摘要
音频分割是提取音频结构和内容语义的重要手段,是基于内容的音频分析、检索的基础。提出分层熵检测音频分割算法,采用定长分析窗分层结构遍历音频流,窗内根据熵变化趋势检测跳变点。实验结果表明,该算法避免了ΔBIC分割算法中的硬门限判决和数据累积问题,是一种更加有效的音频分割方法。
Audio segmentation is an important access to extract audio structure and content, aria is a basis for further audio retrieval and analysis. An audio segmentation algorithm is proposed based on layer entropy detection. The algorithm searches the audio stream using a size-fixed analysis window in which a top-down detection structure is employed and locates the change points according to the entropy trend. The experimental results demonstrate that algorithm avoids detection errors due to data accumulation and establishing hard threshold, it is a more effective segmentation algorithm.
出处
《科学技术与工程》
2009年第17期5012-5016,5023,共6页
Science Technology and Engineering
基金
国家863计划资助项目(2006AA01Z146)资助
关键词
音频分割
分层检测
熵变化趋势
audio segmentation layer detection the entropy trend
