噪声环境下的音频同步算法的研究

发布时间：2018-03-28 01:07

本文选题：PEAQ　切入点：同步　出处：《南京大学》2014年硕士论文

【摘要】：21世纪是网络大爆炸的时代,伴随着因特网的迅速普及和高速发展,依靠互联网传输音频信息的需求也越来越大。流媒体的出现,在很大程度上改善了互联网传输音频难的局面,其将过去传统媒体的"推"式传播,变为受众的"拉"式传播与实时传播。由于流媒体技术在一定程度上突破了网络带宽对多媒体信息传输的限制,因此被广泛应用于网上直播、网络会议、远程教育及企业培训等多种领域。为了更好的进行流媒体的传输,通常需要对流媒体质量进行评测,同时也对流媒体音频的质量评测提出了新的挑战。本文主要讨论了音频客观质量评价算法和音频同步算法,并在理论的基础上对音频同步算法进行了实现,从而保证音频客观质量评价算法的准确性。本文算法的最终目标是用于音频客观质量评价算法,所以在论文前部分对音频客观质量评价算法进行了详细的研究,包括算法的流程、算法中间产生的参数以及算法最终的输出结果。经过对音频客观质量评价算法的研究发现问题,即音频客观质量评价算法在有延迟的情况下,无法正确给出音频客观质量评价的分数。这就引出了本文的主要研究对象——音频同步算法。本文在主体部分对音频同步算法给出了一个有效的解决方案。该方案是采用了二次包络及直方图延迟估计的方法,将参考音频和测试音频以3毫秒和1毫秒为单位分别做包络互相关,在两次包络互相关求取各段延迟的基础上做直方图精同步,并对实验结果做了分析。分析结果发现本文提出的基于短时能量二次包络及直方图延迟估计的方法能够有效的解决上文提出的问题,能够将延迟的样本点数缩小在48个采样点即1ms范围内,小于实际生活中音频通过编解码器产生的延迟,可以满足正常的需求。实验证明经过本同步算法对齐后的音频,再经过音频客观质量评价算法,可以得出正确的结果。
[Abstract]:The 21st century is the era of the Internet explosion. With the rapid popularization and rapid development of the Internet, the need to rely on the Internet to transmit audio information is also growing. To a large extent, it has improved the difficult situation of audio transmission on the Internet, which has "pushed" the traditional media in the past. Streaming media technology has broken through the restrictions of network bandwidth on multimedia information transmission to some extent, so it has been widely used in live webcasting and network meetings. In many fields, such as distance education and enterprise training, in order to transmit streaming media better, we usually need to evaluate the quality of streaming media. At the same time, it also presents a new challenge to the quality evaluation of streaming media audio. This paper mainly discusses the audio objective quality evaluation algorithm and audio synchronization algorithm, and implements the audio synchronization algorithm based on the theory. In order to ensure the accuracy of the audio objective quality evaluation algorithm, the final goal of this algorithm is to use the audio objective quality evaluation algorithm, so in the first part of the paper, the audio objective quality evaluation algorithm is studied in detail. Including the flow of the algorithm, the parameters produced in the middle of the algorithm and the final output of the algorithm. Through the study of the audio objective quality evaluation algorithm, we found the problem, that is, the audio objective quality evaluation algorithm in the case of delay, The score of audio objective quality evaluation can not be given correctly. This leads to the main research object of this paper-audio synchronization algorithm. This paper gives an effective solution to audio synchronization algorithm in the main part. The method of quadratic envelope and histogram delay estimation is used. The reference audio and test audio are correlated in three milliseconds and one millisecond respectively, and the histogram precision synchronization is done on the basis of two envelope cross-correlation to obtain the delay of each segment. The experimental results show that the proposed method based on the short time energy quadratic envelope and histogram delay estimation can effectively solve the above problems. The number of samples can be reduced to 48 sample points, that is, 1ms, which is smaller than the delay caused by audio through codec in real life, which can meet the normal demand. After the audio objective quality evaluation algorithm, we can get the correct results.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN912.3

【相似文献】