本文選題：數(shù)字音樂(lè) + 質(zhì)量檢測(cè)��； 參考：《電子科技大學(xué)》2014年碩士論文

【摘要】：隨著近年來(lái)電信運(yùn)營(yíng)商在彩鈴、無(wú)線音樂(lè)等增值業(yè)務(wù)的高速發(fā)展,對(duì)數(shù)字音樂(lè)數(shù)據(jù)庫(kù)中日益增長(zhǎng)的海量音樂(lè)文件的質(zhì)量保障需求越來(lái)越迫切。電信運(yùn)營(yíng)商為了解決數(shù)字音樂(lè)質(zhì)量問(wèn)題,在數(shù)字音樂(lè)的制作、發(fā)布和播放制定了標(biāo)準(zhǔn)和流程。但如何快速自動(dòng)化的處理數(shù)字音樂(lè)庫(kù)中數(shù)以百萬(wàn)計(jì)的音樂(lè)文件,并進(jìn)行完備的質(zhì)量檢測(cè)、監(jiān)督和預(yù)測(cè),成了運(yùn)營(yíng)商提升業(yè)務(wù)品質(zhì)和用戶(hù)體驗(yàn)的一個(gè)非常重要和緊急的問(wèn)題。本論文將上述問(wèn)題作為出發(fā)點(diǎn)和需求項(xiàng),以數(shù)字音樂(lè)質(zhì)量檢測(cè)軟件系統(tǒng)的噪聲監(jiān)測(cè)及合規(guī)性監(jiān)測(cè)模塊的開(kāi)發(fā)為課題,在音樂(lè)文件特征快速提取技術(shù)、質(zhì)量預(yù)判技術(shù),合規(guī)性檢測(cè)技術(shù)及音樂(lè)質(zhì)量軟件實(shí)現(xiàn)等方面進(jìn)行了詳細(xì)的討論和研究,主要內(nèi)容為:1.提出了一種基于特征碼分析的快速提取方法。該方法是分別對(duì)轉(zhuǎn)換前后的音樂(lè)文件所包含的幅度、過(guò)零率、共振峰等特征碼進(jìn)行提取和數(shù)學(xué)建模,然后通過(guò)特征碼匹配的方式來(lái)判斷音樂(lè)在轉(zhuǎn)換前后是不是出現(xiàn)質(zhì)量問(wèn)題。該方法是數(shù)字音樂(lè)質(zhì)量軟件的分析基礎(chǔ)。2.研究音樂(lè)文件合規(guī)性檢測(cè)方法,提出了一種基于顯式特征分析的合規(guī)性檢測(cè)方法。該方法對(duì)音樂(lè)文件的格式、位數(shù)、碼率、采樣率、單雙聲道等顯式特征進(jìn)行快速提取,然后查詢(xún)規(guī)格范圍數(shù)據(jù)庫(kù)進(jìn)行量度匹配檢測(cè),由此快速發(fā)現(xiàn)音樂(lè)文件不符合規(guī)格的質(zhì)量問(wèn)題。該方法是數(shù)字音樂(lè)質(zhì)量軟件的過(guò)濾器,能快速篩查音樂(lè)文件顯式質(zhì)量問(wèn)題,提升檢測(cè)效率。3.研究噪聲檢測(cè)方法,提出了一種基于無(wú)監(jiān)督分類(lèi)和最小風(fēng)險(xiǎn)率貝葉斯算法相結(jié)合的對(duì)音頻文件進(jìn)行質(zhì)量判定的方法。該方法采用無(wú)監(jiān)督分類(lèi)的方法對(duì)音頻文件進(jìn)行聚類(lèi),并反復(fù)迭代運(yùn)算使分類(lèi)達(dá)到的合理性。再利用貝葉斯條件概率算法確定和修正音樂(lè)文件中包含噪聲的概率,由此確定待測(cè)文件出現(xiàn)噪聲質(zhì)量問(wèn)題的概率。該方法是數(shù)字音樂(lè)質(zhì)量軟件的核心和難點(diǎn)。4.提出數(shù)字音樂(lè)質(zhì)量軟件系統(tǒng)的架構(gòu)和實(shí)現(xiàn)方案,并詳細(xì)論述了各個(gè)功能模塊的細(xì)節(jié)和實(shí)現(xiàn)方法。本論文提出的主要技術(shù)方法在數(shù)字音樂(lè)質(zhì)量軟件中得到了很好的實(shí)現(xiàn)和測(cè)試,實(shí)踐證明能有效的應(yīng)對(duì)電信運(yùn)營(yíng)商對(duì)數(shù)字音樂(lè)庫(kù)的質(zhì)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.
【學(xué)位授予單位】：電子科技大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類(lèi)號(hào)】：TN912.3

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 李昆侖;張偉;代運(yùn)娜;;基于Tri-training的半監(jiān)督SVM[J];計(jì)算機(jī)工程與應(yīng)用;2009年22期

，

本文編號(hào)：1942706