摘要
通过储粮害虫声信号判断害虫活动情况对安全储粮意义重大。本研究采集了2种储粮害虫的活动声信号,首先提取声信号梅尔倒谱系数(MFCC)特征,然后以特征数据建立高斯混合模型(GMM),最后使用聚类方法对2种储粮害虫的4种活动声信号进行识别,识别率均达到80%以上。本研究验证了声检测法识别储粮害虫的可行性和有效性,具有较大的实际应用价值。
It is significant to estimate the pests in stored grain.The acoustical signals of two kinds of stored grain pests were collected in this research.Firstly,Mel Frequency Cepstrum Coefficients（MFCC）of the signals were extracted.Then Gaussian Mixture Models（GMM）were structured.Finally recognition for four varieties acoustical signals was done by clustering algorithm.The results showed that the recognition rates of all kinds of acoustical signals were above 80%.The research verifies the feasibility and availability for recognizing stored grain pests by acoustical detection method,and has good practical applicative prospect.
出处
《南京农业大学学报》
CAS
CSCD
北大核心
2012年第6期44-48,共5页
Journal of Nanjing Agricultural University
基金
国家自然科学基金项目(10974130)
陕西省教育厅科研计划项目(11JK0519)
关键词
高斯混合模型
梅尔倒谱系数
聚类
储粮害虫
声检测
Gaussian Mixture Model（GMM）
Mel Frequency Cepstrum Coefficients（MFCC）
clustering
stored grain pest
acoustical detection
