哈 尔 滨 理 工 大 学
毕 业 设 计
题 目： 音频信号的分类与分割 院 系： 电气与电子工程学院 姓 名： 指导教师： 系 主 任：
2011年6月23日
哈尔滨理工大学学士学位论文 音频信号的分类与分割
摘要
随着计算机技术、网络技术和通讯技术的不断发展，图像、视频、音频等多媒体数据已逐渐成为信息处理领域中主要的信息媒体形式，其中音频信息占有很重要的地位。同时，由于信息获取的方式、手段和技术的不断进步和多样化，使得信息数据量以极高的速度增加，为有效的处理和组织信息带来了挑战，而信息有效的处理和组织是深入分析和充分利用的前提。
原始音频数据是一种非语义符号表示和非结构化的二进制流，缺乏内容语义的描述和结构化的组织，给音频信息的深度处理和分析工作带来了很大的困难。如何提取音频中的结构化信息和内容语义是音频信息深度处理、基于内容检索和辅助视频分析等应用的关键。音频分类与分割技术是解决这一问题的关键技术，是音频结构化的基础。
本文介绍了在MATLAB环境中如何进行语音信号采集后的时频域分析处理，并通过实例分析了应用MATLAB处理语音信号的过程。
本文根据模式识别理论分析了音频分类与分割的技术流程，同时讨论了其中涉及的相关技术;介绍了特征分析与抽取，以及采用的相关音频处理技术。
关键词 MATLAB；语音信号；特征分析
- I -
哈尔滨理工大学学士学位论文 The classification and segmentation of the Audio
Abstract
With the continually evolving of computer technology, network technology and communication technology, images, video, audio and other multimedia data in the field of information processing has become the main form of information media, audio information plays an especially important role.
At the same time, due to the way access to information, tools and technology continues to progress and diversify, the amount of data information increase at very high speed, which has brought challenges for
efficient
processing
and organizing
of
the
information ,
and
effective processing and organization of i information are premise of analysis and full use of the .
The original audio data is a non-semantic notation and unstructured binary stream, lack of content and structure of semantic description of the organization, which has led to great difficulties to the depth of audio information processing and analysis. How to extract structured information in audio and audio information content is the key for the depth of semantic processing, video content-based retrieval and analysis applications supporting. Audio classification and segmentation is a key technology to solve this problem is the structural basis for the audio.
This article describes how the MATLAB environment for voice signal collected after the time-frequency domain analysis and processing, and analysis of the application by example MATLAB to handle voice signals.
Our theoretical analysis is based on pattern recognition, audio classification and segmentation of the technical process, and involving the related
- II -
哈尔滨理工大学学士学位论文 technologies discussed; We describe the characteristics analysis and extraction, and to the corresponding audio processing technology
The last chapter involves the summary and evaluation all the work of the paper, and this research were discussed for future.
Keywords：MATLAB；Voice signal; Characteristics
- III -
哈尔滨理工大学学士学位论文 目录
摘要 ....................................................................................................................... I Abstract ............................................................................................................... II
第1章 绪论 ........................................................................................................ 1
1.1 研究背景 ................................................................................................ 1 1.2 语音信号的采集 .................................................................................... 3
1.2.1 预加重处理 .................................................................................. 3 1.2.2 切分与加窗处理 .......................................................................... 3 1.3 研究的主要内容 .................................................................................... 4 第2章 音频分类与分割技术研究现状 ............................................................ 5
2.1 音频语义内容分析 ................................................................................ 5 2.2 层次化音频结构分析框架 .................................................................... 6 第3章 音频信号特征的提取 ............................................................................ 8
3.1 语音端点检测的基本方法 .................................................................... 8
3.1.1 短时加窗处理 .............................................................................. 8 3.1.2 短时平均能量 .............................................................................. 8 3.2 短时平均过零率 ................................................................................... 11 3.3 基于能量和过零率的语音端点检测 .................................................. 14 第4章 语音信号的短时频阈分析 .................................................................. 16
4.1 语音信号的快速傅里叶变换 .............................................................. 16 4.2 临界频带谱平坦测度函数计算 .......................................................... 18 4.3 基于短时能量比的语音端点检测算法的研究 .................................. 19 4.4 音频信号的功率谱分析 ...................................................................... 20 4.5 音频信号的子带熵分析 ...................................................................... 21 结论 .................................................................................................................... 22 致谢 .................................................................................................................... 23 参考文献 ............................................................................................................ 24 附录A ................................................................................................................ 26 附录B ................................................................................................................ 33
- IV -