谢抢来 杨威

摘  要： 为从海量音频信息中快速准确地获取所需文件，研究基于FP_Growth关联规则挖掘的多轨道数字音频文件分类方法。以数据立方体下具有多维属性的多轨道数字音频文件为研究对象，采用FP_Growth关联规则挖掘算法排序多轨道数字音频文件数据集的频繁1?相集，采用FP tree挖掘频繁项集获取多维关联规则集。识别并提取多轨道数字音频文件中某类型音频文件中的音頻内容，使用删除停用词结合TF/IDF的计算方式获取清洁的音频文件，应用音频文件数据集的多维关联规则集，综合考虑匹配规则数和置信度，搜索规则集，获取该类型音频文件最相配类别，完成多轨道数字音频文件分类。实验结果表明，所提方法能够直观有效地分类多轨道数字音频文件，且分类结果准确率和召回率平均值分别达到96.60%和96.54%，显著高于对比方法。

关键词： 关联规则; 数据挖掘; 多轨道; 数字音频; 文件分类; 数据立方体

中图分类号： TN911.72?34; TP301.6                  文献标识码： A                文章编号： 1004?373X（2020）01?0179?04

Research on multi?track digital audio file classification

based on FP_Growth association rule mining

XIE Qianglai， YANG Wei

Abstract： A multi?track digital audio file classification method based on FP_Growth association rule mining is studied to get the required files quickly and accurately from the massive audio information. By taking multi?track digital audio files with multidimensional attributes under the data cube as the research object， FP_Growth association rules mining algorithm is adopted to sort the frequent 1?phase sets of multi?track digital audio file data sets， and FP tree mining frequent item sets are adopted to obtain multidimensional association rules. The audio content in a certain type of audio file in multi?track digital audio files is identified and extracted， the calculation method of deleting stop words combined with TF/IDF is adopted to obtain clean audio files. The multidimensional association rules set of audio file data set is applied to take into account comprehensively the matching rule number and degree of confidence， search the rule sets， obtain the most suitable type of audio file categories and complete the classification of multi?track digital audio files. The experimental results show that the proposed method can intuitively and effectively classify multi?track digital audio files， and the averages of accuracy and recall rate of classification result reach 96.60% and 96.54% respectively， significantly higher than the comparison method.

Keywords： association rule; data mining; multi?track; digital audio; file classification; data cube

0  引  言

不同类型音频信号经采集、量化编码压缩、多轨录入等数字化处理后，以文件形式存储在计算机硬盘等存储媒体内，最终形成多轨道数字音频文件。从多轨道数字音频文件中快速准确地获取所需类型音频文件，在视频编辑、影像分析等领域具有重要的应用价值[1]。

数据挖掘是挖掘并分析项集之间某种潜在关系，这种潜在关系称为关联规则，关联规则可划分为维间、单维、多维、量化等不同类型[2]。由于多……