基于词激活力的语音关键词检测过滤排序算法研究
首发时间:2014-12-10
摘要:在信息社会,人们从各种途径获取信息,对于大量的爆炸式增长的多媒体数据,其中存在着大量的信息,由于语音的特殊性,人们想要从海量的多媒体数据中直接获取需要的信息是非常低效耗时的,人们希望能像文本检索那样从海量的多媒体数据中快速掌握大量的数据,所以如何从海量多媒体数据中进行信息挖掘是当前信息检索领域的一个热点之一。语音检索是实现从海量多媒体数据进行信息挖掘的主要方向之一,它以语音识别为依托,对识别结果进行处理方便人们对语音文件进行检索,但是语音识别错误的存在,是语音检索的性能无法满足人们的需求,本文通过借助语音模型和词激活力模型在两层索引的基础上,提出一种可靠的语音检索排序算法,提高语音语音关键词检索的性能。
For information in English, please click here
Spoken term detection sort algorithm Based on Word Activition Force
Abstract:In the information society, people obtain information from a variety of ways, there are massive information in the explosive growth of the multimedia data, due to the particularity of multimedia, people want to directly obtain the suitabe information from the huge amounts of multimedia data is very inefficient and time-consuming, people hope to be able to get the huge amount of multimedia information quickly like text retrieval, so how to carry on mining the information from the huge amounts of multimedia data is the one of hotspots in the field of information retrieval. Audio retrieval is one of the main direction of multimeida information mining, which is based on speech recognition, through deal with the recognition results, make the recognition results are convenient for people to retrieve audio files, but because of the existence of the speech recognition error, the performance of audio retrieval can't satisfy people's needs., this paper propose a speech retrieval sort algoritm which based on language model and word activition force, which develop the accuracy in audio retrieval, and develop theperformance of spoken term detection.
Keywords: speech recognition speech retrieval word activition force confusion network
基金:
论文图表:
引用
No.4621753100890314****
同行评议
共计0人参与
勘误表
基于词激活力的语音关键词检测过滤排序算法研究
评论
全部评论0/1000