Research Of Birds Intelligent Recognition Method Based On Chirm

Birds are the most representative wildlife group of wetland, and an important part of thewetland ecosystem, but also important biological indicators on monitoring of wetlands’environmental quality. Determining the bird species provide an important basis to biodiversityand ecological balance of wetland. Like morphological features, bird chirm has uniqueness ofspecies, is important biological characteristics of birds, but also an important basis foridentifying birds.The paper aimed at the problems of the prominent contradictions in China’s economicdevelopment and environmental, and based on the analysis of the existing voice recognitiontechnology principle and system architecture, studied birds intelligent recognition methodbased on chirm, provided technical support for the realization of wetland bird monitoring andmigration parrerns. The main work and conclusions of the paper are as follows:(1) Based on analysis of advantages and disadvantages of the existing research results,and according to the characteristics of bird chirm,the paper proposed to deal with bird callsand bird songs separately, and used the design of double Gaussian mixture model.(2) Considering the the amount of sound samples, the rationality of areas, the differencesof bird subjects, the type of bird chirm and other factors, the paper selected brown-breastedbulbul, Chinese babax, daurian redstart, green-backed tit, gry-headed canary-flycatcher,red-billed leiothrix, red-rumped swallow, yellow-throated bunting, plain-tailed warbler, totalof eight bird species for the study, collect their sounds, and used glodwav software to removalnoise,cutted and other pretreatment.(3) Based on analysis of theoretical basis and implementation scheme of Mel FrequencyCepstrum Coefficient(MFCC), the paper improved MFCC, proposed a new characteristicparameter MFCCA, and did the experiment of correct recognition rate and efficiency. Theresults showed that, MFCCA had better flexibility and robustness and was suitable for birdchirm to extract characteristics.(4) According to the characteristics of birds chirm divided into call and song, the paperbroke the traditional of using single Gaussian Mixture Model(GMM), established bird callGMM and bird song GMM respectively for each bird, and did the experiment of correctrecognition rate and efficiency. The results showed that, when using dual GMM, recognitionresults were the best, and recognition efficiency was not affected. And discussed impact of using different orders of GMM, the results showed that, when the GMM’s order was32, therecognition was the best.(5) The paper selected50kinds of birds of Shaanxi Province, extracted MFCCA,established dual GMM, and designed software for testing, the results showed that, the correctrecognition rate of bird intelligent recognition method proposed in the paper could reach91.52%.