摘要
对于仓储害虫的种类识别问题,提出了一种根据害虫产生的声音来进行基于支持向量机的识别算法.利用已知害虫的声音样本,经过语音预处理、特征提取,用所提取的语音特征向量构造多个支持向量机(SVMs).在识别应用中,在无法看到害虫的情况下,利用传感器采集到的害虫声音样本,预处理之后,利用训练好的支持向量机来判断是什么害虫,从而采取合适的灭虫措施.
An approach based on voice recognition using support vector machines (SVMs) was proposed for stored-product insect recognition. Adaline adaptive noise canceller was used as voice preprocessing unit. Feature vectors were extracted from preprocessed audio signals of known insect samples, and used to train multiply SVMs for insect recognition. The operation is convenient, requiring the insect′s audio signals collected by sensors without insect images or samples.
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2005年第2期34-36,共3页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(79970025)
国家科技型中小企业创新基金资助项目(01C26214200913).
关键词
仓储害虫
声音识别
支持向量机
特征提取
降噪
stored-product insect
voice recognition
support vector machines
feature extraction
de-noise