The Abnormal Recognition Method Of Ecological Breeding Chicken Voice Based On Sparse Representation

With the development of economy and the improvement of the quality of life,more and more people are in favor of ecological breeding chickens.But in the ecological breeding environment,breeders are often unable to ensure whether the chicken flocks are safe and healthy.Although the videos and images of ecological breeding chickens can record the activities of chickens and detect abnormal behavior,the method is easy to be influenced by the light of day and night and wheather.Fortunately,the sound is not affected by light and hides a large number of emotional information,and it is easy to access and process the sound signals.Currently the voice recognition of birds which are in the ecological breeding environment has achieved good results.In view of the fact that birds and chickens belong to the aves,so we can imitate bird voice recognition method to deal with the ecological breeding chickenvoice.On the other hand,since Mallat and Zhang first proposed the idea of sparse representation,it has attracted the attention of scholars and has become a research hotspot in recent years,and it has achieved good results in the application of voice recognition.Ecological breeding chicken voice signals are inevitably mingled with the sound of rain,wind or running water and so on,so the difficulty of recognition lies in how to denoise noisy chicken voice.In this paper,we introduce the idea of sparse representation to improve the noise reduction effect.In addition,we also introduce the method to signal reconstruction,which makes the chicken voice signals more reflect the essential characteristics,so as to reduce the processing cost and improve the efficiency.The key is how to select the voice features,the solution in this paper is to serially fuse three kinds of traditional voice features(including MFCC,LPC and PNCC)with the time frequency parameters.The main contents of this paper are as follows:1.Genetic Algorithm(GA)is used to optimize Orthogonal Matching Pursuit Algorithm(OMPA),GA-OMPA for short.OMPA is a kind of important sparse representation methods,which is often used in voice recognition.But OMPA has limitations,so we use GA to simulate the process of natural evolution to search for the best matching atom of OMPA in order to improve its performance.2.GA-OMPA is used to reduce the noise of chicken voice.In this paper,we choose a multi-band spectral subtraction method as a contrast experiment,because it compensates for the defect of the spectral subtraction method.The method can not only reduce the residual noise,but also avoid the voice distortion.Experimental results show that the algorithm can get a higher recognition rate after noise reduction using the sparse representation algorithm.3.GA-OMPA is used to reconstruct the signal of chicken voice.In this paper,the proposed method is used to reconstruct the chicken voice signals after noise reduction,and to keep the main signal structures,which are the most relevant to the original voice signals.We will get better performance by reconstructing the chicken voice signals.4.The time frequency parameters of Gabor atoms of each voice signal are serially fused with MFCC,LPC and PNCC respectively,which are the three kinds of traditional voice features.And it can achieve higher recognition rate than before fusion.In the last chapter,we summarize and analyze the whole article and point out the shortcomings and new perspectives.