Sliding text recognition in broadcast news

Dikici, Erinç; Saraçlar, Murat

doi:10.1109/siu.2008.4632711

Cited by 8 publications

(10 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…To exemplify, Dikici and Saraclar presents a method that uses linguistic cues to correct sliding texts that are presented in newscasts, and as a result, word and character recognition is increased remarkably 9 . This study is a good example identifying the importance of the knowledge-based approach in improving the quality of the results.…”

Section: Literaturementioning

confidence: 99%

Text recognition and correction for automated data collection by mobile devices

Özarslan

Eren

2014

SPIE Proceedings

View full text Add to dashboard Cite

Participatory sensing is an approach which allows mobile devices such as mobile phones to be used for data collection, analysis and sharing processes by individuals. Data collection is the first and most important part of a participatory sensing system, but it is time consuming for the participants. In this paper, we discuss automatic data collection approaches for reducing the time required for collection, and increasing the amount of collected data. In this context, we explore automated text recognition on images of store receipts which are captured by mobile phone cameras, and the correction of the recognized text. Accordingly, our first goal is to evaluate the performance of the Optical Character Recognition (OCR) method with respect to data collection from store receipt images. Images captured by mobile phones exhibit some typical problems, and common image processing methods cannot handle some of them. Consequently, the second goal is to address these types of problems through our proposed Knowledge Based Correction (KBC) method used in support of the OCR, and also to evaluate the KBC method with respect to the improvement on the accurate recognition rate. Results of the experiments show that the KBC method improves the accurate data recognition rate noticeably.

show abstract

Section: Literaturementioning

confidence: 99%

Text recognition and correction for automated data collection by mobile devices

Özarslan

Eren

2014

SPIE Proceedings

View full text Add to dashboard Cite

show abstract

“…The videos are given as input to a news story segmenter which is a silence detection implementation outputting the segment boundaries corresponding to news stories. These individual segments are fed into the sliding text recognizer presented in [8] and the resulting segment texts are provided as input to the rulebased named entity recognizer [9]. The named entities output by this recognizer constitute the semantic annotations of the news story segments.…”

Section: A News Video Databasementioning

confidence: 99%

“…In order to obtain the speech transcriptions of the videos which are also given as sliding texts in the videos of our data set, we employ the sliding text recognizer presented in [8]. This recognizer first converts the video frames into binary images and utilizes the horizontal and vertical histograms of these images to determine the exact text band.…”

Section: Sliding Text Recognitionmentioning

confidence: 99%

“…Next, connected component analysis is used to segment the text image into individual characters after some noise removal operations. The recognizer then employs template matching for actual character recognition and finally, it carries out a correction procedure based on rules learned through transformation-based learning (TBL) in order to improve character recognition accuracy [8].…”

Section: Sliding Text Recognitionmentioning

confidence: 99%

See 1 more Smart Citation

A text-based fully automated architecture for the semantic annotation and retrieval of Turkish news videos

Kuccuk

Yazıcı

2010

International Conference on Fuzzy Systems

View full text Add to dashboard Cite

Video texts are known to constitute an important source of information for semantic summaries of video archives. In this study, we propose a fully automated architecture for semantic annotation and later retrieval of Turkish news videos based on the corresponding video texts. At the core of the architecture is a named entity recognizer, the output of which on video texts is used as semantic annotations for the corresponding videos. The architecture also comprises components for news story segmentation, sliding text recognition, and video retrieval in addition to a news video database. The news story segmentation module makes use of the audio waveforms of the raw video files to detect the boundaries of individual news stories. The sliding text recognizer is then executed on the video segments corresponding to these news stories to extract their texts. The texts are then fed into the named entity recognizer for Turkish news texts to extract the named entities which are to be used as semantic annotations or index terms for the retrieval of these news videos. Finally, the retrieval interface of the overall architecture enables access to the annotated videos and video segments through boolean queries formed by using the previously extracted named entities. This study is significant for its proposing the first fully automated architecture for the semantic annotation and retrieval of Turkish news video archives.978-1-4244-8126-2/10/$26.00 ©2010 IEEE

show abstract

“…[3] proposed multi-frame combination technique as a robust video text feature extraction method. [4], [5] finds the text location from the horizontal projection of the grey-level image.…”

Section: Introductionmentioning

confidence: 99%

HMM-based sliding video text recognition for Turkish broadcast news

Som

Can

Saraçlar

2009

2009 24th International Symposium on Computer and Information Sciences

Self Cite

View full text Add to dashboard Cite

In this paper, we develop an HMM-based sliding video text recognizer and present our results on Turkish broadcast news for the hearing impaired. We use well known speech recognition techniques to model and recognize sliding video text characters using a minimal amount of labeled data. Baseline system without any language modeling gives a word error rate of 2.2% on 138 minutes of test data. We then provide an analysis of character errors and employ a character-based language model to correct most of them. Finally we decrease the amount of training data to a quarter, split the test data into halves and investigate semi-supervised training. Word error rates after semi-supervised training are significantly lower than to those after baseline training. We see 40% relative reduction in word error rate (1.5 → 0.9) over the test set.

show abstract

Sliding text recognition in broadcast news

Cited by 8 publications

References 8 publications

Text recognition and correction for automated data collection by mobile devices

Text recognition and correction for automated data collection by mobile devices

A text-based fully automated architecture for the semantic annotation and retrieval of Turkish news videos

HMM-based sliding video text recognition for Turkish broadcast news

Contact Info

Product

Resources

About