广播电视监播系统中音频及语音数据处理技术研究

【作者】 王海涛；

【导师】 马建仓；

【作者基本信息】 西北工业大学 ， 电工理论与新技术， 2007， 硕士

【摘要】 随着广播电台和电视台栏目的丰富，以及数字化产业的不断发展，需要对这些广播电视节目的内容和质量进行监督管理，因而产生了广播电视监播系统。本篇论文将对这种监播管理系统的设计背景、主要用途、应用原理、发展趋势和主要设备进行介绍，并开发出监播系统音频文件的管理软件，并嵌入语音识别技术。针对实际应用中广播电视节目的音频信号不易观察的难点，完善整个系统，提高工作效率，满足用户要求。本文的主要工作分为以下几个部分： (1)消化国内外相关的资料；综述监播系统的发展过程；介绍监播系统的组成、工作原理、工作流程和主要设备；结合目前流行的语音识别技术的发展，展望未来该产业的发展趋势。 (2)为了能够对节目当中声音信号进行观察、编辑、处理等操作，运用VC++语言开发了一套应用软件。利用该软件可以对声音信号实时播放，控制操作简便，并最终能够方便的嵌入到整个监播系统的控制软件当中去。 (3)为实现监播系统中对语音信号进行识别的功能。本文研究和分析了语音识别技术的特点。借鉴一套现有的语音识别软件，掌握它的设计特点、结构特征和操作步骤，统计它的识别效率；对其中的关键字识别方案进行改进，提高识别率。 (4)针对含有背景噪声的语音信号识别率将会降低的现象，本文研究语音处理技术的相关理论，区别语音信号与噪声信号的特征。采取谱相减的语音增强方法处理语音信号；并对该方法加以改进，设计一种新的语音增强方案。并在MATLAB编程环境下对上述各种语音增强方法进行了仿真，统计出语音增强后的语音识别率的提高效果。更多还原

【Abstract】 With Broadcasting and TV program being rich and colorful, the Broadcasting and Television digital industry develops rapidly. It is necessary to monitor the contents and quality of these programs so that the Broadcasting and Television repertoire scrutiny system is created by the time. This paper will research about software exploitation by introducing about the background, the purpose, the elements, the developing trend and equipment of this repertoire scrutiny system. Since in practice, it is hard to observe and deal with audio signal in Broadcasting and TV programs, so this paper use advanced computer control and speech recognition technique to solve these problems in order to consummate the whole system function, improve automatic work efficiency and satisfy customers’ need. The primary assignments of this paper include such parts:(1)Collect domestic and foreign documents of the products in this filed, describe the development of scrutiny system; introduce the functions, configuration, theory, work flow and equipments of this system, with currently developed speech recognition technique, prospect the developing trend of this industry.(2)Use VC++ computer language to design an application which can observe, edit and dispose audio signal in Broadcasting and TV programs. This application can play audio signal in time, and easy to use. Besides, this application can be inserted easily to the whole system software.(3)In order to realize the speech recognition function of this system, this paper analyses the characteristic of speech recognition technique. Based on an existed speech recognition software, by studying its designing feature, configuration, working procedure and summing up its recognition efficiency, we insert it into the whole system software to recognize audio files, especially to improve the keywords recognition scheme to increase recognition rate.(4)If the speech signal mix with background noise, the recognition rate will turn down heavily, so in this paper we will research speech processed technique, distinguish the difference between speech signal and noise signal. This paper adopt spectral subtraction approach to process speech signal, and improve it to make new speech strengthen scheme. The realization of these speech strengthen approaches are based on更多还原