摘要
介绍了一种基于Mel倒谱参数和EarthMover'sDistance(EMD)度量的音频检索方法,它利用了人耳的感知特性,并结合特征分布之间的距离度量EMD,获得了良好的性能,具有广阔的发展前景。
An audio retrieval theory using Mel-frequency cepstral coefficients (MFCC) and the Earth Mover's Distance (EMD) is presented. It applies perceptual feature and a novel measurement of signatures. Results show that good performance is achieved,which gives the wide prospect of this approach.
出处
《电声技术》
2005年第7期48-51,共4页
Audio Engineering
关键词
音频检索
EMD
Mel倒谱参数
K-均值聚类
audio retrieval
EMD
Mel-frequency cepstral coefficients
K-Means clustering