数字音乐质量软件噪声监测及合规性检测模块开发

发布时间：2018-05-27 15:31

本文选题：数字音乐 + 质量检测　；参考：《电子科技大学》2014年硕士论文

【摘要】：随着近年来电信运营商在彩铃、无线音乐等增值业务的高速发展,对数字音乐数据库中日益增长的海量音乐文件的质量保障需求越来越迫切。电信运营商为了解决数字音乐质量问题,在数字音乐的制作、发布和播放制定了标准和流程。但如何快速自动化的处理数字音乐库中数以百万计的音乐文件,并进行完备的质量检测、监督和预测,成了运营商提升业务品质和用户体验的一个非常重要和紧急的问题。本论文将上述问题作为出发点和需求项,以数字音乐质量检测软件系统的噪声监测及合规性监测模块的开发为课题,在音乐文件特征快速提取技术、质量预判技术,合规性检测技术及音乐质量软件实现等方面进行了详细的讨论和研究,主要内容为:1.提出了一种基于特征码分析的快速提取方法。该方法是分别对转换前后的音乐文件所包含的幅度、过零率、共振峰等特征码进行提取和数学建模,然后通过特征码匹配的方式来判断音乐在转换前后是不是出现质量问题。该方法是数字音乐质量软件的分析基础。2.研究音乐文件合规性检测方法,提出了一种基于显式特征分析的合规性检测方法。该方法对音乐文件的格式、位数、码率、采样率、单双声道等显式特征进行快速提取,然后查询规格范围数据库进行量度匹配检测,由此快速发现音乐文件不符合规格的质量问题。该方法是数字音乐质量软件的过滤器,能快速筛查音乐文件显式质量问题,提升检测效率。3.研究噪声检测方法,提出了一种基于无监督分类和最小风险率贝叶斯算法相结合的对音频文件进行质量判定的方法。该方法采用无监督分类的方法对音频文件进行聚类,并反复迭代运算使分类达到的合理性。再利用贝叶斯条件概率算法确定和修正音乐文件中包含噪声的概率,由此确定待测文件出现噪声质量问题的概率。该方法是数字音乐质量软件的核心和难点。4.提出数字音乐质量软件系统的架构和实现方案,并详细论述了各个功能模块的细节和实现方法。本论文提出的主要技术方法在数字音乐质量软件中得到了很好的实现和测试,实践证明能有效的应对电信运营商对数字音乐库的质量管理需求。
[Abstract]:With the rapid development of telecom operators' value-added services such as color ring tone and wireless music in recent years, it is more and more urgent to ensure the quality of the mass music files in the digital music database. In order to solve the problem of digital music quality, telecom operators have established standards and procedures for making, publishing and playing digital music. But how to deal with millions of music files in the digital music library quickly and automatically, and how to inspect, supervise and predict the quality of the digital music library has become a very important and urgent problem for the operators to improve the service quality and user experience. This paper regards the above questions as the starting point and demand item, taking the development of the noise monitoring and compliance monitoring module of the digital music quality detection software system as the subject, the technology of fast extraction of the music file features and the quality prediction technology. The technology of compliance testing and the realization of music quality software are discussed and studied in detail, the main content is: 1. 1. A fast extraction method based on signature analysis is proposed. In this method, the amplitude, zero crossing rate and resonance peak of the music file before and after the conversion are extracted and modeled, and then the quality of the music before and after the conversion is judged by the method of signature matching. This method is the analysis foundation of digital music quality software. Based on explicit feature analysis, a method for detecting the compliance of music files is proposed. In this method, the format, bit number, bit rate, sampling rate, mono-dual channel and other explicit features of the music file are quickly extracted, and then the measurement matching detection is carried out by querying the specification range database. This quickly found that the music file does not conform to the quality of the specification. This method is a filter of digital music quality software, which can quickly screen the explicit quality problems of music files and improve the detection efficiency. Based on unsupervised classification and Bayesian algorithm of minimum risk rate, a method for quality determination of audio files is proposed. The method uses unsupervised classification to cluster audio files and iterates repeatedly to make the classification reasonable. Then the Bayesian conditional probability algorithm is used to determine and correct the probability of the noise in the music file, so as to determine the probability of the noise quality problem in the file to be tested. This method is the core and difficulty of digital music quality software. The architecture and implementation of the digital music quality software system are presented, and the details and implementation methods of each function module are discussed in detail. The main technical methods proposed in this paper have been well implemented and tested in the digital music quality software, and the practice has proved that it can effectively meet the quality management requirements of digital music library of telecom operators.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN912.3

【参考文献】