Academic journal article Journal of Digital Information Management

Story Segmentation of Xinwenlianbo Based on Three-Level Criterion of Content Analysis

Academic journal article Journal of Digital Information Management

Story Segmentation of Xinwenlianbo Based on Three-Level Criterion of Content Analysis

Article excerpt

I. Introduction

Xinwenlianbo (XWLB) of China central TV station (CCTV) has the highest audience rating in China and the largest audience rating around the world [1]. From the point of save archives yXWLB is the true record of home and abroad news, which contains politics, economics, science, technology etc. So that it is one of the most possible save archives program. News story is an independent and total content for the former and latter video. Because every piece of home brief news or abroad news is too short in XWLB, so we regard home brief news and abroad news as a story unit respectively. Generally speaking, there are about twenty stories in a day's XWLB. Dividing XWLB into independent story not only help to analyze audio content but also help multimedia information retrieval and understanding [2].

There are mainly three study methods for XWLB story segmentation (SS) in published references. The first uses audio method, which utilizes anchorperson's voice to train model first and then segmentation story according to model beforehand, while many stories have no anchorpersons in XWLB. The second uses video feature to segment story according recognized anchorperson, which also has the limit like the first method. The third combines video feature detection and news title, but it dependent on silence for story boundary (SB) without anchorperson detection, silence is only potential story change point and there is silence in a sentence, that is to say, independent on silence is not a good method.

SB detection is not precisely in the methods of XWLB SS. It is difficult to detect the SB of story without anchorperson for the first and second method. The second and third only used video feature to detect anchorperson, whose disadvantage is removing the false anchorperson, the best method is to utilize audio method to judge for the anchorperson by using video feature. For the story without anchorperson, the SB is not precisely by using silence to seek SB, the best method is to seek SB of story according some rule first and combining shot change point and silence to seek story change point.

In order to detect the SB of XWLB precisely, we collected 180 days' XWLB data of 2009 first, and then marked them by hand, the mark content contain as following: every person appearing in the news story, start and end time of every person, start and end time of every news title and etc. At the base of mark, character of XWLB is concluded. We also go on statistics of story type, the position o and time length of news title. Finally, a story segmentation of XWLB based on three-level criterion of content analysis. For the story with anchorperson, the story beginning boundary (SBB) is gotten by using penalty distance to judge the anchorperson by video feature. For the story without anchorperson, SBB is gotten by judging whether shot change point appear silence region. The experimental result show that the proposed method a detect SBB precisely.

2. Character of Xinwenlianbo

The structure of XWLB is firstly analysed first and then story type is concluded in this section.

2.1 Structure Analysis

XWLB has fixed structure, whose time length is usually half an hour or so. It is often composed six parts: start time, brief news, home news, nei-ji music, abroad news and news end (Figure.1). The time length of start music is 16 s or so and time length of brief news is between 34s and 58s, several pieces of home brief news is often following with home news, there is a 5-second-length segment music between home news and abroad news, which is called nei-ji music, news end is usually at the last 30s of XWLB. There is an anchorman and anchorwoman, they appear in the brief news and news end together, there is only one anchorperson other time. In the process of introduction story, the person, whose's voice can be heard but the body can not be seen, is called as interpreter by us in this paper. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.