Design On The Porcine Abnormal Sound Recognition And Positioning System

With the deepening of the reform and opening process,the rapid popularization of industrial automation and intelligence,intelligent livestock farming has also been greatly developed.Large-scale pig farms have gradually become the mainstream.Audio monitoring is a common technique in large-scale farming.When pigs suffer from influenza or bacterial infections,they will cough;when pigs suffer from pseudorabies or are squeezed,they will startle.When porcine lives are threatened,they will also startle when they are squeezed.By monitoring the abnormal sounds of pigs,sick pigs or threatened pigs can be found as early as possible,so that they can be treated in time to reduce losses.In order to find and locate pigs with abnormal sounds in time,this paper proposes a method to identify and locate the abnormal sounds of pigs.The main research contents and results are as follows:(1)Using field programmable gate array to design a parallel collection system of multiple channels microphones,the microphone array is used to collect porcine sound signals.The collection system has the advantages of low cost and good effect.On the basis of the stable acquisition of sound signals by the microphone array,digital filter,wiener filter and spectrum subtraction method are used to reduce the noise of the collected sound,and then spectrum subtraction method is selected for noise reduction after comparative analysis,which has the advantages of high output signal-to-noise ratio and low root mean square error.(2)By extracting the linear prediction cepstral coefficient,mel frequency cepstrum coeftieient and formant characteristic parameters of the porcine sound signal,the support vector data description recognition model for porcine abnormal sound is trained,and the whale optimization Algorithm is introduced to optimize the model parameters.Combining different features to obtain combined features,the porcine recognition model with combination characteristics of the mel frequency cepstrum coefficient and its first order difference and the formant has a higher recognition accuracy.The spatial characteristics of the microphone array are used to synthesize the recognition results of multiple microphones,which further improves the accuracy ofabnormal sound recognition of pigs.(3)The generalized cross-correlation algorithm based on time delay estimation is used to calculate the time difference of arrival between the signals of each element of the microphone array.The cross-correlation,Roth weighting,smoothed coherence transform weighting and phrase transform weighted cross-correlation algorithms are compared.Finally,the phrase transform weighted cross-correlation algorithm with small calculation amount and low error value is selected.The analysis and comparison between the geometric positioning method and the search positioning method show that the geometric positioning algorithm has a larger error when the time difference of arrival is smaller.Although the search positioning method has a large amount of calculation,its accuracy is high.In this paper,the sound source localization strategy of Phrase Transform weighted cross-correlation algorithm combined with search localization is finally selected.(4)The upper system for porcine abnormal sound recognition and positioning with good human-computer interaction machine interface is designed.The system can display,record and alarm the abnormal situation of porcine sound,when the porcine abnormal sound is found,it can display the positioning result of porcine abnormal sound,timely feedback to the feeders,so that the feeders can quickly deal with the abnormal situation.In this paper,through the study of porcine abnormal sound recognition and localization methods,we present a strategy for identifying and locating porcine abnormal sounds in small pig houses.The combined feature parameters are extracted,the trained support vector data description recognition model is used to identify porcine abnormal sound,and the total recognition rate is up to93.70% based on the recognition results of multiple channels.When porcine abnormal sounds are recognized,the average distance error is within 1m by using the Phase Transformation weighted cross-correlation algorithm and the search positioning method.The results show the method proposed in this paper can provide certain reference value for the modern welfare breeding of pigs.