改进的MELP低速率语音编码器.pdf

ComputerEngineeringandApplications~算机工程与应用 2011，47(29) 131 改进的MELP低速率语音编码器 冯晓荣，刘晓明，田 雨 FENG Xiaorong，LIU Xiaom ing，TIAN Yu 重庆大学 通信工程学院，重庆 400030 College ofCommunication Engineering，Chongqing University，Chongqing 400030，China FENG Xiaorong，LIU Xiaom ing，TIAN Yu．Improved M ELP speech coderfor low bitrate speech coding．ComputerEngi- neeringandApplications。2011，47(29)：131—133． Abstract：In theMixedExcitation LinearPrediction(MELP) algorithm，rfameswith excessive switchingsresultin qualiyt degradation forsynthesized speech，which isa seriousproblem．In orderto solve thisproblem ，a new approach isgiven to an— alyzeand synthesizethetransition segmentswhich are called the transition rfames．Thiscan improvethe accuracy ofparame— terestimation and reduce thenoise caused by classifying the transition flame simply into U／v flame．A new flame type deci— sion algorithm with dynam ic U，V threshold related to the pitch lag is introduced which classifies the speech flame into three types：voiced，unvoiced，and transition．The classifiercan reduce the U／V decision errors and avoid excessive switchings betweenvoicedflameandunvoicedflame．An improvedbitallocation tableisintroducedandthePESQ--MOStestshows thatthesynthesized speechqualiyt hasbeen greatly improved throughthenew MELP speech coder，especially ofrhigh fre— quency femalespeakers． Keywords：MixedExcitationLinearPrediction(MELP)；Unvoiced／Voiced(U／V)decision；speechcoder 摘 要：在混合激励线性预测(MELP)语音编码算法中，语音帧的突变转换导致合成语音质量的下降成为一个突出问题。为解决 该问题，提出一种基于过渡帧判决算法的改进MELP模型，提高了参数估计的准确度，有效实现了语音的降噪处理。引入动态清 浊音判决 (u／V判决)门限将语音帧分为3种类型：浊音帧、清音帧、过渡帧，通过区分过渡帧和清浊音帧，避免了传统的u／v判决 错误和清浊音帧的突变转换。给出了改进MELP编码参数比特分配表，通过PESQ--MOS测试表明，合成语音质量尤其是高频女 声合成语音质量有 了明显的改进。 关键词：混合激励线性预测 (MELP)；清 ／浊音(uV／)判决；语音编码 DOI：l0．3778／j．issn．1002—8331．20l】．29．036 文章编号：1002．8331(2011)29．013l一03 文献标识码：A 中图分类号：TN9123 MELP语音编码器是美国在2．4Kb／s速率上的窄带安全 量下降。 语音通信应用标准 。2．4Kb／sMELP低比特率数字语音编码 本文引入过渡帧来完善MELP算法模型，改进了LSF和基音 技术是