洪巧巧200806052225基于内容的音频检索算法研究2012520.doc

毕 业 设 计 中文题目 基于内容的音频检索算法研究 英文题目 Research on Content-based Audio Retrieval Algorithm 系 别： 电子与电气工程系 年级专业： 2008级通信工程 姓 名： 洪巧巧 学 号： 200806052225 指导教师： 唐骏 职 称： 讲师 2012 年 6 月 1 日 毕业设计诚信声明书 本人郑重声明：在毕业设计工作中严格遵守学校有关规定，恪守学术规范；我所提交的毕业设计是本人在 指导教师的指导下独立研究、撰写的成果，设计中所引用他人的文字、研究成果，均已在设计中加以说明；在本人的毕业设计中未剽窃、抄袭他人的学术观点、思想和成果，未篡改实验数据。 本设计和资料若有不实之处，本人愿承担一切相关责任。 学生签名： 年 月 日 With the rapid development and widely popularity of Internet and network technology, peoples daily lives become increasingly dependent on the large number of multimedia data from the network, and hope to search the things interesting them. Content-based retrieval CBR (Content Based Retrieval) emerged, audio retrieval, image retrieval and video retrieval are the hot topics in the content-based retrieval fields, and one of the areas in the content-based retrieval is Query by Humming, Query by Humming’s development is accelerated, it has broad application prospect in the field of search engine, picking song system of KTV room, digital music library retrieval and so on. There?are?three?modules?in?query-by-humming?system,?construction?of?the?music?database?module,?feature?extraction?module?and?the?matching?module.?The?project?of?graduation?design?is?mainly?to?try?to?optimize?the?humming?program?through?the?research?of?the?speech?enhancement,?feature?extraction?and?feature?matching?algorithm.?It?is?necessary?to?select?the?most?appropriate?algorithm?for?each?module?and?try?to?optimize?part?of?the?algorithm,?also?the?algorithm?improved?effect?and?test?result?are?analyzed?and?discussed?to?reach?the?requirement?of?matching?accuracy?improvement. Keywords： Query by Humming, pre-processing, speech enhancement, feature extraction, melody matching algorithm 目录 第一章 绪论 1 1.1 课题研究背景与意义 1 1.2 哼唱检索系统整体研究现状 2 1.3 哼唱检索系统整体框架 3 1.4 论文的研究内容积总体框架 4 第二章 哼唱检索预处理模块的研究 5 2.1 语音增强的背景 5 2.2 语音增强算法 6 2.2.1 统计方法 6 2.2.2 参数方法 6 2.2.3 基于小波分解的方法 7 2.2.4 基于短时谱估计的方法 7 2.2.5 适合哼唱信号的语音