Research On Pig Audio Recognition Based On Audio Compression Transmission And Endpoint Detection

The problems of poor ventilation,small space and limited management in intensive breeding of live pigs lead to the frequent occurrence of disease and difficult observation of various habits of live pigs,and the management is time-consuming and labour-consuming,how to monitor and quickly judge the habits and possible diseases of live pigs has become an urgent problem.The live pig’s squeal contains rich biological information.The analysis and research of the live pig’s squeal can further understand the physiological and behavioral mechanism of the live pig,and help to improve the diagnosis and treatment efficiency of the disease of the live pig,to understand the behavior and individual characteristics of pig population.Therefore,it is very important to monitor and identify the live pig’s voice in real time.In this paper,the external microphone is used to build the development board equipment to collect,improve and optimize the transmission of the node’s live pig sound,and the transfer program is used to transfer the reception,after preprocessing,the signal is detected by an improved EMD-TEO(Teager energy operator based on empirical mode decomposition)cepstrum distance endpoint,then the Mel Cepstrum Coefficient(MFCC)is extracted and the Hidden Markov Model(HMM)is studied,this paper optimizes the pig audio recognition system and explores a new way to identify the pig voice using the voice recognition technology.The main work is introduced from the following aspects:(1)Aiming at the acquisition and transmission of the pig audio signal,by using the compression sensing concept and three core compression sensing knowledge,the system is applied to the data acquisition node of the Development Board,and the corresponding hardware and software are designed.The results show that the reconstruction rate is increased by 20.75%,the reconstruction time is shortened by0.20 s,and the Peak signal-to-noise ratio is increased by 1.50 d B compared with the Peak signal-to-noise ratio.The invention reduces the information transmission amount and saves the time of signal processing,storage and transmission.Then,the transfer program and interface are designed to send the pig audio signal to the server and to download and record it in real time through the client.Finally,the established pig audio signal recognition system is called.(2)Aiming at the problems of poor anti-noise ability and low accuracy of traditional endpoint detection methods in pig audio signal recognition,an endpoint detection method which improves the cepstrum distance of EMD-TEO is applied to pig audio signal endpoint detection,the two characteristic parameters of end-point detection(Teager energy parameter and cepstrum distance parameter)of pig audio signal are studied.The results showed that when the Signal-noise ratio(SNR)of the Additive white Gaussian noise WAS-5 d B and 10 d B,and the SNR of the background noise was 10 d B,compared with EMD-TEO and cepstrum distance endpoint detection,the improved algorithm is better than the other two algorithms in the accuracy of end-point detection of pig audio signal,which is 84.725%,94.281% and 90.293%,respectively,and the improved algorithm has good endpoint detection effect for different types of live pig audio signal,and has a certain robustness.(3)The audio recordings of live pigs downloaded and recorded on the above platform were classified into 10 categories,including the sound numbers of live pigs(bellowing,oestrus,eating and getting sick,etc.).Each category was divided into 30 groups,of which 25 groups were used as training samples,another 5 groups as a test set.A pig audio signal recognition system based on HMM is designed by using Matlab,and the effect of the improved algorithm on the performance of the recognition system is compared with the traditional method,the improved algorithm improves the recognition rate of the recognition system,and fully reflects the effectiveness and feasibility of the improved algorithm.Finally,a pig audio signal recognition system based on HMM is built on Matlab Platform,and the GUI function is used to demonstrate each part of the improved recognition system.At last,the recognition effect is compared and analyzed,verify the effectiveness of the improved method.