Design Of Pig Cough Monitoring System In Fattening Pig Houses

In the process of the pig breeding industry’s transformation from traditional retail farming to large-scale farming,in order to pursue economic benefits,farmers have reduced the breeding area of a single pig and at the same time increased the risk of pig diseases.Since the main symptom of common pig respiratory diseases is cough,this article aims to monitor the coughing of pigs in pig houses through intelligent information processing and voice recognition technology,and improve the efficiency of diagnosis and treatment of pig respiratory diseases in pig farms.First of all,with the help of the pig farmer,data collection of pig calls in the fattening pig barn is carried out to obtain the original pig call audio.After screening and manual segmentation,the pig call endpoint detection data set and pig cough sound recognition data set were produced.One sample in the endpoint detection data set contains multiple and multiple pig calls,and the cough sound recognition data set includes three types of sounds: pig humming,howling,and coughing.Through preprocessing and feature extraction of samples,feature parameters for subsequent model research are obtained.Secondly,in view of the problem that the fan noise in the finishing pig house will affect the detection effect of endpoint detection,this paper proposes a pig cry endpoint detection model based on LSTM,which is different from the dual-threshold endpoint detection and SVM endpoint detection.Compare the detection effect under fan noise,and then objectively measure the performance of the model based on the detection accuracy,missed detection rate and false detection rate.Experimental results show that,compared to the other two endpoint detections,the pig cry endpoint detection model based on LSTM has a higher detection accuracy in a fan noise environment,and as the fan noise intensity increases,the accuracy fluctuation range is smaller,which verifies this model has good robustness.Thirdly,since the samples will be divided into samples of different lengths after endpoint detection,this paper adopts the RNN structure model with unlimited input length of the sample to be tested in the pig cough recognition algorithm.By comparing the characteristics of MFCC and MFSC,which are commonly used in voice recognition,on the classification and recognition of pigs’ humming,howling and coughing,it explores the acoustic features with higher discrimination and better noise immunity for the three kinds of pig calls.In addition,by comparing the performance of RNN and its variant structure on the pig cough sound recognition task under fan noise,the recognition rate,false recognition rate and accuracy rate of the pig cough sound are used as the evaluation criteria of the model,and add attention mechanism to optimize the model.Finally choose a good structure which the classification recognition effect is better to builds the pig cough sound recognition model.The experimental results show that the pig cough sound recognition model based on AM-GRU and MFCC performs best in the task.Finally,this paper deploys two models on the basis of pig cry endpoint detection and pig cough sound recognition,and according to the actual needs of the pig farm,designs and implements a web application-based pig cough sound monitoring system.The system realizes the functions of user management,pig cough sound detection,cough record and historical operation query and display,and provides help for breeders to understand pig cough conditions in fattening pig houses.