基于隐含语义分析的音乐检索  

Music retrieval based on latent semantic analysis

作　　者：周皓东[1] 刘炜[1] 

机构地区：[1]江西省科学院高科技开发中心,江西南昌330029

出　　处：《计算机工程与设计》2013年第6期2071-2074,2146,共5页Computer Engineering and Design

基　　金：江西省科学院科研开发(青年创新)基金项目(2010-YQC-03)

摘　　要：为提高音乐检索效率,使检索结果与搜索目的更接近,提出了基于隐含语义分析的音乐检索方法。将曲谱表示为标准音符和音转的交替串,基于每个交替串使用频率高于包含它的多交替串排列的事实,设计了音乐词汇统计算法。为使各分句能整齐地转化为相同维数的向量,使用最长的分句长度作为标准维数,基于增加频率和的原则进行单词的重新分割。实验结果表明,基于隐含语义分析的检索能获得令人满意的检索结果。To improve the efficiency and find the music nearest to search purpose, a music retrieval method based on latent semantic analysis is proposed. Firstly, the scores are expressed as alternant string of standard notes and tones string, and then based on the fact that the use frequency of each alternant string is higher than any one including it, statistical algorithm is designed for music vocabulary. For each clause can neatly converted to vectors of the same number of dimensions, using the longest clause length as the number of standard dimensions. Based on the principle to increase frequency, the words are re-segmented. The experiments show that the new algorithm can obtain satisfactory retrieval results.

关 键 词：音乐排序 音乐检索 检索效率 隐含语义分析 词汇统计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]