基于ＭＰ时频特征的电影音频场景语义推理研究

项目名称： 基于ＭＰ时频特征的电影音频场景语义推理研究

项目编号： No.61301300

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 无线电电子学、电信技术

项目作者： 杨继臣

作者单位： 华南理工大学

项目金额： 24万元

中文摘要： 随着电影的日益增加，为了更好地分析、索引和管理越来越庞大的电影数据，方便人们快速检索到自己想要的信息，对电影进行内容分析和理解变得越来越迫切，针对目前电影音频内容分析与理解策略所存在的音频事件改变检测没受到重视、场景语义推理没被有效处理等问题，本项目主要探索电影音频内容分析与理解的实用方法，重点研究以下四个方面的问题：1）时频特征提取；2）音频事件改变检测；3）音频事件识别；4）场景语义推理。旨在通过上述四个问题的研究，弥补当前电影音频处理策略所存在的缺陷。 本项目的研究内容是语音及音频信号处理、模式识别与理解等领域的研究热点，理论上具有一定的挑战性，在应用上是研发音频内容分析及理解系统的重要基础。同时，对提高音频处理系统的实用性，加快我国信息产业的发展具有重要作用。预期发表EI索引论文3篇、SCI索引论文2篇、申报专利1项。

中文关键词： 时频特征提取；音频事件改变检测；音频事件识别；音频场景识别；

英文摘要： With the increasing of movies every day, in order to analyze,index and organize more and more vast movie data, making it esaier for people to retrieve wanted information, it become more and more urgent to analyze and understand movie content.Aiming at the problems of audio event change detection been not payed attention to and scene semantic been not effectively processed in the movie content analysis and understanding tactics, this project mainly explores practical methods , four major problems will be studied: 1) time frequency feature extraction, 2) audio event change detection , 3) audio event identification，4）scene semantic inference. Focusing on aforementioned four problems, this project aims to recuperate the defects in the current tactics of movie audio processing. The content includes hot topics in speech and audio signal processing, pattern recognition and understanding, etc., all of which have particular challenge in theory and the important base of study and development audio content analysis and understanding system in application. In addition, it plays an important role to improve system practicability and to speed up the information industry development of our country. Three papers indexed by EI , two papers indexed by SCI, are expected to publish and one patent will be declared.

英文关键词： time-frequency feature extraction；audio enent change detrction；audio event recognition；audio scene recognition；