母羊发情声音数字化识别模型的建立

Establishment of Digital Recognition Model for Ewe Estrus Sound

本试验旨在通过构建模型,利用计算机技术识别母羊发情的叫声。对40只崂山奶山羊发情、羔羊寻母、母羊饥饿和饲料刺激4种类型声音进行录制,通过搭建程序对原始声音进行声音信号分析,然后将声音进行预处理并进行特征参数提取,发现母羊发情、饥饿、羔羊寻母状态下羊的声音强度和频率有显著差异。利用隐马尔可夫模型(HMM)、支持向量机(SVM)和迭代算法(Adaboost)3种声音模型对羊声音进行识别准确率验证,结果显示3种声音模型识别羊的发情状态与试情结果吻合率为96.67%、84.17%和87.92%。从动物行为学的角度分析,可以利用计算机处理羊的声音识别出母羊发情等状态。

This experiment aims to identify the estrus of the ewes by constructing a model using computer technology.Four types of the sound,including coming from estrus,lamb looking for mother,ewes and feed stimuli,from 40 Laoshan dairy goats were recorded..The sound signal was analyzed by setup program,and then the sound was preprocessed and the feature parameters were extracted.It was found that there were significant differences in sound intensity and frequency of ewes in estrus,hunger and seeking ewes.Models trained using HMM,SVM and Adaboost three sound models can be verified by testing the recognition accuracy.The results showed that the coincidence rate of sound recognition of estrusestrus state and estrus test were 96.67%,84.17%and 87.92,respectively.According to the analysis of animal behavior,we can use the computer to process the sound of the goat to identify the status of the ewes.