摘要
针对音频数据库中存在的问题 ,提出了一种基于索引的变换 .分析了离散余弦变换的特点 ,得出Ⅱ型和Ⅲ型快速离散余弦变换算法 ,并把这两种算法用于音频数据索引特征的抽取和音频信号的重构 .所提算法具有搜索速度快、回取精度高的特点 ,同时也使得音频数据的索引对噪声不敏感 ,与原数据搜索相比 ,具有更高的成功率 .通过对峰值信号噪音率和回取精度两个指标的评估 ,验证了这种方法对加快音频数据的搜索速度和提高回取精度的有效性 ,为音频数据自动分析和分类、基于内容的数据索引和查询、基于近似的搜索提供了快速而有效的手段 .
A transformation based on index was developed for the audio database. The attribute of the discrete cosine transform (DCT) were analyzed and the type-II and the type-III of the fast discrete cosine transform (FDCT-II and FDCT-III) were derived. The FDCTs were used to extract attribute of the audio data index and to reconstruct the data signal. Proposed arithmetic was provided with fast speed in search and high retrieval precision. At the same time, it also made the index of audio data insensitive to the noise. Compared with the original data index, the proposed arithmetic has higher success rate. A test was done with the new method and showed the validity of picking up the search of the audio data and improving the retrieval precision by evaluating two targets of the peak signal noise radio (PSNR) and the retrieval precision. This method also offers fast and effective measures for audio data automatic analysis and classification, content-based data index and query, and approximate-based search.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2001年第8期854-857,共4页
Journal of Xi'an Jiaotong University
关键词
音频数据
离散余弦变换
特征抽取
数据库
Audio systems
Database systems
Discrete Fourier transforms
Information retrieval
