基于MCF5249的VoIP语音处理模块设计与实现

【作者】 田峰；

【导师】 李广军；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2006， 硕士

【摘要】 VoIP,是Voice over IP的简称,也就是IP语音技术,其基本思想是通过IP包的传送来实现语音业务。其最大的优势是能广泛地采用Internet和全球IP互连的环境,提供比传统业务更多、更好的服务。目前,VoIP己经在全球范围内得到了迅速的发展,并能够实现传统的PSTN的绝大部分功能,在这基础上人们对VoIP的语音通话质量提出了进一步的要求。但是,由于网络存在时延抖动、带宽有限等固有特点,导致VoIP语音质量还不尽如人意,一度成为了制约VoIP发展的瓶颈。为了提高语音质量,需要采取一系列的语音处理技术,主要包括语音压缩技术、静音处理技术、回声消除技术、抖动缓冲技术等。本文基于“ColdFire平台的IP电话设计”项目,目的在于设计语音处理模块,并在目标平台M5249C3开发板上实现。该语音处理模块由两部分组成,一方面是对语音的处理,包括语音压缩模块、静音处理模块和回声消除模块,主要为了提高VoIP的语音质量;另一方面是对电话通信的控制和处理,包括双音多频模块和呼叫进程音模块,主要为了产生和检测IP电话通信中一些必须的电话信号。本文首先介绍了VoIP语音处理模块的主要技术,针对项目中需要设计实现模块的相关算法进行了分析,找到了适合的算法。接着介绍了基于MCF5249 ColdFire芯片的语音模块具体设计方法与实现过程。其中对语音编解码器的设计采用优化G.729A代码达到设计要求,并在此基础上加入G.729B的静音检测模块,以进一步降低网络传输带宽;对回声消除器的设计采用NLMS算法,通过设计自适应FIR滤波器和语音检测器达到回声消除目的;对双音多频设计,信号发生端采用构造静态参数表并通过二阶正弦振荡器产生信号,信号检测端提取频率信息以检测信号;对呼叫进程音设计,除了类似双音多频的信号发生及频率检测设计外,还需要检测信号持续时间,作者设计了一种基于匹配状态表的方法以检测信号持续时间。最后介绍了语音处理模块在硬件平台上的测试方法,并分析了测试结果。作者已经在Freescale的MCF5249芯片上实现了VoIP语音处理模块的设计,该模块工作正常,主要指标满足相关技术标准,达到了期望目标。更多还原

【Abstract】 VoIP, which stands for Voice over Internet Protocol, is the IP voice and speech technology. Its basic idea is to realize voice and speech business though transmitting IP packet. The most outstanding advantage of VoIP is capable of utilizing the world Internet networking to provide IP telephone service, which provides more and better value-added services than traditional telephone business.Up to now, VoIP has developed rapidly all over the world and implemented a majority part of traditional PSTN functions, based on which people raised additional requirement on speech quality that is not satisfied in VoIP now. But, as the inherent trait of delay and lack of bandwidth on the internet, the speech quality of VoIP is not as good as desired, which once prevent the development of VoIP. To improve the speech quality, it is necessary to adopt such speech processing techniques like speech compression, voice activity detection, echo cancellation, Jitter Buffer and so on.Built upon the Project Designing VoIP Terminal Based on ColdFire Platform, this paper designs speech processing module and successfully applied this solution on target platform. The speech processing module can be divided into two parts, the first part includes speech compression, voice activity detection and echo cancellation module, which improve speech quality; the other part includes DTMF and CPT module, which generate and detect some necessary telephony signal in the communication.This thesis is organized as follows. Section I introduces the main techniques in the VoIP speech module and analyze the relative algorithm. Section II describes the design approach and implementation of speech module on MCF5249 ColdFire core. The speech codec optimizes G.729A codes and added voice activity detection of G.729B to save bandwidth; the implementation of acoustic echo cancellation uses NLMS algorithm and it can reduce echo though designing adaptive FIR filter and speech detector; the DTMF and CPT generate signal using two second order digital sinusoidal oscillators and detect signal by picking up the frequency information. But only get the frequency information is not enough in CPT detector, this thesis introduces a method更多还原