Research On The Monitoring Method Of Colony Healthy State Based On Sound Signal

China is the world's largest producer of honey,and its demand for honey is among the world's highest.In the bee farming industry,the health of bees has a direct impact on the production of honey and is directly related to the economic benefits of bee farmers,so the health of bees has been widely concerned by beekeepers.In recent years,with the rapid development of science and technology,bee breeding industry has also added a variety of new technologies,which have injected new vitality into the traditional beekeeping industry.However,the rapid identification of the health status of honeybees and the alert of the abnormal status of honeybees in advance are still the most urgent and important problems to be solved in the intensive bee breeding industry.At present,the domestic monitoring of bee health mainly relies on experienced beekeepers.This method is not objective enough,and human subjective factors account for a large proportion.Meanwhile,the invasive inspection of beehive activities by the beekeepers will also cause a strong reaction from the bee colony,which has a great impact on the honeybee yield cycle.Audio recognition technology has the characteristics of high recognition accuracy,no direct contact with bees,strong objectivity and so on.In this paper,the audio recognition technology is used to collect and analyze the various states of the bees,and finally the state recognition rate of the corresponding bees is obtained.The main research contents of this paper are as follows:1.The bee colony sound recording system is constructed to obtain the bee colony sound signal.The system included power supply module,microphone,DSpic,RPI2 and other modules.The bee colony sound sampling rate was set to 44100 Hz and the storage format was wav.2.There is a lot of ambient noise in the sounds of the bees collected from the hive.In this paper,spectrum subtraction,wavelet threshold function and adaptive filter cancellation method are used to reduce the noise of the acquired sound signal.The experimental results show that the adaptive filter cancellation method can effectively remove the noise of bees.3.Four characteristic parameters were extracted from the sound signal of bee colony: time-domain characteristic parameters,LPCC and its first-order difference characteristic parameters,MFCC and its first-order difference characteristic parameters.The results show that the sound recognition effect is best when MFCC and its first-order difference coefficient are used as the characteristic parameters.4.Using the single classification detection capability of SVDD model,a multifunctional classifier suitable for SVM model structure was constructed,and SVDD and SVM detection recognition models were obtained.The model is used to detect and identify the audio signal of abnormal state of bees,and the SVM model can effectively identify the abnormal state of bees.