  • Step S102 inputting a plurality of pig cough sample data into a recurrent neural network model for training to construct a pig disease recognition model, and the pig disease recognition model records a mapping relationship between pig cough data and pig disease types;
  • the pig cough sample data includes cough data of different pig disease types
  • the pig cough sample data needs to be labeled.
  • pigs carrying different disease types are mainly used. Only the cough sample data is labeled, and of course, the cough sample data of pigs carrying different disease stages of the same disease type can be labeled, which is not limited in the embodiment of the present application.
  • the recurrent neural network model here is a network structure that can train pig cough sample data and pig disease type mapping.
  • the network structure is equivalent to the pig disease recognition model.
  • Pig cough data identifies types of pig disease.
  • the structure of a specific recurrent neural network model can be implemented by a 6-layer coding-decoding structure.
  • other recurrent neural network model structures can also be selected according to the actual characteristics of the data characteristics of training pig cough sample data, which is not limited in the examples of this application.
  • step S103 the cough data of the pig to be identified is input to a pig disease recognition model, and the pig disease type corresponding to the cough data of the pig to be identified is identified.
  • the pig cough data to be identified is the collected audio data of unknown disease or unknown pig types, and the pig cough data to be identified is the same as the pig cough sample data collection process.
  • the pig disease recognition model output classification is set to pig disease stage classification, the pig disease recognition model can also identify the stage of the pig disease.
  • the application examples are not limited.
  • the output of the pig disease recognition model is set to 7 types of pig diseases, and the disease stage of each disease type is not set. Of course, other will be set to indicate the health of the pig, and the cough data of the pig to be identified will be passed through the pig.
  • the disease-only recognition model can output pig disease types, and output other if the pigs are healthy.
  • a pig cough sample is input to a recurrent neural network model for training to construct a pig disease recognition model.
  • the pig disease recognition model records the mapping relationship between pig cough data and pig disease types.
  • the identification model can accurately identify pig diseases.
  • the embodiment of the present application recognizes pig diseases by constructing a pig identification model. When abnormalities in the audio data of the pigs are monitored, the veterinarian inspects them one by one and judges the types of pig diseases one by one, which saves manpower time. Colleagues facilitate the pig breeding sites to quickly deploy pig treatments and corresponding treatment plans, reducing the economics caused by the diseases. loss.
  • FIG. 2 is a flowchart of a method for identifying pig diseases according to a preferred embodiment of the present application. As shown in FIG. 2, the method includes the following steps:
  • Step S201 Acquire audio sample data of pigs of different pig disease types within a preset time period.
  • each pig has an independent activity space to ensure that the audio samples of the pigs are not interfered by the sound of other pigs.
  • the pig's audio here also includes the pig's bark, the pig's breathing, the pig's heartbeat, and the pig's crackling sound.
  • step S202 the audio sample data of the pig is labeled according to the type of the pig disease carried on each pig sample to obtain multiple pieces of audio sample data of the pig that carry the label of the type of the pig disease.
  • the specific way of labeling the audio sample data of each pig can be manually labeled on the PC using a sound editor. For example, if the disease type of the pig sample is asthma, then the asthma is marked on the audio sample data of the pig. The labeling of other pig audio sample data is the same.
  • the audio sample data of pigs is labeled according to the type of pig disease carried on each pig sample, and the audio sample data of pigs of different disease types are classified in advance to determine the sample data input mode.
  • the classified pig sample data is input into the neural network structure as a positive sample, which is convenient for subsequent training of the pig audio sample data.
  • Step S203 Perform frame processing on the pig audio sample data within a preset time period to obtain a plurality of frames of the pig audio sample sequence with a preset duration.
  • a random Gaussian distribution duration period can be generated from the audio sample data of the pig in advance.
  • the audio sample data is quantized so that the average duration of all sounds is fixed.
  • the cough duration in all the collected audio data of the pigs is from 250 milliseconds to a maximum of 800 milliseconds.
  • the draw duration of all the audio data of the pigs is 320 milliseconds.
  • step S204 the pig audio sample sequence is filtered, and the pig cough sample data is retained.
  • the pig audio sample sequence includes pig cough sounds, as well as other pig cough sounds, such as pig breathing sounds, pig heartbeat sounds, and other irrelevant audio data.
  • the pig audio sample sequences need to be filtered. Only the pig cough sample data can be retained to extract the pig cough audio data.
  • Step S205 The pig cough sample data is input to a recurrent neural network model for training, and a pig disease recognition model is constructed.
  • the pig disease recognition model records a mapping relationship between pig cough data and pig disease types.
  • the recurrent neural network may have a multi-layer structure, each layer structure has different input and output parameters and implement different functions.
  • the recurrent neural network is used to iterate the cough sample data of pigs carrying different disease type labels. After training, a mapping relationship between pig cough sample data and different types of pig diseases is obtained, and the mapping relationship is equivalent to a pig disease recognition model.
  • a recurrent neural network is composed of a 5-layer structure, and the first layer is a coding layer structure.
  • the coding layer is composed of 3 layers of encoders, including 128 neurons, 2 bidirectional recurrent layers of 64 neurons, and 32 recurrent neurons.
  • the unidirectional layer of the encoder is set to handle any sequence with a maximum length of the set value, and all recurrent neurons in the encoder can adopt the GRU structure.
  • the GRU structure is simple.
  • the second layer structure is a fixed coding layer, fixed
  • the coding layer can include a 32-element activation layer with fixed parameters. Through the fixed coding layer, the characteristic parameters of each disease stage are summarized, and the feature matrix of each disease stage of different pig disease types is output.
  • the third layer structure is a decoding layer.
  • the decoding layer can include a separate loop layer with 64 long and short memory LSTM units, combined with the attention mechanism, so that The network mainly focuses on the significant part of the input features to improve the classification performance.
  • the pig cough sample data is used to optimize the feature parameters in the feature matrix of each disease stage of different pig disease types, and output different pigs.
  • the feature representation maps to the sample tag space, learns the overall features in combination, and uses the fully connected layer to fuse the pig cough sample data at the feature matrix of each disease stage of different pig disease types, and outputs features carrying different pig disease type labels.
  • Matrix; the fifth layer is a classification layer.
  • the classification layer can use softmax output classification labels.
  • the softmax function can map the input to a value of (0,1). This value can be understood as a probability. The maximum probability is used as the classification result.
  • Classification layer pairs of characteristic moments carrying different pig disease labels Classification, to build pig disease recognition model.
  • the training process of a specific pig disease recognition model can be shown in Figure 3A.
  • the pig cough sample data is classified, labeled, and preprocessed to obtain a pig cough sample sequence, which carries different diseases.
  • the labeled pig cough sample sequence is trained to obtain a pig disease recognition model.
  • step S206 the cough data of the pig to be identified is input to a pig disease recognition model, and the pig disease type corresponding to the cough data of the pig to be identified is identified.
  • the disease can be selected within a preset time period. Pig cough data was compared to determine pig disease types.
  • the recognition process of the specific pig disease recognition model can be shown in Figure 3B. After the cough data of the pigs to be identified is input, the pig cough data is pre-processed to obtain the pig cough sample sequence, and the pig cough sample sequence is input to the treating doctor. Identify the model and classify the type of pig disease.
  • the pig disease recognition model in the embodiment of the present application is equivalent to using a sequence-to-sequence recurrent neural network to map an undivided input sequence to another output label.
  • the recurrent nerve can capture the dependency between time and spectrum, and perform the task of identifying the type of disease corresponding to the cough data according to the characteristic parameters of the cough data in each disease stage, so that the output of each layer will depend on the calculation of the previous layer, thereby improving pigs. Only disease recognition accuracy.
  • step S207 pig disease information is established according to the identified disease type, and the pig disease information records the pig information, the identified pig disease type, and the disease stage at which the pig disease type is determined.
  • the pig information may include, but is not limited to, the breeding place where the pig is located.
  • the pig code and other information used to identify the identity of the pig.
  • the pig breeding place is used to isolate the pigs.
  • Each pig pen has identification information, and each pig is recorded with a code to distinguish it from other pigs.
  • the information of the diseased pigs is organized by establishing the pig disease information to facilitate the subsequent treatment of the pig diseases.
  • step S208 different treatment strategies are set for the pigs according to the disease stage in which the pig disease type is in the pig disease information.
  • pigs at different disease stages can be set.
  • Treatment strategy For example, pigs at the beginning of the disease can be treated by adding therapeutic drugs to the pig's feed, or simply observed. Pigs at the later stage of the disease only need to notify the veterinarian for emergency treatment or need some emergency measures.
  • the pig information carries pig identification information
  • Institutional information of different pig breeding institutions and a preset database of pig identification information in different pig breeding institutions Based on the pig identification information, look up the information of the institution responsible for raising pigs, and the pig disease information
  • different treatment strategies are sent to the institutions responsible for raising pigs to facilitate early treatment of diseased pigs.
  • a pig cough sample is input to a recurrent neural network model for training to construct a pig disease recognition model.
  • the pig disease recognition model records a mapping relationship between pig cough data and pig disease types. Only the disease recognition model can accurately identify pig diseases.
  • the embodiment of the present application recognizes pig diseases by constructing a pig identification model. Without the need to monitor the abnormal audio data of the pigs, the veterinarians check each one and judge the types of pig diseases one by one, saving manpower time. Colleagues can facilitate the deployment of pig treatments and corresponding treatment schemes at pig breeding sites to reduce the incidence of diseases. Economic loss.
  • FIG. 4 is a flowchart of a method for identifying pig diseases according to a preferred embodiment of the present application. As shown in FIG. 4, the method includes the following steps:
  • Step S301 Acquire audio sample data of pigs of different pig disease types within a preset time period.
  • a specific implementation method for specifically collecting audio data of the pig is the same as step S201, and details are not described herein.
  • a wearable device is designed in the embodiment of the present application to acquire pig audio sample data in real time, which includes audio data such as pig breathing sounds, cries, and cough sounds, and can be Transfer the collected data to the cloud in real time.
  • the device consists of a piezoelectric sensor and additional components.
  • the signal conditioning circuit enables the sensor to amplify breathing sounds and completely eliminate ambient sounds.
  • the device can be fixed to the pig's ears on.
  • Step S302 labeling the audio sample data of the pig according to the different disease stages of the pig disease type carried on each pig sample to obtain multiple audio samples of the pigs carrying the tags of the different disease stages of the pig disease type. data.
  • step S202 the labeling method of the audio sample data of each pig is the same as that of step S202, and will not be described again here.
  • pig samples in order to accurately train the effects of different disease types, can also be selected as training samples for different disease stages in which the disease types are located, and according to different diseases of the pig disease types carried in the pig samples, In this stage, the pig audio sample data is labeled, and the pig audio sample data of different disease stages are classified in advance to determine the sample data input mode. The classified pig sample data is input into the neural network structure as a positive sample. Facilitate subsequent training of pig audio sample data.
  • Step S303 Perform frame processing on the pig audio sample data within a preset period of time to obtain a plurality of frames of the pig audio sample sequence with a preset duration.
  • the pig audio sample data is not a stable audio signal.
  • the pig audio sample data is framed to obtain a multi-frame sequence of pig audio samples with a preset duration, and then each frame of the pig audio is The sample sequence is treated as a stable audio signal to facilitate subsequent processing of pig cough data.
  • step S304 the signal energy of the audio sample sequence of the pig is detected through the energy window.
  • different pig audio sample sequences have different signal energies.
  • the signal energy of each frame of the pig audio sample sequences is detected through an energy window to determine the audio characteristics of the pig audio sample sequences, and further The pig audio sample sequence is processed.
  • step S305 the audio sample sequence of the pig which is lower than the preset signal energy is filtered, and the remaining audio sample sequence of the pig obtained by the filtering is used as the pig cough sample data.
  • the preset signal energy is usually set to be low , Filtering the pig audio sample sequence below the preset signal energy to remove the interfering pig audio data, and filtering the remaining pig audio sample sequence as the pig cough sample data.
  • Step S306 the pig cough sample data is input into a recurrent neural network model for training, and a pig disease recognition model is constructed.
  • the pig disease recognition model records a mapping of the pig cough data and different disease stages of the pig disease type. relationship.
  • the recurrent neural network is used to repeatedly train the audio sample data of the pigs carrying the tags of different disease stages of the pig's disease type to obtain a mapping of the pig's cough data and the different stages of the disease's disease type This mapping relationship is equivalent to a pig disease recognition model.
  • step S205 the specific implementation of training the pig cough sample data here is the same as that of step S205, and will not be described again here.
  • step S307 the cough data of the pig to be identified is input to the pig disease recognition model, and the pig disease type corresponding to the pig cough data to be identified and the disease stage of the pig disease type are identified.
  • a pig cough sample is input to a recurrent neural network model for training, and a pig disease recognition model is constructed.
  • the pig disease recognition model also records a mapping of pig cough data and different disease stages in the type of pig disease Relationship, the pig disease identification model can accurately identify the pig disease type corresponding to the pig cough data and the disease stage of the pig disease type, there is no need to monitor the pig audio data when abnormal, the veterinarian one by one Inspection and individual determination of the disease type of the pig and the stage of the disease, saving labor time. Colleagues facilitate the rapid deployment of pig treatment and corresponding treatment plans in pig breeding sites, and reduce economic losses due to disease.
  • FIG. 5 is a structural block diagram of a pig disease identification device according to an embodiment of the present application.
  • the device includes an obtaining unit 41, a constructing unit 42, and a recognition unit 43.
  • the obtaining unit 41 may be used to obtain multiple pig cough sample data, and each pig cough sample data carries a corresponding pig disease type label;
  • a constructing unit 42 may be used to input a plurality of pig cough sample data into a recurrent neural network model for training to construct a pig disease recognition model.
  • the pig disease recognition model records pig cough data and pig disease types. Mapping relations;
  • the identifying unit 43 may be configured to input the cough data of the pig to be identified into a pig disease recognition model, and identify a pig disease type corresponding to the cough data of the pig to be identified.
  • a pig cough sample is input to a recurrent neural network model for training to construct a pig disease recognition model.
  • the pig disease recognition model records the mapping relationship between pig cough data and pig disease types.
  • the identification model can accurately identify pig diseases.
  • the embodiment of the present application recognizes pig diseases by constructing a pig identification model. When abnormalities in the audio data of the pigs are monitored, the veterinarian inspects them one by one and judges the types of pig diseases one by one, which saves manpower time. Colleagues facilitate the pig breeding sites to quickly deploy pig treatments and corresponding treatment plans, reducing the economics caused by the diseases loss.
  • FIG. 6 is a schematic structural diagram of another pig disease identification device according to an embodiment of the present application. As shown in FIG. 6, the device further includes:
  • the establishing unit 44 may be configured to input pig cough data to be identified into a pig disease recognition model, and identify pig disease types according to the identified disease types.
  • the pig disease information is recorded in the pig disease information. Have pig information, identified pig disease types, and the stage of the disease to determine the type of pig disease;
  • the setting unit 45 can be used to set different treatment strategies for the pigs according to the disease stage of the pig disease type in the pig disease information.
  • the pig information carries pig identification information
  • the device further includes:
  • the searching unit 46 may be configured to search for a pig responsible for raising pigs from a preset database according to the pig identification information after the pigs are initially treated with different treatment strategies in the disease stage of the pig disease type in the pig disease information.
  • Institutional information the preset database contains institutional information of different pig breeding institutions and pig identification information in different pig breeding institutions;
  • the sending unit 47 may be configured to send the pig disease information and the different treatment strategies to the institution responsible for raising pigs.
  • the obtaining unit 41 includes:
  • the obtaining module 411 may be used for obtaining audio sample data of pigs of different types of pig diseases within a preset time period;
  • the frame framing module 412 may be used to perform frame processing on the pig audio sample data within a preset time period to obtain a plurality of frames of the pig audio sample sequence with a preset duration;
  • the filtering module 414 may be configured to filter a plurality of pig audio sample sequences of preset durations, and use the plurality of pig audio sample sequences obtained after filtering as a plurality of pig cough sample data.
  • the obtaining unit 41 further includes:
  • the first labeling module 414 may be configured to label the audio sample data of the pig according to the type of the pig disease carried on each pig sample to obtain multiple pieces of audio sample data of the pig that carry the label of the type of the pig disease;
  • the recurrent neural network is a multi-layered network model
  • the construction unit 32 includes:
  • the extraction module 421 can be used for the first layer structure of the recurrent neural network to extract the characteristic parameters of the pig cough sample data at each disease stage of different pig disease types;
  • the summary module 422 can be used for the second layer structure of the recurrent neural network to summarize the characteristic parameters of each disease stage, and output the feature matrix of each disease stage of different pig disease types;
  • the optimization module 423 can be used in the third layer structure of the recurrent neural network to optimize the characteristic parameters of the pig cough sample data in the characteristic matrix of each disease stage of different pig disease types according to different pig disease type labels, and output different pigs. Mapping relationship between disease type labels and feature matrix of pig cough sample data at different disease stages of different pig disease types;
  • the fusion module 424 can be used in the fourth layer structure of the recurrent neural network to fuse the feature matrix of pig cough sample data at each disease stage of different pig disease types, and output the feature matrix carrying labels of different pig disease types;
  • the classification module 425 can be used for the fifth layer structure of the recurrent neural network to classify the feature matrix carrying different types of pig disease type labels, and to construct a pig disease recognition model.
  • FIG. 7 is a schematic structural diagram of another pig disease identification device according to an embodiment of the present application.
  • the cough sample data of each pig is also It carries the corresponding disease stage tags of the corresponding pig disease types.
  • the pig disease recognition model also records the mapping relationship between the pig cough data and the different disease stages of the pig disease types.
  • the obtaining unit 41 further includes:
  • the second labeling module 415 can be used to label the audio sample data of the pig according to the different disease stages of the pig disease type carried on each pig sample to obtain the tags carrying the different disease stages of the pig disease type. Pig audio sample data.
  • the identification unit 43 may be further configured to input the cough data of the pig to be identified into the pig disease recognition model, and identify the pig disease type corresponding to the cough data of the pig to be identified and the type of the pig disease type. Disease stage.
  • the filtering module 413 may be specifically configured to detect the signal energy of the pig audio sample sequence through the energy window;
  • the filtering module 413 may also be specifically configured to filter a sequence of audio samples of pigs whose energy is lower than a preset signal, and use the remaining sequence of audio samples of pigs obtained after filtering as sample data of pig cough.
  • Fig. 8 is a block diagram of a pig disease recognition device according to an exemplary embodiment.
  • the device 500 may be a mobile phone, a computer, a digital broadcasting terminal, a messaging device, a game console, a tablet device, a medical device, a fitness equipment, a personal digital assistant, and the like.
  • the device 500 may include one or more of the following components: a processing component 502, a memory 504, a power component 506, a multimedia component 508, an audio component 510, an I / O (Input / Output) interface 512, A sensor component 514, and a communication component 516.
  • the processing component 502 generally controls overall operations of the device 500, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations.
  • the processing component 502 may include one or more processors 520 to execute instructions to complete all or part of the steps of the method described above.
  • the processing component 502 may include one or more modules to facilitate the interaction between the processing component 502 and other components.
  • the processing component 502 may include a multimedia module to facilitate the interaction between the multimedia component 508 and the processing component 502.
  • the memory 504 is configured to store various types of data to support operation at the device 500. Examples of such data include instructions for any application or method operating on the device 500, contact data, phone book data, messages, pictures, videos, and the like.
  • the memory 504 may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as SRAM (Static Random Access Memory, Static Random Access Memory), EEPROM (Electrically-Erasable Programmable Read-Only Memory, Electrical Erasable Programmable Read Only Memory (EPROM), EPROM (Erasable Programmable Read Only Memory), PROM (Programmable Read-Only Memory, Programmable Read Only Memory), ROM (Read-Only Memory, Read-only memory), magnetic memory, flash memory, magnetic or optical disks.
  • SRAM Static Random Access Memory, Static Random Access Memory
  • EEPROM Electrical Erasable Programmable Read Only Memory
  • EPROM Electrical Erasable Programmable Read Only Memory
  • PROM Programmable Read-Only Memory, Programmable Read Only Memory
  • the power supply component 506 provides power to various components of the device 500.
  • the power component 506 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the device 500.
  • the multimedia component 508 includes a screen that provides an output interface between the device 500 and a user.
  • the screen may include an LCD (Liquid Crystal Display) and a TP (Touch Panel). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive an input signal from a user.
  • the touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure related to the touch or slide operation.
  • the multimedia component 508 includes a front camera and / or a rear camera. When the device 500 is in an operation mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.
  • the audio component 510 is configured to output and / or input audio signals.
  • the audio component 510 includes a MIC (Microphone, microphone).
  • the microphone is configured to receive an external audio signal.
  • the received audio signal may be further stored in the memory 504 or transmitted via the communication component 516.
  • the audio component 510 further includes a speaker for outputting audio signals.
  • the I / O interface 512 provides an interface between the processing component 502 and a peripheral interface module.
  • the peripheral interface module may be a keyboard, a click wheel, a button, or the like. These buttons can include, but are not limited to: a home button, a volume button, a start button, and a lock button.
  • the sensor component 514 includes one or more sensors for providing status assessment of various aspects of the device 500.
  • the sensor component 514 can detect the on / off state of the device 500, and the relative positioning of the components, such as the display and keypad of the device 400.
  • the sensor component 514 can also detect the change in the position of the device 500 or a component of the device 500. The presence or absence of contact with the device 500, the orientation or acceleration / deceleration of the device 500, and the temperature change of the device 500.
  • the sensor component 514 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact.
  • the sensor component 514 may further include a light sensor, such as a CMOS (Complementary Metal Oxide Semiconductor) or a CCD (Charge-coupled Device) image sensor, for use in imaging applications.
  • a light sensor such as a CMOS (Complementary Metal Oxide Semiconductor) or a CCD (Charge-coupled Device) image sensor, for use in imaging applications.
  • the sensor component 514 may further include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • the communication component 516 is configured to facilitate wired or wireless communication between the device 500 and other devices.
  • the device 500 may access a wireless network based on a communication standard, such as WiFi, 2G, or 3G, or a combination thereof.
  • the communication component 516 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel.
  • the communication component 516 further includes an NFC (Near Field Communication) module to facilitate short-range communication.
  • the NFC module can be based on RFID (Radio Frequency Identification) technology, IrDA (Infra-red Data Association) technology, UWB (Ultra Wideband) technology, BT (Bluetooth, Bluetooth) technology and Other technologies to achieve.
  • the device 500 may be implemented by one or more ASIC (Application Specific Integrated Circuit), DSP (Digital Signal Processor), DSPD (Digital Signal Processor, Device) Equipment), PLD (Programmable Logic Device), FPGA) (Field Programmable Gate Array), controller, microcontroller, microprocessor or other electronic components to implement the above Identification of pig diseases.
  • ASIC Application Specific Integrated Circuit
  • DSP Digital Signal Processor
  • DSPD Digital Signal Processor, Device
  • PLD Programmable Logic Device
  • FPGA Field Programmable Gate Array
  • controller microcontroller
  • microprocessor or other electronic components to implement the above Identification of pig diseases.
  • a non-transitory computer non-volatile readable storage medium including instructions may be executed by the processor 520 of the apparatus 500 to complete the above method.
  • the non-transitory computer non-volatile storage medium may be ROM, RAM (Random Access Memory, Random Access Memory), CD-ROM (Compact Disc Read-Only Memory), magnetic tape , Floppy disks, and optical data storage devices.
  • a non-transitory computer non-volatile readable storage medium when instructions in the non-volatile readable storage medium are executed by a processor of a pig disease identification device, enable the pig disease identification device to execute the above Identification of pig diseases.
  • modules or steps of the present application can be implemented by general-purpose computer equipment, which can be centralized on a single computer equipment or distributed on a network composed of multiple computer equipment
  • they may be implemented with computer-readable instructions of a computer device, so that they may be stored in a storage device and executed by the computer device, and in some cases, may be in a different order than here
  • the steps shown or described are performed, or they are separately made into individual integrated circuit modules, or multiple modules or steps in them are made into a single integrated circuit module for implementation.
  • this application is not limited to any particular combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Environmental Sciences (AREA)
  • Public Health (AREA)
  • Pathology (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Databases & Information Systems (AREA)
  • Biophysics (AREA)
  • Animal Husbandry (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Data Mining & Analysis (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

一种猪只疾病识别方法、装置及计算机非易失性可读存储介质，涉及动物疾病识别技术领域，可以无需兽医逐个检查，逐个判断猪只疾病类型，简化猪只疾病识别过程，提高了猪只疾病识别速度。所述方法包括：获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签（S101）；将所述多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系（S102）；将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型（S103）。

Description

猪只疾病识别方法、装置及计算机可读存储介质
本申请要求于2018年7月5日提交中国专利局、申请号为2018107292691、申请名称为“猪只疾病识别方法、装置及电子设备”的中国专利申请的优先权，其全部内容通过引用结合在申请中。
技术领域
本申请涉及动物疾病识别技术领域，尤其是涉及到猪只疾病识别方法、装置及计算机非易失性可读存储介质。
背景技术
家畜、家禽生产是人类食物供应链的重要来源之一。为了达到健康养殖，在养殖场所中监控猪只的健康状况对猪只的可持续管理至关重要。随着先进传感器和信息网络技术的不断发展，使得对畜禽实时全程监控成为可能。
咳嗽作为猪只身体的一种防御呼吸道感染的机制，可能是呼吸系统紊乱或感染的征兆，猪只咳嗽声音大小直接与猪只的呼吸道的疾病有关，通常在养猪场中设置有猪只咳嗽声音监控系统，用来监控猪只咳嗽声音，进而根据猪只咳嗽声音大小对猪只疾病进行预警，经验丰富的兽医可以根据咳嗽声音进行疾病诊断，防止猪场造成更大的损失。然而，现有技术的猪只监控系统仅可以对疾病进行预警，无法对猪只疾病类型进行判断，使得饲养人员无法在猪只发病的初期判断出猪只疾病类型，即使在监控到猪只咳嗽声音异常时，也需要兽医逐个检查以及逐个判断猪只疾病类型，使得猪只疾病识别过程繁琐，导致猪只疾病识别速度较慢。
发明内容
本申请实施例提供了一种猪只疾病识别方法、装置及计算机非易失性可读存储介质，解决了相关技术中猪只疾病识别过程繁琐，导致猪只疾病识别速度较慢的问题。
根据本申请实施例的第一方面，提供一种猪只疾病识别方法，所述方法包括：
获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；
将所述多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；
将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型。
根据本申请实施例的第二方面，提供一种猪只疾病识别装置，所述装置包括：
获取单元，用于获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；
构建单元，用于将所述多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；
识别单元，用于将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型。
根据本申请实施例的第三方面，提供一种计算机非易失性可读存储介质，其上存储有计算机可读指令，该程序被处理器执行时实现以下步骤：
获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；
将所述多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；
将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型。
根据本申请的第四个方面，提供一种计算机设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令，所述处理器执行所述程序时实现以下步骤：
获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；
将所述多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；
将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型。
通过本申请，将猪只咳嗽样本输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型记录有猪只咳嗽数据与猪只疾病类型的映射关系，通过猪只疾病识别模型即可准确识别出猪只疾病，与现有技术通过监控猪只音频数据的方式识别猪只疾病的方法相比，本申请实施例通过构建猪只识别模型对猪只疾病进行识别，无需在监控到猪只音频数据发生异常时，兽医逐个检查以及逐个去判断猪只疾病类型，节省人力时间，同事方便猪只饲养场所快速部署猪只治疗以及对应的治疗方案，降低因疾病产生的经济损失。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1是根据本申请实施例的一种猪只疾病识别方法的流程图；
图2是根据本申请实施例的另一种猪只疾病识别方法的流程图；
图3A是根据本申请实施例的猪只疾病识别模型训练过程的框图；
图3B是根据本申请实施例的猪只疾病识别模型识别过程的框图；
图4是根据本申请实施例的另一种猪只疾病识别方法的流程图；
图5是根据本申请实施例的一种猪只疾病识别装置的结构框图；
图6是根据本申请实施例的另一种猪只疾病识别装置的结构框图；
图7是根据本申请实施例的另一种猪只疾病识别装置的结构框图；
图8是根据本申请实施例的猪只疾病识别装置的框图。
具体实施方式
下文中将参考附图并结合实施例来详细说明本申请。需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互组合。
在本实施例中提供了一种猪只疾病识别方法，图1是根据本申请实施例的流程图一，如图1所示，该流程包括如下步骤：
步骤S101，获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；
其中，猪只咳嗽样本数据为从选取的猪只样本身上采集的音频数据，并且猪只咳嗽样本身上携带有所需要识别的猪只疾病类型，如肺疫、传染性萎缩鼻炎、胸膜肺炎、喘病、肺丝虫病或其他疾病等，具体根据实际需求选取不同疾病类型的猪只样本，当然还会包括健康猪只样本，以便于在未知猪只疾病与否的情况下对猪只进行疾病识别。
需要说明的是，本申请实施例对选取猪只样本的数量不进行限定，通常情况下可以每种疾病类型选取相同数量的猪只样本，例如，每种疾病类型的猪只样本选取5只，当然也可以根据不同猪只疾病类型在疾病各个阶段的疾病症状来选取猪只样本数量，例如，为了保证采集到的猪只样本覆盖足够多的疾病类型以及足够多的疾病症状，为各个阶段疾病症状明显或者疾病症状较多的猪只疾病类型选取多量的猪只样本，为其他的猪只疾病类型选取少量的猪只样本。
本申请实施例具体可以通过在猪只样本身上安装采集设备来获取猪只咳嗽样本数据，还可以通过在猪只养殖场所内安装采集设备，本申请实施例不进行限定，为了保证获取到 猪只咳嗽样本数据的准确性以及减少噪声干扰，通常在猪只身上安装可穿戴传感器来实时获取猪只的音频数据。
步骤S102，将多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；
由于猪只咳嗽样本数据中包括不同猪只疾病类型的咳嗽数据，为了对不同疾病类型的猪只样本咳嗽数据进行区分，需要对猪只咳嗽样本数据进行标记，这里主要对携带不同疾病类型的猪只咳嗽样本数据进行标记，当然可以对携带相同疾病类型所处不同疾病阶段的猪只咳嗽样本数据进行标记，本申请实施例不进行限定。
对于本申请实施例，这里的循环神经网络模型为可以训练出猪只咳嗽样本数据与猪只疾病类型映射的网络结构，该网络结构相当于猪只疾病识别模型，通过猪只疾病识别模型可以根据猪只咳嗽数据对猪只疾病类型进行识别。
具体循环神经网络模型的结构可以通过6层的编码-解码结构实现，当然也可以根据实际需要训练猪只咳嗽样本数据的数据特征选择其他循环神经网络模型的结构，本申请实施例不进行限定。
步骤S103，将待识别猪只咳嗽数据输入至猪只疾病识别模型，识别出待识别猪只咳嗽数据对应的猪只疾病类型。
其中，待识别猪只咳嗽数据为采集到的未知疾病与否、未知疾病类型的猪只音频数据，并且待识别猪只咳嗽数据与猪只咳嗽样本数据的采集过程相同，通过猪只疾病识别模型可以识别出猪只疾病与否以及猪只疾病类型，当然如果猪只疾病识别模型的输出分类设置有猪只疾病阶段分类，通过猪只疾病识别模型也可以识别出猪只疾病所处阶段，本申请实施例不进行限定。
例如，猪只疾病识别模型的输出设置有7种猪只疾病类型，并未设置每种疾病类型所处的疾病阶段，当然还会设置其他来表明猪只健康，将待识别猪只咳嗽数据通过猪只疾病识别模型可以输出猪只疾病类型，如果猪只健康则输出其他。
通过本申请，将猪只咳嗽样本输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型记录有猪只咳嗽数据与猪只疾病类型的映射关系，通过猪只疾病识别模型即可准确识别出猪只疾病，与现有技术通过监控猪只音频数据的方式识别猪只疾病的方法相比，本申请实施例通过构建猪只识别模型对猪只疾病进行识别，无需在监控到猪只音频数据发生异常时，兽医逐个检查以及逐个去判断猪只疾病类型，节省人力时间，同事方便猪只饲养场所快速部署猪只治疗以及对应的治疗方案，降低因疾病产生的经济损失。
图2是根据本申请优选实施例的猪只疾病识别方法的流程图，如图2所示，该方法包括以下步骤：
步骤S201，获取预设时间段内不同猪只疾病类型的猪只音频样本数据。
在本申请实施例中，将不同疾病类型以及健康猪只分离的方式来进行养殖，每头猪只拥有独立的活动空间，保证采集到猪只音频样本不会受到其他猪只声音的干扰，进一步通过安置传感器来实时采集猪只音频样本数据。
考虑到猪只样本数量问题，如果记录猪只音频数据时间过久，则每只猪只样本的音频数据需要花费大量的处理时间，通过设置预设时间段来控制音频数据采集的时间长度，方便后续猪只音频样本数据的处理。
例如，创建一个50头携带不同疾病类型的猪只音频的数据库，每头猪只记录2个小时内的音频数据，而在这两个小时过程中每头猪只会发出至少30个咳嗽声音，覆盖了不同疾病类型，这里的猪只音频除了猪只咳嗽声，还有猪只叫声、猪只呼吸声、猪只心跳声、猪只撞栏的噼啪声等。
步骤S202，根据每个猪只样本身上所携带的猪只疾病类型对猪只音频样本数据进行标记，得到多个携带猪只疾病类型标签的猪只音频样本数据。
具体对每个猪只音频样本数据进行标记的方式可以使用声音编辑器在PC上手工标记，例如，猪只样本的疾病类型为喘病，则在该猪只音频样本数据上标记喘病，对于其他猪只音频样本数据的标记同理。
对于本申请实施例，通过根据每个猪只样本身上所携带的猪只疾病类型对猪只音频样本数据进行标记，预先对不同疾病类型的猪只音频样本数据进行分类，从而确定样本数据输入模式，将分类后的猪只样本数据作为正例样本输入至神经网络结构中，方便后续对猪只音频样本数据进行训练。
步骤S203，对预设时间段内的猪只音频样本数据进行分帧处理，得到多帧预设时长的猪只音频样本序列。
需要说明的是，由于猪只音频样本数据持续时间较长，在对猪只音频数据进行分帧处理之前，可以预先从猪只音频样本数据中生成随机的高斯分布的持续时间段，对猪只音频样本数据进行量化，使得所有声音的平均持续时间固定。
例如在所有采集的猪只音频数据范围内咳嗽持续时间从250毫秒到最多800毫秒，通过对猪只音频数据进行量化，使得所有猪只音频数据的平局持续时间为320毫秒。
步骤S204，对猪只音频样本序列进行过滤，保留猪只咳嗽样本数据。
由于猪只音频样本序列中除了猪只咳嗽声音外，还包括其他的猪只咳嗽声音，如猪只 呼吸声、猪只心跳声音等不相关的音频数据，需要对猪只音频样本序列进行过滤，保留猪只咳嗽样本数据，才能提取到猪只咳嗽音频数据。
步骤S205，将猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系。
对于本申请实施例，循环神经网络可以有多层结构组成，每层结构具有不同的输入输出参数以及实现不同的功能，通过循环神经网络来对携带不同疾病类型标签的猪只咳嗽样本数据进行反复训练，得到猪只咳嗽样本数据与不同猪只疾病类型的映射关系，该映射关系相当于猪只疾病识别模型。
例如，循环神经网络为5层结构组成，第一层结构为编码层结构，该编码层由3层编码器组成，包括128神经元、64神经元的2个双向循环层以及32个循环神经元的单向层，这里的编码器被设置为可以处理最大长度为设定值的任意序列，并且编码器中所有的循环神经元可以采用GRU结构，GRU结构简单，通过更新门和重置门来决定之前状态的依赖程度，从而结局远距离依赖问题，通过编码层结构提取出猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征参数；第二层结构为固定的编码层，固定的编码层可以包括一个固定参数的32神经元的激活层，通过固定的编码层来汇总各个疾病阶段的特征参数，输出不同猪只疾病类型各个疾病阶段的特征矩阵；第三层结构为解码层，解码层可以包括一个单独的循环层，具有64个长短记忆LSTM单元，结合注意力机制，使得网络主要关注输入特征的显著部分，提高分类性能，通过解码层根据不同猪只疾病类型标签来优化猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵内的特征参数，输出不同猪只疾病类型标签与猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵对应的映射关系；第四层结构为全连接层，全连接层可以包括256个神经元，将训练的分布式特征表示映射到样本标记空间，组合学习整体特征，通过全连接层将猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵进行融合，输出携带有不同猪只疾病类型标签的特征矩阵；第五层结构为分类层，分类层可以使用softmax输出分类标签，该softmax函数可以将输入映射成为(0,1)的数值，该数值可以理解为概率，取最大概率作为分类结果，通过分类层对携带有不同猪只疾病标签的特征矩阵进行分类，构建猪只疾病识别模型。
具体猪只疾病识别模型的训练过程可以如图3A所示，在输入猪只咳嗽样本数据后，对猪只咳嗽样本数据进行分类标注以及预处理，得到猪只咳嗽样本序列，对携带有不同疾病标签的猪只咳嗽样本序列进行训练，得到猪只疾病识别模型。
步骤S206，将待识别猪只咳嗽数据输入至猪只疾病识别模型，识别出待识别猪只咳嗽数据对应的猪只疾病类型。
需要说明的是，由于猪只疾病类型所处的不同疾病阶段可能存在一定的潜伏期，为了进一步准确判断猪只疾病类型，在识别出猪只疾病类型后，可以选取预设时间段内的该疾病猪只咳嗽数据进行比对，从而确定猪只疾病类型。
具体猪只疾病识别模型的识别过程可以如图3B所示，在输入待识别猪只咳嗽数据后，对猪只咳嗽数据预处理，得到猪只咳嗽样本序列，将猪只咳嗽样本序列输入至主治识别模型，分类得到猪只疾病类型。
本申请实施例的猪只疾病识别模型相当于使用序列到序列的循环神经网络将一个未分割的输入序列映射到另一个输出标签，考虑到猪只咳嗽数据在各个疾病阶段的依赖性，循环神经网络可以捕获时间和光谱之间的依赖性，根据咳嗽数据在各个疾病阶段的特征参数，来进行咳嗽数据对应疾病类型的判别任务，使得各层输出会依赖于上一层的计算，从而提高猪只疾病识别精度。
步骤S207，根据识别出的疾病类型建立猪只疾病信息，该猪只疾病信息中记录有猪只信息、识别出的猪只疾病类型以及判断猪只疾病类型所处的疾病阶段。
其中，猪只信息可以包括但不局限于猪只所处饲养场所，猪只编码等用于标识猪只身份的信息，例如，猪只饲养场所为了方便对猪只管理，为猪只进行隔离圈养，每个猪圈都有标识信息，并且每个猪只身上记录有用于区别于其他猪只的编码。
对于本申请实施例，在识别出猪只疾病类型后，为了便于后续猪只疾病的治疗，通过建立猪只疾病信息的方式来整理疾病猪只的信息，以便于后续的猪只疾病治疗。
步骤S208，根据猪只疾病信息中猪只疾病类型所处的疾病阶段对猪只设置不同的治疗策略。
由于所处不同疾病阶段的猪只所表现症状有所不同，并且需要的治疗策略有所不同，为了节省人力治疗时间，可以在建立猪只疾病信息后，对不同疾病阶段的猪只设置不同的治疗策略。例如，对于疾病初期的猪只可通过在猪只饲料中添加治疗药物，或者简单观察即可，对于疾病晚期的猪只需要通知兽医紧急治疗或者需要一些应急措施。
为了保证疾病猪只的安全性，在根据猪只疾病信息中猪只疾病类型所处的疾病阶段对猪只设置不同的治疗策略之后，由于猪只信息中携带有猪只标识信息，通过建立包含不同猪只饲养机构的机构信息以及不同猪只饲养机构内的猪只标识信息的预设数据库，根据猪只标识信息从预设数据库中查找负责饲养猪只的机构信息，并将猪只疾病信息以及不同的治疗策略发送至负责饲养猪只的机构，以便于对疾病猪只尽早治疗。
通过本申请实施例，将猪只咳嗽样本输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型记录有猪只咳嗽数据与猪只疾病类型的映射关系，通过 猪只疾病识别模型即可准确识别出猪只疾病，与现有技术通过监控猪只音频数据的方式识别猪只疾病的方法相比，本申请实施例通过构建猪只识别模型对猪只疾病进行识别，无需在监控到猪只音频数据发生异常时，兽医逐个检查以及逐个去判断猪只疾病类型，节省人力时间，同事方便猪只饲养场所快速部署猪只治疗以及对应的治疗方案，降低因疾病产生的经济损失。
图4是根据本申请优选实施例的猪只疾病识别方法的流程图，如图4所示，该方法包括以下步骤：
步骤S301，获取预设时间段内不同猪只疾病类型的猪只音频样本数据。
对于本申请实施例，具体采集猪只音频样本数据的具体实现方法与步骤S201相同，在此不进行赘述。
需要说明的是，为了方便猪只音频数据采集，本申请实施例设计一种可穿戴设备来实时获取猪只音频样本数据，这里包括猪只呼吸声、叫声以及咳嗽声等音频数据，并且可以将采集到的数据实时传输到云端。该设备由压电传感器和附加元件组成，通过信号调理电路使得传感器可以放大呼吸声和完全消除环境声音，另外，考虑到猪只身上的其他部位容易被磨蹭，可以将设备固定在猪只的耳朵上。
步骤S302，根据每个猪只样本身上所携带的猪只疾病类型所处不同疾病阶段对猪只音频样本数据进行标记，得到多个携带猪只疾病类型所处不同疾病阶段标签的猪只音频样本数据。
需要说明的是，这里对每个猪只音频样本数据的标记方式与步骤S202采用相同的方式，再此不进行赘述。
对于本申请实施例，为了精确不同疾病类型的训练效果，还可以针对疾病类型所处不同疾病阶段选取猪只样本作为训练样本，通过根据猪只样本身上所携带的猪只疾病类型所处不同疾病阶段对猪只音频样本数据进行标记，预先对不同疾病阶段的猪只音频样本数据进行分类，从而确定样本数据输入模式，将分类后的猪只样本数据作为正例样本输入至神经网络结构中，方便后续对猪只音频样本数据进行训练。
步骤S303，对预设时间段内的猪只音频样本数据进行分帧处理，得到多帧预设时长的猪只音频样本序列。
对于本申请实施例，猪只音频样本数据并非是稳定的音频信号，通过对猪只音频样本数据进行分帧，得到多帧预设时长的猪只音频样本序列，进而将每一帧猪只音频样本序列看成稳定的音频信号，以便于后续对猪只咳嗽数据进行处理。
步骤S304，通过能量窗口检测猪只音频样本序列的信号能量。
对于本申请实施例，不同猪只音频样本序列具有不同的信号能量，通过能量窗口检测每帧猪只音频样本序列的信号能量，来判断猪只音频样本序列的音频特征，进而对不同音频特征的猪只音频样本序列进行处理。
步骤S305，过滤低于预设信号能量的猪只音频样本序列，将过滤得到的剩余的猪只音频样本序列作为猪只咳嗽样本数据。
由于猪只呼吸声以及猪只心跳声等音频序列的信号能量通常较低，为了从多种音频特征的猪只音频样本序列中筛选出猪只咳嗽样本序列，通常将预设信号能量设置较低，将低于预设信号能量的猪只音频样本序列进行过滤，以去除干扰的猪只音频数据，并将过滤得到剩余猪只音频样本序列作为猪只咳嗽样本数据。
步骤S306，将猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型所处不同疾病阶段的映射关系。
对于本申请实施例，通过循环神经网络来对携带猪只疾病类型所处不同疾病阶段标签的猪只音频样本数据进行反复训练，得到猪只咳嗽数据与猪只疾病类型所处不同疾病阶段的映射关系，该映射关系相当于猪只疾病识别模型。
需要说明的是，这里对猪只咳嗽样本数据进行训练的具体实现方式与步骤S205采用相同的方式，再此不进行赘述。
步骤S307，将待识别猪只咳嗽数据输入至猪只疾病识别模型，识别出待识别猪只咳嗽数据对应的猪只疾病类型以及猪只疾病类型所处的疾病阶段。
通过本申请，将猪只咳嗽样本输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型还记录有猪只咳嗽数据与猪只疾病类型所处不同疾病阶段的映射关系，通过猪只疾病识别模型即可准确识别出待识别猪只咳嗽数据对应的猪只疾病类型以及猪只疾病类型所处的疾病阶段，无需在监控到猪只音频数据发生异常时，兽医逐个检查以及逐个去判断猪只疾病类型以及疾病类型所处的疾病阶段，节省人力时间，同事方便猪只饲养场所快速部署猪只治疗以及对应的治疗方案，降低因疾病产生的经济损失。
图5是根据本申请实施例的一种猪只疾病识别装置的结构框图。参照图5，该装置包括获取单元41，构建单元42和识别单元43。
获取单元41，可以用于获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；
构建单元42，可以用于将多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映 射关系；
识别单元43，可以用于将待识别猪只咳嗽数据输入至猪只疾病识别模型，识别出待识别猪只咳嗽数据对应的猪只疾病类型。
通过本申请，将猪只咳嗽样本输入至循环神经网络模型进行训练，构建猪只疾病识别模型，该猪只疾病识别模型记录有猪只咳嗽数据与猪只疾病类型的映射关系，通过猪只疾病识别模型即可准确识别出猪只疾病，与现有技术通过监控猪只音频数据的方式识别猪只疾病的方法相比，本申请实施例通过构建猪只识别模型对猪只疾病进行识别，无需在监控到猪只音频数据发生异常时，兽医逐个检查以及逐个去判断猪只疾病类型，节省人力时间，同事方便猪只饲养场所快速部署猪只治疗以及对应的治疗方案，降低因疾病产生的经济损失。
作为图5中所示猪只疾病识别装置的进一步说明，图6是根据本申请实施例另一种猪只疾病识别装置的结构示意图，如图6所示，该装置还包括：
建立单元44，可以用于在将待识别猪只咳嗽数据输入至猪只疾病识别模型，识别出猪只疾病类型之后，根据识别出的疾病类型建立猪只疾病信息，该猪只疾病信息中记录有猪只信息、识别出的猪只疾病类型以及判断猪只疾病类型所处的疾病阶段；
设置单元45，可以用于根据猪只疾病信息中猪只疾病类型所处的疾病阶段对猪只设置不同的治疗策略。
进一步地，该猪只信息中携带有猪只标识信息，该装置还包括：
查找单元46，可以用于在将猪只疾病信息中猪只疾病类型所处的疾病阶段对猪只初步采取不同的治疗策略之后，根据猪只标识信息从预设数据库中查找负责饲养猪只的机构信息，该预设数据库中包含不同猪只饲养机构的机构信息以及不同猪只饲养机构内的猪只标识信息；
发送单元47，可以用于将猪只疾病信息以及所述不同的治疗策略发送至负责饲养猪只的机构。
进一步地，该获取单元41包括：
获取模块411，可以用于获取预设时间段内不同猪只疾病类型的猪只音频样本数据；
分帧模块412，可以用于对预设时间段内的猪只音频样本数据进行分帧处理，得到多帧预设时长的猪只音频样本序列；
过滤模块414，可以用于对多帧预设时长的猪只音频样本序列进行过滤，将过滤后得到的多个猪只音频样本序列作为多个猪只咳嗽样本数据。
进一步地，该获取单元41还包括：
第一标记模块414，可以用于根据每个猪只样本身上所携带的猪只疾病类型对猪只音频样本数据进行标记，得到多个携带猪只疾病类型标签的猪只音频样本数据；
进一步地，上述循环神经网络为多层结构的网络模型，该构建单元32包括：
提取模块421，可以用于循环神经网络的第一层结构提取出猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征参数；
汇总模块422，可以用于循环神经网络的第二层结构汇总各个疾病阶段的特征参数，输出不同猪只疾病类型各个疾病阶段的特征矩阵；
优化模块423，可以用于循环神经网络的第三层结构根据不同猪只疾病类型标签来优化猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵内的特征参数，输出不同猪只疾病类型标签与猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵对应的映射关系；
融合模块424，可以用于循环神经网络的第四层结构将猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵进行融合，输出携带有不同猪只疾病类型标签的特征矩阵；
分类模块425，可以用于循环神经网络的第五层结构对携带有不同猪只疾病类型标签的特征矩阵进行分类，构建猪只疾病识别模型。
作为图5中所示猪只疾病识别装置的进一步说明，图7是根据本申请实施例另一种猪只疾病识别装置的结构示意图，如图7所示，所述每个猪只咳嗽样本数据还携带对应的猪只疾病类型所处不同疾病阶段标签，所述猪只疾病识别模型中还记录有猪只咳嗽数据与猪只疾病类型所处不同疾病阶段的映射关系，该获取单元41还包括：
第二标记模块415，可以用于根据每个猪只样本身上所携带的猪只疾病类型所处不同疾病阶段对猪只音频样本数据进行标记，得到携带猪只疾病类型所处不同疾病阶段标签的猪只音频样本数据。
所述识别单元43，还可以用于将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型以及猪只疾病类型所处的疾病阶段。
进一步地，过滤模块413，具体可以用于通过能量窗口检测猪只音频样本序列的信号能量；
过滤模块413，具体还可以用于过滤低于预设信号能量的猪只音频样本序列，将过滤后得到的剩余的猪只音频样本序列作为猪只咳嗽样本数据。
图8是根据一示例性实施例示出的一种猪只疾病识别装置的框图。例如，可以为一个计算机设备，装置500可以是移动电话，计算机，数字广播终端，消息收发设备，游戏控 制台，平板设备，医疗设备，健身设备，个人数字助理等。
参照图8，装置500可以包括以下一个或多个组件：处理组件502，存储器504，电源组件506，多媒体组件508，音频组件510，I/O(Input/Output，输入/输出)的接口512，传感器组件514，以及通信组件516。
处理组件502通常控制装置500的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件502可以包括一个或多个处理器520来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件502可以包括一个或多个模块，便于处理组件502和其他组件之间的交互。例如，处理组件502可以包括多媒体模块，以方便多媒体组件508和处理组件502之间的交互。
存储器504被配置为存储各种类型的数据以支持在装置500的操作。这些数据的示例包括用于在装置500上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器504可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如SRAM(Static Random Access Memory,静态随机存取存储器)，EEPROM(Electrically-Erasable Programmable Read-Only Memory,电可擦除可编程只读存储器)，EPROM(Erasable Programmable Read Only Memory,可擦除可编程只读存储器)，PROM(Programmable Read-Only Memory,可编程只读存储器)，ROM(Read-Only Memory,只读存储器)，磁存储器，快闪存储器，磁盘或光盘。
电源组件506为装置500的各种组件提供电力。电源组件506可以包括电源管理系统，一个或多个电源，及其他与为装置500生成、管理和分配电力相关联的组件。
多媒体组件508包括在所述装置500和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括LCD(Liquid Crystal Display，液晶显示器)和TP(Touch Panel，触摸面板)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件508包括一个前置摄像头和/或后置摄像头。当装置500处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件510被配置为输出和/或输入音频信号。例如，音频组件510包括一个MIC(Microphone,麦克风)，当装置500处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器 504或经由通信组件516发送。在一些实施例中，音频组件510还包括一个扬声器，用于输出音频信号。
I/O接口512为处理组件502和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件514包括一个或多个传感器，用于为装置500提供各个方面的状态评估。例如，传感器组件514可以检测到设备500的打开/关闭状态，组件的相对定位，例如组件为装置400的显示器和小键盘，传感器组件514还可以检测装置500或装置500一个组件的位置改变，用户与装置500接触的存在或不存在，装置500方位或加速/减速和装置500的温度变化。传感器组件514可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件514还可以包括光传感器，如CMOS(Complementary Metal Oxide Semiconductor，互补金属氧化物)或CCD(Charge-coupled Device，电荷耦合元件)图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件514还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件516被配置为便于装置500和其他设备之间有线或无线方式的通信。装置500可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信组件516经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件516还包括NFC(Near Field Communication,近场通信)模块，以促进短程通信。例如，在NFC模块可基于RFID(Radio Frequency Identification,射频识别)技术，IrDA(Infra-red Data Association,红外数据协会)技术，UWB(Ultra Wideband,超宽带)技术，BT(Bluetooth,蓝牙)技术和其他技术来实现。
在示例性实施例中，装置500可以被一个或多个ASIC(Application Specific Integrated Circuit,应用专用集成电路)、DSP(Digital signal Processor,数字信号处理器)、DSPD(Digital signal Processor Device，数字信号处理设备)、PLD(Programmable Logic Device,可编程逻辑器件)、FPGA)(Field Programmable Gate Array,现场可编程门阵列)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述猪只疾病识别方法。
在示例性实施例中，还提供了一种包括指令的非临时性计算机非易失性可读存储介质，例如包括指令的存储器504，上述指令可由装置500的处理器520执行以完成上述方法。例如，所述非临时性计算机非易失性可读存储介质可以是ROM、RAM(Random Access Memory,随机存取存储器)、CD-ROM(Compact Disc Read-Only Memory,光盘只读存储器)、磁带、软盘和光数据存储设备等。
一种非临时性计算机非易失性可读存储介质，当所述非易失性可读存储介质中的指令由猪只疾病识别装置的处理器执行时，使得猪只疾病识别装置能够执行上述猪只疾病识别方法。
显然，本领域的技术人员应该明白，上述的本申请的各模块或各步骤可以用通用的计算机设备来实现，它们可以集中在单个的计算机设备上，或者分布在多个计算机设备所组成的网络上，可选地，它们可以用计算机设备的计算机可读指令来实现，从而，可以将它们存储在存储装置中由计算机设备来执行，并且在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤，或者将它们分别制作成各个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样，本申请不限制于任何特定的硬件和软件结合。
以上所述仅为本申请的优选实施例而已，并不用于限制本申请，对于本领域的技术人员来说，本申请可以有各种更改和变化。凡在本申请的精神和原则之内，所作的任何修改、等同替换、改进等，均应包括在本申请的保护范围之内。

Claims (20)

  1. 一种猪只疾病识别方法，其特征在于，所述方法包括：
    获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；
    将所述多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；
    将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型。
  2. 根据权利要求1所述的方法，其特征在于，所述获取猪只咳嗽样本数据包括：
    获取预设时间段内不同猪只疾病类型的猪只音频样本数据；
    对预设时间段内的猪只音频样本数据进行分帧处理，得到多帧预设时长的猪只音频样本序列；
    对所述多帧预设时长的猪只音频样本序列进行过滤，将过滤后得到的多个猪只音频样本序列作为多个猪只咳嗽样本数据。
  3. 根据权利要求2所述的方法，其特征在于，所述在所述获取预设时间段内不同猪只疾病类型的猪只音频样本数据之后，所述方法还包括：
    根据每个猪只样本身上所携带的猪只疾病类型对猪只音频样本数据进行标记，得到多个携带猪只疾病类型标签的猪只音频样本数据。
  4. 根据权利要求2所述的方法，其特征在于，所述对猪只音频样本序列进行过滤，保留猪只咳嗽样本数据包括：
    通过能量窗口检测所述猪只音频样本序列的信号能量；
    过滤低于预设信号能量的猪只音频样本序列，将过滤后得到的剩余的猪只音频样本序列作为猪只咳嗽样本数据。
  5. 根据权利要求2所述的方法，其特征在于，所述每个猪只咳嗽样本数据还携带对应的猪只疾病类型所处不同疾病阶段标签，所述猪只疾病识别模型中还记录有猪只咳嗽数据与猪只疾病类型所处不同疾病阶段的映射关系；
    所述将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型包括：
    将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型以及猪只疾病类型所处的疾病阶段。
  6. 根据权利要求5所述的方法，其特征在于，所述在所述获取预设时间段内不同猪 只疾病类型的猪只音频样本数据之后，所述方法还包括：
    根据每个猪只样本身上所携带的猪只疾病类型所处不同疾病阶段对猪只音频样本数据进行标记，得到多个携带猪只疾病类型所处不同疾病阶段标签的猪只音频样本数据。
  7. 根据权利要求5或6所述的方法，其特征在于，所述循环神经网络为多层结构的网络模型，所述将所述猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型包括：
    所述循环神经网络的第一层结构提取出猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征参数；
    所述循环神经网络的第二层结构汇总所述各个疾病阶段的特征参数，输出不同猪只疾病类型各个疾病阶段的特征矩阵；
    所述循环神经网络的第三层结构根据不同猪只疾病类型标签来优化猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵内的特征参数，输出不同猪只疾病类型标签与猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵对应的映射关系；
    所述循环神经网络的第四层结构将猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵进行融合，输出携带有不同猪只疾病类型标签的特征矩阵；
    所述循环神经网络的第五层结构对携带有不同猪只疾病类型标签的特征矩阵进行分类，构建猪只疾病识别模型。
  8. 一种猪只疾病识别装置，其特征在于，所述装置包括：
    获取单元，用于获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；
    构建单元，用于将所述猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；
    识别单元，用于将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型。
  9. 根据权利要求8所述的装置，其特征在于，所述获取单元包括：获取模块，用于获取预设时间段内不同猪只疾病类型的猪只音频样本数据；分帧模块，用于对预设时间段内的猪只音频样本数据进行分帧处理，得到多帧预设时长的猪只音频样本序列；过滤模块，用于对所述多帧预设时长的猪只音频样本序列进行过滤，将过滤后得到的多个猪只音频样本序列作为多个猪只咳嗽样本数据。
  10. 根据权利要求9所述的装置，其特征在于，所述获取单元还包括：第一标记模块， 用于根据每个猪只样本身上所携带的猪只疾病类型对猪只音频样本数据进行标记，得到多个携带猪只疾病类型标签的猪只音频样本数据。
  11. 根据权利要求9所述的装置，其特征在于，所述过滤模块，具体用于通过能量窗口检测所述猪只音频样本序列的信号能量；所述过滤模块，具体还用于过滤低于预设信号能量的猪只音频样本序列，将过滤后得到的剩余的猪只音频样本序列作为猪只咳嗽样本数据。
  12. 根据权利要求9所述的装置，其特征在于，所述每个猪只咳嗽样本数据还携带对应的猪只疾病类型所处不同疾病阶段标签，所述猪只疾病识别模型中还记录有猪只咳嗽数据与猪只疾病类型所处不同疾病阶段的映射关系；所述识别单元，还用于所述将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型以及猪只疾病类型所处的疾病阶段。
  13. 根据权利要求12所述的装置，其特征在于，所述获取单元还包括：第二标记模块，用于根据每个猪只样本身上所携带的猪只疾病类型所处不同疾病阶段对猪只音频样本数据进行标记，得到多个携带猪只疾病类型所处不同疾病阶段标签的猪只音频样本数据。
  14. 根据权利要求12或13所述的装置，其特征在于，所述循环神经网络为多层结构的网络模型，所述构建单元包括：提取模块，用于所述循环神经网络的第一层结构提取出猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征参数；汇总模块，用于所述循环神经网络的第二层结构汇总所述各个疾病阶段的特征参数，输出不同猪只疾病类型各个疾病阶段的特征矩阵；优化模块，用于所述循环神经网络的第三层结构根据不同猪只疾病类型标签来优化猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵内的特征参数，输出不同猪只疾病类型标签与猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵对应的映射关系；融合模块，用于所述循环神经网络的第四层结构将猪只咳嗽样本数据在不同猪只疾病类型各个疾病阶段的特征矩阵进行融合，输出携带有不同猪只疾病类型标签的特征矩阵；分类模块，用于所述循环神经网络的第五层结构对携带有不同猪只疾病类型标签的特征矩阵进行分类，构建猪只疾病识别模型。
  15. 一种计算机非易失性可读存储介质，其上存储有计算机可读指令，其特征在于，所述计算机可读指令被处理器执行时实现猪只疾病识别方法，包括：获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；将所述多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型。
  16. 根据权利要求15所述的计算机非易失性可读存储介质，其特征在于，所述计算机可读指令被处理器执行时实现所述获取猪只咳嗽样本数据包括：获取预设时间段内不同猪只疾病类型的猪只音频样本数据；对预设时间段内的猪只音频样本数据进行分帧处理，得到多帧预设时长的猪只音频样本序列；对所述多帧预设时长的猪只音频样本序列进行过滤，将过滤后得到的多个猪只音频样本序列作为多个猪只咳嗽样本数据。
  17. 根据权利要求16所述的计算机非易失性可读存储介质，其特征在于，所述计算机可读指令被处理器执行时实现所述在所述获取预设时间段内不同猪只疾病类型的猪只音频样本数据之后，所述方法还包括：根据每个猪只样本身上所携带的猪只疾病类型对猪只音频样本数据进行标记，得到多个携带猪只疾病类型标签的猪只音频样本数据。
  18. 一种计算机设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令，其特征在于，所述处理器执行所述计算机可读指令时实现猪只疾病识别方法，包括：获取多个猪只咳嗽样本数据，每个猪只咳嗽样本数据携带对应的猪只疾病类型标签；将所述多个猪只咳嗽样本数据输入至循环神经网络模型进行训练，构建猪只疾病识别模型，所述猪只疾病识别模型中记录有猪只咳嗽数据与猪只疾病类型的映射关系；将待识别猪只咳嗽数据输入至所述猪只疾病识别模型，识别出所述待识别猪只咳嗽数据对应的猪只疾病类型。
  19. 根据权利要求18所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时实现所述获取猪只咳嗽样本数据包括：获取预设时间段内不同猪只疾病类型的猪只音频样本数据；对预设时间段内的猪只音频样本数据进行分帧处理，得到多帧预设时长的猪只音频样本序列；对所述多帧预设时长的猪只音频样本序列进行过滤，将过滤后得到的多个猪只音频样本序列作为多个猪只咳嗽样本数据。
  20. 根据权利要求19所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时实现所述在所述获取预设时间段内不同猪只疾病类型的猪只音频样本数据之后，所述方法还包括：根据每个猪只样本身上所携带的猪只疾病类型对猪只音频样本数据进行标记，得到多个携带猪只疾病类型标签的猪只音频样本数据。
PCT/CN2018/111654 2018-07-05 2018-10-24 猪只疾病识别方法、装置及计算机可读存储介质 Ceased WO2020006934A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810729269.1A CN108935188A (zh) 2018-07-05 2018-07-05 猪只疾病识别方法、装置及电子设备
CN201810729269.1 2018-07-05

Publications (1)

Publication Number Publication Date
WO2020006934A1 true WO2020006934A1 (zh) 2020-01-09

Family

ID=64485824

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/111654 Ceased WO2020006934A1 (zh) 2018-07-05 2018-10-24 猪只疾病识别方法、装置及计算机可读存储介质

Country Status (2)

Country Link
CN (1) CN108935188A (zh)
WO (1) WO2020006934A1 (zh)

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109637549A (zh) * 2018-12-13 2019-04-16 北京小龙潜行科技有限公司 一种对猪只进行声音检测的方法、装置和检测系统
BE1026886B1 (nl) * 2018-12-18 2020-07-22 Soundtalks Nv Werkwijze voor het monitoren van een veeteeltfaciliteit en/of veeteeltdieren in een veeteeltfaciliteit met behulp van verbeterde geluidsverwerkingstechnieken
CN109602421A (zh) * 2019-01-04 2019-04-12 平安科技（深圳）有限公司 健康监测方法、装置及计算机可读存储介质
CN110188669B (zh) * 2019-05-29 2021-01-19 华南理工大学 一种基于注意力机制的空中手写字符轨迹恢复方法
JP7312037B2 (ja) * 2019-06-25 2023-07-20 パナソニック インテレクチュアル プロパティ コーポレーション オブ アメリカ 咳検出装置、咳検出装置の作動方法及びプログラム
CN111183917B (zh) * 2019-11-14 2022-02-01 京东科技信息技术有限公司 一种动物异常监测、图像处理方法及装置
CN111160271B (zh) * 2019-12-30 2023-08-18 哈尔滨商业大学 一种基于深度学习的室内养鸡场病鸡检测系统
CN111681755A (zh) * 2020-04-28 2020-09-18 北京农业信息技术研究中心 一种生猪疾病诊疗系统及方法
CN112185523B (zh) * 2020-09-30 2023-09-08 南京大学 基于多尺度卷积神经网络的糖尿病视网膜病变分类方法
CN113671164A (zh) * 2021-09-26 2021-11-19 吾征智能技术(北京)有限公司 一种基于大便颜色及气味判断疾病的系统、设备及介质
CN114041779A (zh) * 2021-11-26 2022-02-15 河南牧原智能科技有限公司 用于对牲畜呼吸道疾病进行识别的识别系统和计算机设备
CN114970772B (zh) * 2022-07-26 2022-11-04 北京纳通医用机器人科技有限公司 用于肺部疾病检测的模型训练方法、设备、装置和介质
CN115620184B (zh) * 2022-08-09 2026-02-03 武汉长江通信智联技术有限公司 基于红外热成像的旅客突发心脏病识别系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106175726A (zh) * 2016-07-11 2016-12-07 夏茂 基于bp神经网络的人体健康状况识别方法
CN106251880A (zh) * 2015-06-03 2016-12-21 创心医电股份有限公司 识别生理声音的方法以及系统
CN106778014A (zh) * 2016-12-29 2017-05-31 浙江大学 一种基于循环神经网络的患病风险预测方法
CN107818366A (zh) * 2017-10-25 2018-03-20 成都力创昆仑网络科技有限公司 一种基于卷积神经网络的肺音分类方法、系统及用途

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB0711524D0 (en) * 2007-06-15 2007-07-25 Univ Leuven Kath Online recognition and localisation of sick pig cough sounds
CN105336331A (zh) * 2015-11-21 2016-02-17 太原理工大学 基于声音的猪异常行为智能监测方法及智能监测系统
CN106847262A (zh) * 2016-12-28 2017-06-13 华中农业大学 一种猪呼吸道疾病自动识别报警方法
CN107799114A (zh) * 2017-04-26 2018-03-13 珠海智牧互联科技有限公司 一种猪只咳嗽声音识别方法及系统
CN107094633A (zh) * 2017-05-25 2017-08-29 仁怀市泓熙养殖专业合作社 具有呼吸道疾病监测功能的猪舍
CN107680597B (zh) * 2017-10-23 2019-07-09 平安科技（深圳）有限公司 语音识别方法、装置、设备以及计算机可读存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106251880A (zh) * 2015-06-03 2016-12-21 创心医电股份有限公司 识别生理声音的方法以及系统
CN106175726A (zh) * 2016-07-11 2016-12-07 夏茂 基于bp神经网络的人体健康状况识别方法
CN106778014A (zh) * 2016-12-29 2017-05-31 浙江大学 一种基于循环神经网络的患病风险预测方法
CN107818366A (zh) * 2017-10-25 2018-03-20 成都力创昆仑网络科技有限公司 一种基于卷积神经网络的肺音分类方法、系统及用途

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHONG CHANGLE ET AL: "Design and Implementation of Animal Disease Diagnosis Expert System Based on BP neural network", MODERN COMPUTER, vol. 4, 30 April 2010 (2010-04-30), XP055671021 *

Also Published As

Publication number Publication date
CN108935188A (zh) 2018-12-07

Similar Documents

Publication Publication Date Title
WO2020006934A1 (zh) 猪只疾病识别方法、装置及计算机可读存储介质
EP3693966B1 (en) System and method for continuous privacy-preserved audio collection
US11013467B2 (en) System, method and computer program product for physiological monitoring
US20210338103A1 (en) Screening of individuals for a respiratory disease using artificial intelligence
WO2020010694A1 (zh) 动物健康监测方法、装置及计算机可读存储介质
CN111467074B (zh) 用于检测牲畜状态的方法和装置
KR102552576B1 (ko) 반려동물 모니터링 방법 및 그 장치
CN109637549A (zh) 一种对猪只进行声音检测的方法、装置和检测系统
CN107368567B (zh) 动物语言识别方法及用户终端
CN107992199A (zh) 一种用于电子设备的情绪识别方法、系统及电子设备
CN109602421A (zh) 健康监测方法、装置及计算机可读存储介质
TW201828238A (zh) 認知訓練系統
CN113569671B (zh) 异常行为报警方法、装置
CN105975981A (zh) 疫情监测方法及装置
CN110381296A (zh) 一种空巢老人智能健康管理系统及方法
CN109077741A (zh) 心理状态识别方法及系统
KR20140023458A (ko) 언어 유창성 기반 치매 진단 방법 및 이를 위한 장치
CN119625779A (zh) 一种基于深度学习的羊反刍行为识别方法及系统
US12419509B1 (en) Camera accessory device for a laryngoscope and an artificial intelligence and pattern recognition system using the collected images
Dwivedi et al. Real-time digital dermatitis detection in dairy cows on Android and iOS apps using computer vision techniques
KR102171742B1 (ko) 노인 케어 시스템 및 그 방법
DE112021003125T5 (de) Systeme und Verfahren zur Erzeugung früher gesundheitsbasierter Alarme aus kontinuierlich erfassten Daten
KR102406249B1 (ko) 브러시, 상기 브러시의 동작 방법 및 상기 브러시를 포함하는 병변 관리 시스템
CN117973923A (zh) 护理水平评估及其模型训练方法、装置和设备
US11751774B2 (en) Electronic auscultation and improved identification of auscultation audio samples

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18925636

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18925636

Country of ref document: EP

Kind code of ref document: A1