以人工智慧方法判別電視廣告所在位置

Using Artificial Intelligent Approaches to Locating TV Commercial Films

許良維(Liang-wei Hsu)

指導教授 ： 黃有評

大同大學/電機資訊學院/資訊工程學系所/碩士(2005年)

摘要

本篇論文提出了一個針對錄下來的電視節目(包含節目內容與廣告)裡檢出並判定廣告位置的智慧型系統。其中整合了ANFIS，ART網路和基因演算法等人工智慧技術，以實現一個能檢出廣告片段可能位置的建議系統。依節目類型不同，我們可以組合不同策略去找出影片內容中能用來分辦節目內容和廣告片段的特徵，並找出廣告片段的可能位置。最後以一個人機互動的界面呈現，使得使用者能快速簡便地找出廣告和節目內容的交界位置。
由於電視廣告和節目的內容是完全不相關，使得使用者通常都不希望觀看廣告，或必需先把廣告片段從錄下的電視節目中挑出之後，才能使用錄下的內容當成多媒體編輯的材料。我們提出了一種可行的系統，運用人工智慧的一些方法，幫助使用者在離線的狀態下找出可能的廣告片段位置，以方便進行後續的處理。
在論文的最後我們以錄下的幾個電視影片來驗證這個系統的可行性，並得到90%以上的準確率和召回率，此結果驗証所提系統之可用性。

關鍵字

廣告位置判別 ； 人工智慧 ； 廣告偵測

並列摘要

An artificial intelligent (AI) system, integrating ANFIS, ART network and genetic algorithm (GA), is proposed in this thesis to implement the commercial film (CF) detection and locating for the recorded TV programs. According to the types of the recorded TV programs, this content-based detection system can take different strategies to find out the characteristics, which the program content (PC) and CF differ, and then locate CFs. With the probable locations of CFs, an interactive user interface is proposed to enable CF cutting operation. Users can easily find out the precise boundary between CFs and PCs.
Since commercial films are not related to the program content at all, they are usually not wanted in TV watching. Or users have to trim off the CF parts, and then extract the useful contents from the recorded TV program video for further multimedia editing or authoring. The proposed system uses some AI-based approaches to locate CFs automatically and makes the locating procedure efficient.
At the end of this thesis, we take some cases to verify the feasibility of this system. Based on the experimental results, the precision rate and recall rate can both reach 90% at least, and the results verify the high usability of the proposed system.

並列關鍵字

Artificial Intelligence ； Commercial Film Detection ； Commercial Film Locating

參考文獻

[1] Y. Taniguchi, A. Akutsu, Y. Tonomura and H. Hamada, "An intuitive and efficient access interface to real-time incoming video based on automatic indexing," Proc. of ACM Multimedia 95, San Francisco, Calnifornia, 1995.

[2] G. Davenport, T. Smith and N. Princever, "Cinematic primitives for multimedia," IEEE Trans. on Computer Graphics and Applications, vol. 11, No.4, pp.67-74, 1991.

[3] R. Lienhart, C. Kuhmunch, W. Effelsberg, "On the detection and recognition of television commercials," Proc. of IEEE International Conference on Multimedia Computing and Systems, pp.509-516, 1997.

[4] A. Hauptmann, M. Witbrock, "Story segmentation and detection of commercials in broadcast news video," Proc. of Advances in Digital Libearies Conference (ADL '98), Santa Barbara, CA, Apr. 22-24, pp.168-179, 1998.

[5] P. Duygulu, M.-Y. Chen and A. Hauptmann, "Comparison and combination of two novel commercial detection methods," Proc. of the 2004 IEEE International Conference on Multimedia and Expo (ICME 2004), Taipei, Taiwan, 2004.