摘要
为利用生态环境中各种声音包含的信息,提出一种将流形学习算法和支持向量机(SVM)相结合的生态环境声音分类技术。提取音频强度、音色、音调和音频节奏的特征集合并计算对应的特征向量,采用改进的拉普拉斯特征映射流形学习算法对特征向量进行维数约简,从而降低数据处理的复杂性。使用SVM对降维后的特征向量进行分类,发挥SVM在处理小样本、非线性及高维数据方面的优势,从而提高分类准确率。实验结果表明,该技术能对生态环境声音进行快速准确的分类。
In order to take full advantage of the information contained in the eco-environmental sounds,this paper presents a ecological environmental sounds classification technology based on manifold learning algorithm and Support Vector Machine（SVM）.Select four different kinds of audio characteristics those are dynamics,timbre,pitch and rhythm and then calculate the feature vectors corresponding to those four audio characteristics.So as to reduce the complexity of data processing,it makes use of an improved Laplacian feature mapping for dimensionality reduction.To improve the accuracy,the SVM classifier is used to classify the dimension-reduced feature vectors because SVM have advantages in dealing with the data that is of few samples,nonlinear and high dimension.Experimental results show that the technology can be used to classify ecological environmental sounds quickly and accurately.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第7期288-290,共3页
Computer Engineering
基金
国家自然科学基金资助项目(61075022)
福建省教育厅A类科技基金资助项目(JA09021)
关键词
生态环境声音分类
流形学习
支持向量机
ecological environmental sound classification
manifold learning
Support Vector Machine（SVM）
