CN116486819A - Pig respiratory disease monitoring method and system and farm inspection robot - Google Patents
Pig respiratory disease monitoring method and system and farm inspection robot Download PDFInfo
- Publication number
- CN116486819A CN116486819A CN202310746188.3A CN202310746188A CN116486819A CN 116486819 A CN116486819 A CN 116486819A CN 202310746188 A CN202310746188 A CN 202310746188A CN 116486819 A CN116486819 A CN 116486819A
- Authority
- CN
- China
- Prior art keywords
- pig
- feature vector
- monitoring
- biological
- result
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 44
- 208000023504 respiratory system disease Diseases 0.000 title claims abstract description 41
- 238000007689 inspection Methods 0.000 title claims abstract description 26
- 239000013598 vector Substances 0.000 claims abstract description 85
- 238000004458 analytical method Methods 0.000 claims abstract description 61
- 238000012545 processing Methods 0.000 claims abstract description 37
- 230000008451 emotion Effects 0.000 claims abstract description 33
- 210000001260 vocal cord Anatomy 0.000 claims abstract description 28
- 210000002345 respiratory system Anatomy 0.000 claims abstract description 27
- 238000003745 diagnosis Methods 0.000 claims abstract description 20
- 238000007781 pre-processing Methods 0.000 claims abstract description 19
- 239000012634 fragment Substances 0.000 claims abstract description 12
- 230000002776 aggregation Effects 0.000 claims abstract description 11
- 238000004220 aggregation Methods 0.000 claims abstract description 11
- 238000012351 Integrated analysis Methods 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 14
- 230000015556 catabolic process Effects 0.000 claims description 11
- 238000006731 degradation reaction Methods 0.000 claims description 11
- 210000003205 muscle Anatomy 0.000 claims description 11
- 201000010099 disease Diseases 0.000 claims description 10
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 7
- 230000001953 sensory effect Effects 0.000 claims description 7
- 238000000556 factor analysis Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000004088 simulation Methods 0.000 claims description 5
- 238000013526 transfer learning Methods 0.000 claims description 4
- 206010063659 Aversion Diseases 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 3
- 230000007935 neutral effect Effects 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 description 24
- 241000282887 Suidae Species 0.000 description 22
- 230000005236 sound signal Effects 0.000 description 12
- 206010011224 Cough Diseases 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 6
- 230000003862 health status Effects 0.000 description 3
- 244000144972 livestock Species 0.000 description 3
- 210000004072 lung Anatomy 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000009395 breeding Methods 0.000 description 2
- 230000001488 breeding effect Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 244000144977 poultry Species 0.000 description 2
- 230000000241 respiratory effect Effects 0.000 description 2
- 206010041232 sneezing Diseases 0.000 description 2
- 206010027940 Mood altered Diseases 0.000 description 1
- 206010028289 Muscle atrophy Diseases 0.000 description 1
- 206010037660 Pyrexia Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000000090 biomarker Substances 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 235000013372 meat Nutrition 0.000 description 1
- 230000005541 medical transmission Effects 0.000 description 1
- 230000007510 mood change Effects 0.000 description 1
- 238000009304 pastoral farming Methods 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000006116 polymerization reaction Methods 0.000 description 1
- 238000009374 poultry farming Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000002685 pulmonary effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000004148 unit process Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000001755 vocal effect Effects 0.000 description 1
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, e.g. detecting heat or mating
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
- G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B31/00—Predictive alarm systems characterised by extrapolation or other computation using updated historic data
-
- 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
- G10L17/02—Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction
-
- 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
- G10L17/04—Training, enrolment or model building
-
- 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
- G10L17/06—Decision making techniques; Pattern matching strategies
- G10L17/14—Use of phonemic categorisation or speech recognition prior to speaker recognition or verification
-
- 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
- G10L17/18—Artificial neural networks; Connectionist approaches
-
- 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
- G10L25/48—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
- G10L25/63—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 estimating an emotional state
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
- Y02A40/70—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in livestock or poultry
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Environmental Sciences (AREA)
- Business, Economics & Management (AREA)
- Animal Husbandry (AREA)
- Game Theory and Decision Science (AREA)
- Emergency Management (AREA)
- Signal Processing (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Psychiatry (AREA)
- Hospice & Palliative Care (AREA)
- Child & Adolescent Psychology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a monitoring method and a system for porcine respiratory disease and a farm inspection robot, which comprise the following steps: preprocessing the collected pig audio fragments, wherein the preprocessing comprises MFCC processing and Poisson attenuation processing to obtain acoustic feature vectors; inputting the acoustic feature vector into a biological reasoning model, the biological reasoning model at least comprising: a first ResNet residual network for extracting vocal cord feature vectors, a second ResNet residual network for extracting respiratory tract feature vectors, a third ResNet residual network for extracting emotion feature vectors; carrying out integrated analysis on the vocal cord feature vector, the respiratory tract feature vector and the emotion feature vector through an aggregation model, and outputting a pig state diagnosis result and a biological latitude significance analysis result; thereby improving the accuracy of the monitoring result of the porcine respiratory disease and the interpretability of the diagnosis result.
Description
Technical Field
The invention relates to the technical field of pig breeding, in particular to a method for monitoring porcine respiratory disease, a system for monitoring porcine respiratory disease by using the method and a farm inspection robot.
Background
As people pay more attention to food safety and health, the supervision of livestock and poultry farming is also becoming more and more strict. In livestock and poultry cultivation, respiratory diseases are one of common diseases, and can not only negatively affect the growth and health of pigs, but also cause the problems of disease transmission, meat quality reduction and the like. Therefore, the research of an effective pig respiratory disease monitoring method has important significance for the development of the livestock and poultry raising industry and the guarantee of food safety.
At present, two main methods for monitoring the respiratory diseases of pigs exist in the traditional methods:
(1) By manually observing the behaviors and physical conditions of pigs, etc.: however, the method not only consumes a great deal of time and energy, but also has the problems of strong subjectivity, large workload, low efficiency and the like, and can not discover abnormal conditions in time;
(2) By computer aided monitoring: mainly comprises a video monitoring technology and an audio analysis technology. The existing method for monitoring the porcine respiratory disease based on the audio signal mainly comprises the steps of collecting porcine audio data, judging whether the porcine audio data contains cough sound or not, and sending out an alarm signal of the porcine respiratory disease if the cough sound is monitored.
The computer-aided monitoring method can improve the effectiveness and timeliness of diagnosis, but the cough sound of pigs is possibly physiological or pathological or caused by other reasons, so that misjudgment is easy to cause, the accuracy of the monitoring result is lower, and the daily work of farm workers is greatly disturbed.
Disclosure of Invention
The invention mainly aims to provide a method and a system for monitoring porcine respiratory diseases and a farm inspection robot, and aims to solve the technical problem that the existing method for monitoring porcine respiratory diseases is easy to misjudge, and the accuracy of monitoring results of porcine respiratory diseases is improved through multidimensional biological analysis.
In order to achieve the above object, the present invention provides a method for monitoring porcine respiratory disease, comprising the steps of:
preprocessing the collected pig audio fragments, wherein the preprocessing comprises MFCC processing and Poisson attenuation processing to obtain acoustic feature vectors;
inputting the acoustic feature vector into a biological reasoning model, the biological reasoning model at least comprising: a first ResNet residual network for extracting vocal cord feature vectors, a second ResNet residual network for extracting respiratory tract feature vectors, a third ResNet residual network for extracting emotion feature vectors;
and carrying out integrated analysis on the vocal cord feature vector, the respiratory tract feature vector and the emotion feature vector through an aggregation model, and outputting a pig status diagnosis result and a biological latitude significance analysis result.
Preferably, the pig audio segment is obtained by inspecting a pig column through an inspection robot, collecting pig sound through a sound sensor when the pig audio segment is close to the pig column, and carrying out segmentation processing on the collected pig sound; and recording the positioning information corresponding to the pig sound collecting position, or obtaining the corresponding positioning information by analyzing the sound source positioning of the pig sound.
Preferably, the poisson attenuation process is to perform simulation calculation of an attenuation effect caused by muscle degradation on the acoustic feature vector through a poisson attenuation template.
Preferably, the attenuation effect simulation calculation is calculated by adopting a muscle degradation index function, and the index form of the muscle degradation index function is as follows:
where point Ix is a point obtained by the MFCC and λ is the average poisson mask of all values in the MFCC; poiss (x=k) is a specific poisson distribution form thereof, and represents the probability that the value of the random variable X is k; wherein X is a discrete random variable representing the number of events occurring within a certain time period or spatial region; x=k represents that the random variable X takes on a value of k, i.e., the number of times an event occurs is k.
Preferably, the first, second and third ResNet residual networks are connected in parallel using ResNet50 models, with the 4-D tensor output layers of 7×7X108 for each ResNet50 model.
Preferably, the aggregation model includes:
the pooling layer is used for pooling the vocal cord feature vector, the respiratory tract feature vector and the emotion feature vector;
the full-connection layer is used for carrying out intensive connection treatment on the pooling result, comparing the prediction result with a preset threshold value and outputting a classification result as the pig status diagnosis result; the full connection layer adopts a Sigmoid activation function, and the classification result comprises: the diagnosis result of the pig disease is that respiratory diseases exist or no respiratory diseases exist.
Preferably, the vocal cord feature vector is formed by detecting pronunciation of m similar to English word 'Them' on collected pig audio fragments through ResNet model, and measuring pig vocal cord change according to the pronunciation; the respiratory tract feature vector is obtained by taking a data set containing similar sounds of a disease pig and a normal pig as training data, training a classification acoustic model by using transfer learning, and predicting respiratory tract feature changes according to the classification acoustic model; the emotion feature vector is used for learning emotion features on the RAVDESS voice data set by training an entity voice classifier model, wherein the emotion feature vector comprises tones of organisms in 8 emotion states, and the tones comprise: neutral, calm, happy, sad, anger, fear, aversion and surprise.
Preferably, the biological latitude significance analysis result is displayed by adopting a biological latitude significance map, and each element of the biological latitude significance map is marked manually during training and automatically generated during prediction; the biological latitude saliency map display dimension includes three major aspects: the core index significance analysis result, the sensory flow data significance analysis result and the feature mark combination significance analysis result; wherein the significance analysis result of the core index further subdivides the significance analysis result of each audio segment; the results of the significance analysis of the sensory flow data further comprise: vocal cord variation analysis results, emotion variation analysis results, muscle degradation analysis results, and respiratory tract variation analysis results; the significance analysis results of the signature combination further include: pig morphological factor analysis result, significance factor analysis result, voting credibility analysis result and pig state factor analysis result.
Corresponding to the method for monitoring the porcine respiratory disease, the invention provides a monitoring system for the porcine respiratory disease, which comprises the following components:
the audio preprocessing module is used for preprocessing the collected pig audio fragments, wherein the preprocessing comprises MFCC processing and Poisson attenuation processing to obtain acoustic feature vectors;
the feature processing module is used for inputting the acoustic feature vector into a biological reasoning model, and the biological reasoning model at least comprises: a first ResNet residual network for extracting vocal cord feature vectors, a second ResNet residual network for extracting respiratory tract feature vectors, a third ResNet residual network for extracting emotion feature vectors;
and the characteristic analysis module is used for carrying out integrated analysis on the vocal cord characteristic vector, the respiratory tract characteristic vector and the emotion characteristic vector through an aggregation model and outputting a pig state diagnosis result and a biological latitude significance analysis result.
In addition, in order to achieve the above purpose, the invention also provides a farm inspection robot, which comprises a sound sensor for collecting pig audio, wherein the farm inspection robot is connected with a central processing unit and is used for executing the steps of the method for monitoring the pig respiratory diseases.
The beneficial effects of the invention are as follows:
(1) According to the invention, the accuracy of the monitoring result of the porcine respiratory disease is improved through extraction of the vocal cord feature vector, the respiratory tract feature vector and the emotion feature vector and biological analysis of multiple dimensions;
(2) According to the invention, the inspection robot is used for inspecting pigs at regular intervals, checking whether the sound sensor detects abnormal sound in each pig house, positioning and counting the abnormality, so that the disease is early-warned in time, and the accuracy and the efficiency are higher;
(3) The invention provides a reference for diagnosis by carrying out biological latitude significance analysis on the polymerization result, and can improve the accuracy of diagnosis and the interpretability of the diagnosis result.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention and do not constitute a limitation on the invention. In the drawings:
FIG. 1 is a flow chart of a method for monitoring porcine respiratory disease according to the present invention;
fig. 2 is a significant plot of biological latitude in fig. 1.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Before the method for monitoring the porcine respiratory disease is implemented, firstly, an audio sample is required to be collected and trained, and the method specifically comprises the following steps:
(1) Collecting variable length pig sound recordings (average 3 times sound segments of each tested pig), and covering tested pigs with different ages, sexes, varieties, whether the tested pigs are in ill state or not, etc.;
(2) And converting the collected audio window in real time by using the Mel frequency cepstral coefficient. In terms of data distribution, the present invention uses all positive samples in the dataset (positive samples, i.e., containing coughs, sneezes) and randomly selects the same number of negative pigs to achieve balanced distribution. The audio frequency and the diagnosis result of the tested pig are used for training and verification and are transmitted to a discriminator constructed by a neural network algorithm;
(3) The audio convolutional neural network algorithm trains a classifier to classify normal sound and abnormal sound, and the abnormal classification comprises: cough, sneeze, etc.;
(4) And finally, giving a prediction result output by the classifier to an output program, comparing the prediction result with a set threshold value, and determining whether the input signal is abnormal sound. If the predicted result is abnormal sound, the CPU locates and records the abnormal sound.
After training, the method for monitoring the porcine respiratory disease is carried out by adopting a trained model, as shown in figure 1, and specifically comprises the following steps:
preprocessing the collected pig Audio fragment Audio Chunks, wherein the preprocessing comprises MFCC processing and Poisson attenuation processing to obtain acoustic feature vectors;
inputting the acoustic feature vector into a biological reasoning model Biomarker Models, wherein the biological reasoning model at least comprises: a first ResNet residual network for extracting vocal cord feature vectors, a second ResNet residual network for extracting respiratory tract feature vectors, a third ResNet residual network for extracting emotion feature vectors;
and carrying out integrated analysis on the vocal cord feature vector, the respiratory tract feature vector and the emotion feature vector through an aggregation model Competing Aggregator Models, and outputting Pig status diagnosis results Pig-status diagnosis and biological latitude significance analysis results Longitudinal Saliency Map.
The CNN architecture of the invention consists of three parallel ResNet50 (Vocal, lungs, sentiment) which respectively correspond to a first residual error network, a second residual error network and a third residual error network shown in figure 1, namely, a first ResNet residual error network, a second ResNet residual error network and a third ResNet residual error network; specifically, the three models are used for training the VocalCords acoustic markers respectively; lungs & RespTract respiratory contractile markers; sentiment emotion marking, the three models being respectively pre-trained coding models for the 3 aspects; the 4-D tensor output layers of 7×7×2048 for each ResNet50 model are connected in parallel.
And outputting and integrating the vectors output by the three ResNet50 models through the aggregation model, namely, taking the vectors as three dimensions, and finally, obtaining the final judgment of whether the pig state of the PigStatus is ill or not through full connection processing of Dense Dense by Sigmoid activation.
The Sigmoid function is an activation function and is mainly used for logistic regression (logistic regression) to realize a classification function; dense represents a Dense fully connected layer, a ReLU-activated 1024-neuron deep connected neural network layer, with sigmoid classification outputs.
In this embodiment, the pig audio segment is obtained by inspecting a pig field by an inspection robot, collecting pig sound by a sound sensor when the pig audio segment is close to the pig field, and performing segmentation processing on the collected pig sound; and recording the positioning information corresponding to the pig sound collecting position, or obtaining the corresponding positioning information by analyzing the sound source positioning of the pig sound.
In this embodiment, the poisson attenuation processing is a simulation calculation of an attenuation effect musculor caused by muscle degradation of the acoustic feature vector through poisson attenuation template poisson nmask. The attenuation effect simulation calculation is performed by adopting a muscle degradation index function, and the index form of the muscle degradation index function is as follows:
where point Ix is a point obtained by the MFCC and λ is the average poisson mask of all values in the MFCC; poiss (x=k) is a specific poisson distribution form thereof, and represents the probability that the value of the random variable X is k; wherein X is a discrete random variable representing the number of events occurring within a certain time period or spatial region; x=k represents that the random variable X takes on a value of k, i.e., the number of times an event occurs is k.
The first ResNet residual network, the second ResNet residual network and the third ResNet residual network are connected in parallel by using ResNet50 models, and 7X 2048 4-D tensor output layers of each ResNet50 model.
The aggregation model includes:
the pooling layer is used for pooling the vocal cord feature vector, the respiratory tract feature vector and the emotion feature vector;
the full-connection layer is used for carrying out intensive connection treatment on the pooling result, comparing the prediction result with a preset threshold value and outputting a classification result as the pig status diagnosis result; the full connection layer adopts a Sigmoid activation function, and the classification result comprises: the diagnosis result of the pig disease is that respiratory diseases exist or no respiratory diseases exist.
And the vocal cord feature vector is formed by detecting pronunciation of m similar to English words 'Them' of collected pig audio fragments through a ResNet model, and measuring pig vocal cord change according to the pronunciation.
The respiratory tract feature vector is obtained by taking a data set containing similar sounds of a disease pig and a normal pig as training data, training a classification acoustic model by using transfer learning, and predicting respiratory tract feature changes according to the classification acoustic model; the respiratory characteristic changes include pulmonary changes that can be used to distinguish between the same type of sound in uninfected and infected pigs: such as sneezing in a fever pig, is not the same as sneezing in a normal pig. The learning process starts training from a classified acoustic model, and the training data is a data set which simultaneously contains similar sounds of the sick pigs and the normal pigs by using transfer learning.
The emotion feature vector is used for learning emotion features on the RAVDESS voice data set by training an entity voice classifier model, wherein the emotion feature vector comprises tones of organisms in 8 emotion states, and the tones comprise: neutral, calm, happy, sad, anger, fear, aversion and surprise.
As shown in fig. 2, in this embodiment, the biological latitude saliency analysis result is displayed by using a biological latitude saliency map Longitudinal Saliency Map, and each element of the biological latitude saliency map is manually marked during training and automatically generated during prediction; the biological latitude saliency map display dimension includes three major aspects: a significance analysis result of a core index (Brain OS), a significance analysis result of Sensory Stream data (Sensory Stream), and a significance analysis result of a feature tag combination (Symbolic Compositional); wherein the significance analysis result of the core index further subdivides the significance analysis result of each audio segment (Cough); the results of the significance analysis of the sensory flow data further comprise: vocal cord changes (v.chord) analysis results, mood changes (Sentiment) analysis results, muscle degeneration (musculor) analysis results, respiratory tract changes (Lungs) analysis results; the significance analysis results of the signature combination further include: PIG morphological Factor (OVBMPrection) analysis results, saliency Factor (salt Factor) analysis results, voting reliability (Voting Confidence) analysis results, PIG status Factor (PIG-status Factor) analysis results. The biological latitude significance analysis result is displayed by adopting a biological latitude significance graph.
The invention also correspondingly provides a monitoring system for the porcine respiratory disease, which comprises a sound sensor, a patrol robot, a central processing unit, a data memory, a display and the like, wherein the central processing unit at least comprises:
the audio preprocessing module is used for preprocessing the collected pig audio fragments, wherein the preprocessing comprises MFCC processing and Poisson attenuation processing to obtain acoustic feature vectors;
the feature processing module is used for inputting the acoustic feature vector into a biological reasoning model, and the biological reasoning model at least comprises: a first ResNet residual network for extracting vocal cord feature vectors, a second ResNet residual network for extracting respiratory tract feature vectors, a third ResNet residual network for extracting emotion feature vectors;
and the characteristic analysis module is used for carrying out integrated analysis on the vocal cord characteristic vector, the respiratory tract characteristic vector and the emotion characteristic vector through an aggregation model and outputting a pig state diagnosis result and a biological latitude significance analysis result.
The audio preprocessing module, the feature processing module and the feature analysis module can be partially or completely arranged in the inspection robot, and/or the audio preprocessing module, the feature processing module and the feature analysis module can be partially or completely arranged in the central processing unit.
The specific implementation process in this embodiment is as follows:
1. the inspection robot reaches pig columns, and sound sensors are carried to monitor sound in each pig house:
the inspection robot is core equipment for realizing sound monitoring and can automatically travel to a pig field for inspection. The robot is provided with a sound sensor for collecting sound signals emitted by pigs. When each pig house monitors sound, the inspection robot can enable the sound sensor to be close to pigs and collect sound signals of the pigs.
2. The sound sensor transmits sound signals sent by pigs to the central processing unit:
after the sound sensor on the inspection robot collects the sound signals sent by pigs, the signals are transmitted to the central processing unit through the transmission line connected with the inspection robot. The transmission line may be wireless or wired, depending on the actual situation.
3. The central processing unit processes and analyzes the sound signals, and judges whether the sound signals belong to abnormal sounds or not by utilizing an audio recognition technology:
in this embodiment, the central processing unit is the core control device of this patent, and is mainly responsible for processing and analyzing the transmitted sound signals. During the processing, the voice signal is analyzed by utilizing the audio recognition technology to judge whether the signal belongs to abnormal voice, such as cough, sneeze and the like. When the signal is judged to be abnormal sound, the CPU positions and records the abnormal sound.
In this embodiment, when the central processor determines that the sound signal is an abnormal sound, the information of the abnormal sound is located and recorded in the data memory. The data storage can be a storage device such as a hard disk, a solid state disk and the like.
And the CPU performs statistics and analysis on the abnormal sound information recorded in the data memory to generate a report of the abnormal sound. The abnormal sound report can be output through the display, so that timely early warning of diseases is realized. The abnormal sound report can include information such as the type, occurrence time, occurrence place and the like of the abnormal sound, so that a user can conveniently analyze and process the porcine respiratory disease.
In addition, the invention also provides a farm inspection robot which comprises a sound sensor for collecting pig audio, wherein the farm inspection robot is connected with a central processing unit and used for executing the steps of the method for monitoring the pig respiratory diseases according to any one of the above steps, and the inspection robot is used for monitoring, analyzing and processing the sound of pigs in different time periods and providing real-time monitoring and early warning functions.
The specific implementation steps are as follows:
step one: the inspection robot reaches the pig columns, and sound sensors are carried to monitor sound in each pig house.
Step two: the sound sensor transmits sound signals sent by pigs to the central processing unit.
Step three: the CPU processes and analyzes the sound signal and determines whether the sound signal belongs to abnormal sounds, such as coughs, sneezes, etc., by using an advanced audio recognition technology.
Step four: the CPU locates the abnormal sound information and records it in the data memory.
Step five: the CPU performs statistics and analysis on the recorded abnormal sound information to generate an abnormal sound report.
Step six: the abnormal sound report is output through the display, so that the timely early warning of diseases is realized.
The specific implementation principle and technical effect of the inspection robot are similar to those of the method embodiment, and details can be found in the related description in the above embodiment, which is not repeated here.
It should be noted that, in the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described as different from other embodiments, and identical and similar parts between the embodiments are all enough to be referred to each other. For the system embodiment and the robot embodiment, since they are substantially similar to the method embodiment, the description is relatively simple, and the relevant points are referred to in the description of the method embodiment.
Compared with the traditional monitoring mode of the breeding industry, the technical scheme of the invention has the following advantages:
1. the traditional monitoring mode needs to manually patrol the pig house, so that time consumption and labor cost are high, and the scheme adopts an automatic patrol robot, so that the labor cost can be saved, and the monitoring efficiency is improved;
2. the traditional monitoring mode mainly relies on manual observation of behaviors and physical conditions of pigs, omission and misjudgment are easy to occur, and the scheme adopts an advanced audio recognition technology, so that abnormal sounds emitted by the pigs can be accurately recognized and positioned;
3. the traditional monitoring mode can only realize intermittent monitoring of the health status of pigs, and the scheme adopts the functions of real-time monitoring and early warning, so that the health status of pigs can be monitored and early warned in time;
4. traditional monitoring mode needs professional to operate and analyze, and the operation of this scheme is simple easily understood, only need open inspection robot can carry out pig health status's monitoring, has reduced technical threshold.
Also, herein, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
While the foregoing description illustrates and describes the preferred embodiments of the present invention, it is to be understood that the invention is not limited to the forms disclosed herein, but is not to be construed as limited to other embodiments, but is capable of use in various other combinations, modifications and environments and is capable of changes or modifications within the scope of the inventive concept, either as described above or as a matter of skill or knowledge in the relevant art. And that modifications and variations which do not depart from the spirit and scope of the invention are intended to be within the scope of the appended claims.
Claims (10)
1. The method for monitoring the porcine respiratory disease is characterized by comprising the following steps of:
preprocessing the collected pig audio fragments, wherein the preprocessing comprises MFCC processing and Poisson attenuation processing to obtain acoustic feature vectors;
inputting the acoustic feature vector into a biological reasoning model, the biological reasoning model at least comprising: a first ResNet residual network for extracting vocal cord feature vectors, a second ResNet residual network for extracting respiratory tract feature vectors, a third ResNet residual network for extracting emotion feature vectors;
and carrying out integrated analysis on the vocal cord feature vector, the respiratory tract feature vector and the emotion feature vector through an aggregation model, and outputting a pig status diagnosis result and a biological latitude significance analysis result.
2. The method for monitoring porcine respiratory disease according to claim 1, wherein: the pig audio fragment is obtained by inspecting pig columns through an inspection robot, collecting pig sounds through a sound sensor when the pig audio fragment is close to the pig columns, and carrying out segmentation processing on the collected pig sounds; and recording the positioning information corresponding to the pig sound collecting position, or obtaining the corresponding positioning information by analyzing the sound source positioning of the pig sound.
3. The method for monitoring porcine respiratory disease according to claim 1, wherein: the Poisson attenuation treatment is to simulate and calculate the attenuation effect caused by muscle degradation of the acoustic feature vector through a Poisson attenuation template.
4. A method of monitoring porcine respiratory disease as claimed in claim 3, wherein: the attenuation effect simulation calculation is calculated by adopting a muscle degradation index function, and the index form of the muscle degradation index function is as follows:
where point Ix is a point obtained by the MFCC and λ is the average poisson mask of all values in the MFCC; poiss (x=k) is a specific poisson distribution form thereof, and represents the probability that the value of the random variable X is k; wherein X is a discrete random variable representing the number of events occurring within a certain time period or spatial region; x=k represents that the random variable X takes on a value of k, i.e., the number of times an event occurs is k.
5. The method for monitoring porcine respiratory disease according to claim 1, wherein: the first ResNet residual network, the second ResNet residual network and the third ResNet residual network are connected in parallel by using ResNet50 models, and 7X 2048 4-D tensor output layers of each ResNet50 model.
6. A method of monitoring porcine respiratory disease as claimed in claim 1 or 5, wherein: the aggregation model includes:
the pooling layer is used for pooling the vocal cord feature vector, the respiratory tract feature vector and the emotion feature vector;
the full-connection layer is used for carrying out intensive connection treatment on the pooling result, comparing the prediction result with a preset threshold value and outputting a classification result as the pig status diagnosis result; the full connection layer adopts a Sigmoid activation function, and the classification result comprises: the diagnosis result of the pig disease is that respiratory diseases exist or no respiratory diseases exist.
7. The method for monitoring porcine respiratory disease according to claim 1, wherein:
the vocal cord feature vector is characterized in that the ResNet model is used for detecting m-like pronunciation of English word 'Them' on collected pig audio fragments, and the vocal cord change of the pig is measured according to the pronunciation;
the respiratory tract feature vector is obtained by taking a data set containing similar sounds of a disease pig and a normal pig as training data, training a classification acoustic model by using transfer learning, and predicting respiratory tract feature changes according to the classification acoustic model;
the emotion feature vector is used for learning emotion features on the RAVDESS voice data set by training an entity voice classifier model, wherein the emotion feature vector comprises tones of organisms in 8 emotion states, and the tones comprise: neutral, calm, happy, sad, anger, fear, aversion and surprise.
8. The method for monitoring porcine respiratory disease according to claim 1, wherein: the biological latitude significance analysis result is displayed by adopting a biological latitude significance map, and each element of the biological latitude significance map is marked manually during training and automatically generated during prediction; the biological latitude saliency map display dimension includes three major aspects: the core index significance analysis result, the sensory flow data significance analysis result and the feature mark combination significance analysis result; wherein the significance analysis result of the core index further subdivides the significance analysis result of each audio segment; the results of the significance analysis of the sensory flow data further comprise: vocal cord variation analysis results, emotion variation analysis results, muscle degradation analysis results, and respiratory tract variation analysis results; the significance analysis results of the signature combination further include: pig morphological factor analysis result, significance factor analysis result, voting credibility analysis result and pig state factor analysis result.
9. A monitoring system for porcine respiratory disease comprising:
the audio preprocessing module is used for preprocessing the collected pig audio fragments, wherein the preprocessing comprises MFCC processing and Poisson attenuation processing to obtain acoustic feature vectors;
the feature processing module is used for inputting the acoustic feature vector into a biological reasoning model, and the biological reasoning model at least comprises: a first ResNet residual network for extracting vocal cord feature vectors, a second ResNet residual network for extracting respiratory tract feature vectors, a third ResNet residual network for extracting emotion feature vectors;
and the characteristic analysis module is used for carrying out integrated analysis on the vocal cord characteristic vector, the respiratory tract characteristic vector and the emotion characteristic vector through an aggregation model and outputting a pig state diagnosis result and a biological latitude significance analysis result.
10. A farm inspection robot comprising a sound sensor for collecting pig audio, characterized in that the farm inspection robot is connected with a central processor for performing the steps of the method for monitoring a pig respiratory disease according to any of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310746188.3A CN116486819A (en) | 2023-06-25 | 2023-06-25 | Pig respiratory disease monitoring method and system and farm inspection robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310746188.3A CN116486819A (en) | 2023-06-25 | 2023-06-25 | Pig respiratory disease monitoring method and system and farm inspection robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116486819A true CN116486819A (en) | 2023-07-25 |
Family
ID=87212313
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310746188.3A Pending CN116486819A (en) | 2023-06-25 | 2023-06-25 | Pig respiratory disease monitoring method and system and farm inspection robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116486819A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118072936A (en) * | 2023-12-19 | 2024-05-24 | 南京裕隆生物医学发展有限公司 | Intelligent sound health analysis system and method for disease type identification |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102009037097A1 (en) * | 2009-08-11 | 2011-02-17 | Kouemou, Guy Leonard, Dr. Ing. | Device for automatically monitoring respiration and ventilation of patient suffering from e.g. asthma, in hospital, has personal computer connected to sensor, and classification result fed to learning system for result confirmation |
| US20200245873A1 (en) * | 2015-06-14 | 2020-08-06 | Facense Ltd. | Detecting respiratory tract infection based on changes in coughing sounds |
| US20210319804A1 (en) * | 2020-04-01 | 2021-10-14 | University Of Washington | Systems and methods using neural networks to identify producers of health sounds |
| US20220338756A1 (en) * | 2020-11-02 | 2022-10-27 | Insubiq Inc. | System and method for automatic detection of disease-associated respiratory sounds |
-
2023
- 2023-06-25 CN CN202310746188.3A patent/CN116486819A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102009037097A1 (en) * | 2009-08-11 | 2011-02-17 | Kouemou, Guy Leonard, Dr. Ing. | Device for automatically monitoring respiration and ventilation of patient suffering from e.g. asthma, in hospital, has personal computer connected to sensor, and classification result fed to learning system for result confirmation |
| US20200245873A1 (en) * | 2015-06-14 | 2020-08-06 | Facense Ltd. | Detecting respiratory tract infection based on changes in coughing sounds |
| US20210319804A1 (en) * | 2020-04-01 | 2021-10-14 | University Of Washington | Systems and methods using neural networks to identify producers of health sounds |
| US20220338756A1 (en) * | 2020-11-02 | 2022-10-27 | Insubiq Inc. | System and method for automatic detection of disease-associated respiratory sounds |
Non-Patent Citations (1)
| Title |
|---|
| JORDI LAGUARTA 等: "COVID-19 Artificial Intelligence Diagnosis Using Only Cough Recordings", 《IEEE OPEN JOURNAL OF ENGINEERING IN MEDICINE AND BIOLOGY》, pages 275 - 281 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118072936A (en) * | 2023-12-19 | 2024-05-24 | 南京裕隆生物医学发展有限公司 | Intelligent sound health analysis system and method for disease type identification |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pramono et al. | A cough-based algorithm for automatic diagnosis of pertussis | |
| KR102630580B1 (en) | Cough sound analysis method using disease signature for respiratory disease diagnosis | |
| CN109258509B (en) | Intelligent monitoring system and method for abnormal sound of live pigs | |
| Cuan et al. | Automatic Newcastle disease detection using sound technology and deep learning method | |
| US8712740B2 (en) | Signal judgment method, signal judgment apparatus, program, and signal judgment system | |
| CN106580282A (en) | Human body health monitoring device, system and method | |
| CN106725532A (en) | Depression automatic evaluation system and method based on phonetic feature and machine learning | |
| Wei et al. | A real-time robot-based auxiliary system for risk evaluation of COVID-19 infection | |
| CN116486819A (en) | Pig respiratory disease monitoring method and system and farm inspection robot | |
| Tian et al. | Real-time behavioral recognition in dairy cows based on geomagnetism and acceleration information | |
| CN112037929A (en) | Classification method based on multi-modal machine learning, online new coronary pneumonia early warning model training method and early warning method | |
| US20240050006A1 (en) | System and method for prediction and control of attention deficit hyperactivity (adhd) disorders | |
| CN117034143B (en) | Distributed system fault diagnosis method and device based on machine learning | |
| Cong Phi Khanh et al. | The new design of cows' behavior classifier based on acceleration data and proposed feature set | |
| CN114299551A (en) | Model training method, animal behavior identification method, device and equipment | |
| CN114118219A (en) | Data-driven real-time abnormal detection method for health state of long-term power-on equipment | |
| Chen et al. | Improving the assessment of mild cognitive impairment in advanced age with a novel multi-feature automated speech and language analysis of verbal fluency | |
| Duan et al. | Short-term feeding behaviour sound classification method for sheep using LSTM networks | |
| CN115510901A (en) | Fault identification method and device for carrier roller of belt conveyor | |
| Sun et al. | Research on recognition method of broiler overlapping sounds based on random forest and confidence interval | |
| CN113842135A (en) | BilSTM-based sleep breathing abnormality automatic screening method | |
| Sun et al. | Research on white feather broiler health monitoring method based on sound detection and transfer learning | |
| KR102601051B1 (en) | Livestock disease real-time surveillance technology and intelligent quarantine and sanitation system | |
| CN117334337A (en) | Cancer patient pain intelligent evaluation and early warning system based on image recognition technology | |
| CN117349693A (en) | Axle fatigue crack acoustic emission signal identification method based on PSO-DBN |
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 |
Application publication date: 20230725 |
|
| RJ01 | Rejection of invention patent application after publication |