| Largemouth bass(Micropterus salmoides),commonly known as California perch,is the main species of freshwater perch breeding in China,with the largest breeding area and the largest breeding scale.Largemouth bass has the advantages of short growth cycle,easy to catch,solid meat,delicious meat,fragrance,rich nutrition,no muscle thorns,so it is deeply loved by farmers and consumers.According to the data of the 2022 Fishery Statistical Yearbook,the output of freshwater perch increased from619,500 tons in 20 years to 702,100 tons,an increase of more than 13%,making it the fastest growing species with the best economic benefits in recent years.According to the cost analysis of the industry report,the feed cost accounts for 67.35% of the total cost.Although the feeding method has been upgraded from manual feeding to mechanical feeding,reducing the feed cost within the feasible scope is still the driving force for the continuous development of intelligent farming.In order to break through the technical bottleneck of intelligent feeding system,passive acoustics was used to study the feeding behavior of cultured fish.However,at present,the study on the feeding acoustic signals of cultured fish is still in the initial stage.Because the feeding acoustic signals in groups are chaotic,it is not conducive to the analysis and extraction of acoustic characteristics,and the relationship between the acoustic characteristic parameters of group feeding and feeding activity is complicated.Therefore,this study mainly focuses on the characteristics and types of feeding acoustic signals,and uses passive acoustic techniques to obtain feeding acoustic signals of largemouth bass,starting from the monomeric and small-size breeding densities.The feeding acoustic signals of fish were extracted from the mixed signals,and the acoustic characteristic parameters of largemouth bass were screened,and different types of acoustic signals during feeding were classified according to such acoustic characteristic parameters,so as to quantify the feeding activity under different breeding densities.In this study,hydrophones were used to collect the complete feeding acoustic signals of largemouth bass during feeding.Then,according to the synchronous audio and video records during feeding,behaviors corresponding to the acoustic signals generated by largemouth bass during feeding were determined and labeled.The required acoustic signal types are extracted and analyzed.The research mainly uses speech signal processing methods for reference to mark them.The required signal types are extracted and analyzed.This study mainly uses speech signal processing methods to process and analyze feeding sounds,and screens acoustic parameters that can be used to distinguish feeding signal types.To study the difference of feeding acoustic signal under different body size and culture density.Different size specifications and different breeding density were set respectively.The body size of about 25 cm were selected to analyze the individual sound signal,to explore the composition and stability of the feeding sound signal.Four small breeding density groups of 4,8,12,and 16,respectively,were set up to analyze the differences of feeding signals of largemouth bass at different densities and the acoustic parameters that could be used for identification.The main research work of this paper is as follows:1.In the preprocessing part of largemouth bass feeding sound,the collected acoustic signals are firstly converted into spectral graphs by Fourier transform;There is a large amount of background noise in the original signal.According to the collection of background noise 5 min before sampling,the acoustic characteristics of the background noise are analyzed.After frame division and window addition,spectral subtraction,Wiener filter and subspace speech enhancement method are compared to obtain the optimal feeding signal noise reduction method.2.In the study of largemouth bass feeding acoustics,it was found that the feeding acoustic signals of largemouth bass were mainly classified into swallowing signals and chewing signals.The swallowing signal was the signal generated by the largemmouth bass when they swallowed the feed.According to the video of the sound and image synchronization,the "squeak" sound emitted after swallowing the feed was the chewing sound.Four acoustic features(short-time average energy,average Mel cepstrum coefficient,main peak value and main peak frequency)of the two signals were extracted and combined to establish the feeding sound classification model of largemmouth bass.3.In the feeding acoustic study of the population largemmouth bass,it was found that the frequency spectrum of feeding sound and background noise could be distinguished from 1-20 k Hz at different breeding densities.The frequency of the main peak of the power spectrum of feeding sound was in the range of 1.2-3 k Hz,and the main peak of the power spectrum was positively correlated with breeding density.In the time domain analysis,the short-time mean amplitude and short-time mean zero crossing rate could be used as the characteristic parameters for the recognition of the swallowing and chewing process signals of largemmouth bass.Spectral entropy and power spectrum could be used to identify the feeding stages of largemouth bass at different breeding densities... |
PDF Full Text Request