L.W. He, E. Sanocki, A. Gupta and J. Grudin, Auto-Summarization of Audio-Video Presentations, in Proc. of ACM MM 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....face detectors and performed motion analysis for additional cues. The MoCA project [13] worked on automatic generation of film trailers. They used heuristics on the trailers, along with a set of rules to detect certain objects (e.g. faces) or events (e.g. explosions) Work at Microsoft Research [9] dealt with informational videos; there, they looked at slide changes, user statistics and pitch activity to detect important segments. Recent work [11] has dealt with the problem of preview generation by generating interesting regions based on viewer activity in conjunction with topical phrase ....

....determining useful segments in the audio stream. Let us assume that we wish to compress an audio track that is 100 sec. long, by 90 . Then: a) downsampling the audio by 90 will leave the audio to be severely degraded since the pitch of the speech segments will increase dramatically. b) PR SOLA [9] is a non linear time compression technique that eliminates long pauses, and attempts to preserve the original pitch in the output. User studies indicate that users do not prefer to have the speech sped up beyond 1.6x (i.e. 40 compression) c) selecting only those segments that are synchronous ....

[Article contains additional citation context not shown here]

L. He et. al. Auto-Summarization of Audio-Video Presentations, ACM MM '99, Orlando FL, Nov. 1999.

....associated with each video type are reported. Huang et al. 24] use a set of features (Features a.2, a.3, b.2, b.3, c.1, and more) to form feature vectors for audio break detection. The result is combined with color and motion break detection results to segment videos into scenes. He et al. [22] use Feature b.1 to identify the speaker s emphasis in his her oral presentation, based on the observation that the speaker s introduction of a new topic often corresponds to an increased pitch range in his her voice. This information, combined with slide transition information of the ....

L. He, E. Sanocki, A. Gupta, and J. Grudin, "Auto-summarization of audio-video presentations," Proc. ACM Multimedia 99, pp. 489-498, Oct. 1999, Orlando, FL.

....information. Some interesting early work was done by Arons [12] in trying to aggressively speed up informational talks. He noticed that relative pitch increases for people when they are emphasizing points. In his Speech Skimmer system, he used that for prioritizing regions within a talk. He et al. [13] further built upon Aron s work and constructed presentation summaries based on pitch analysis, knowledge of slide transitions in the presentation, and information about previous users access patterns. The study showed that the automatically generated summaries were of considerable value to the ....

He, L., et al. Auto-Summarization of audio-video presentations.in ACM Multimedia. 1999.

....associated with each video type are reported. Huang et al. 24] use a set of features (Features a.2, a.3, b.2, b.3, c.1, and more) to form feature vectors for audio break detection. The result is combined with color and motion break detection results to segment videos into scenes. He et al. [22] use Feature b.1 to identify the speaker s emphasis in his her oral presentation, based on the observation that the speaker s introduction of a new topic often corresponds to an increased pitch range in his her voice. This information, combined with slide transition information of the ....

L. He, E. Sanocki, A. Gupta, and J. Grudin, "Auto-summarization of audio-video presentations," Proc. ACM Multimedia 99, pp. 489-498, Oct. 1999, Orlando, FL. 50

....use a simple technique to make a distinction between thematic and actual spoken dialogue. From test data we observe that the average shot length for a thematic dialogue is much shorter than for a spoken dialogue. The reason is that there is a minimum time required to utter a meaningful phrase. In [7], the authors assume that phrases last between 5 15 sec. An analysis of hand labeled data reveals that dialogues with average shot length of less than 4 sec. are thematic. 4.2.4 The Sliding Window Algorithm We use a sliding window algorithm to detect the presence of a dialogue (thematic or true ....

Liwei He et. al. Auto-Summarization of Audio-Video Presentations, ACM MM #99, Orlando FL, Nov. 1999.

....providing another factor related to the precedence of a frame, so that earlier appearing frames are more heavily weighted than later ones in the same similarity class. There are several reasons for using precedence as a criterion. For video, it has been noticed in video playback usage studies (see [5]) that the earlier appearances of an event are accessed more. For images of people or slides, the earlier ones may introduce or define things that the later ones will refer to. For video from surveillance or wearable personal video cameras, the frames can be processed backwards (or invert our ....

He, L., Sanocki, E., Gupta, A., Grudin, J. Autosummarization of audio-video presentations. Proceedings of ACM Multimedia '99. ACM Press, pp. 489-498.

....less versatile. Efforts are underway to make multimedia equally useful. Compression and indexing, automated wholly or in part, can facilitate skimming and browsing, reducing the time to watch. When available, such capability will provide further incentive to view material asynchronously (e.g. [6, 8]) One technology that facilitates collaboration around multimedia is distributed tutored video instruction (DTVI) in which students meet virtually to watch and discuss a lecture video. Sun and Microsoft researchers have tested systems that allow viewers in different locations, linked by a ....

He, L., Sanocki, E., Gupta, A., and Grudin, J. (1999). Auto-summarization of audio-video presentations. Proc. Multimedia 99, 489-498.

....we may select only the first 30 seconds of audio video after each of the slide transitions in the presentation, or have a human identify key portions of the talk and include only those segments, or base it on the access patterns of users who have watched the talk before us. In an earlier paper [8], we studied three automatic methods for creating audio video summaries for presentations with slides. These were compared to author generated summaries. While users preferred author generated summaries, as may be expected, they showed good comprehension with automated summaries and were overall ....

....this study extends. Next, the experimental design of the current study is presented, followed by the results section. Finally, we discuss related work and draw conclusions. AUTOMATIC AUDIO VIDEO SUMMARIZATION We briefly summarize our earlier study on automated audio video summarization methods [8]. The combination of the current study and this older study enable us to build a more complete picture of the overall tradeoffs. Our study used a combination of information sources in talks to determine segments to be included in the summary. These were: 1) analysis of speech signal, for example, ....

He, L., Sanocki, E., Gupta, A., & Grudin, J., 1999. Auto-summarization of audio-video presentations. In Proc. Multimedia'99. ACM.

....particularly appeal to on line viewers (some of whom suggested this feature) The results also guide those building tools to support the viewing and authoring of digitized presentations. Greater support for skimming and browsing is possible, including automatic generation of multimedia summaries [7]. Mixedinitiative authoring tools are possible. Presenters invest time preparing a talk for a live audience; when more than half of the total audience could be on demand, they may be happy to contribute to post processing the talk. Even a few minutes could contribute significantly to preparing ....

He, L., Sanocki, E., Gupta, A., & Grudin, J., 1999. Auto-summarization of audio-video presentations. In Proc. Multimedia'99. ACM.

No context found.

L.W. He, E. Sanocki, A. Gupta and J. Grudin, Auto-Summarization of Audio-Video Presentations, in Proc. of ACM MM 1999.

No context found.

He, L., Sanocki, E., Gupta, A., & Grudin, J. (1999). Auto summarization of audio-video presentations. ACM International Conference on Multimedia. p. 489498.

No context found.

L. He et al., "Auto-summarization of audio-video presentations," Proc.ACM Multimedia, pp. 489--498, 1999.

No context found.

L. He et al., "Auto-Summarization of Audio-Video Presentations," Proc. ACM Multimedia 99, ACM Press, New York, 1999, pp. 489-498.

No context found.

L. He et. al. Auto-Summarization of Audio-Video Presentations, ACM MM #99, Orlando FL, Nov. 1999.

No context found.

He, L., Sanocki, E., Gupta, A., Grudin, J. (1999) Autosummarization of audio-video presentations. Proceedings of ACM Multimedia '99. ACM Press, pp. 489-498.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC