Scholarlink

<p>Video summarization is a method of deducing the content of video content for generating a summary in video format. The generated summary should have the significant segments of raw video. Recently, the content of video has been rapidly increasing, thus automatic video summarization is beneficial for individuals who want to keep time and learn more in a specific time. However, online courses do not fully control the content since it has diverse open challenges in video indexing, customization requirements, and summarization for particular courses. To overcome this gap, a semantic analysis of a summarized video named Fuzzy-based Deep Convolutional Kronecker Network (Fuzzy-based DCKN) is proposed. The input lecture audio and video are carried out on video shot segmentation is done by the YCbCr space color model. Afterwards, segment the audio and video from every slot using the Honey Badger Based Bald Eagle Algorithm (HBBEA), and the features are extracted from this phase. Then, select the important segments by Deep Residual Network (DRN). Furthermore, combine the relevant audio and video to obtain the summarized video. Here, the summarized video is given to frame extraction individually, where the text content extraction and residual image extraction are accomplished by optical character recognition (OCR) and Deep Residual Network (DRN) respectively. The score generation is done with the Deep Belief Network (DBN). Hence, semantic summarization is achieved by Fuzzy-based DCKN along with summarized video. Finally, the proposed semantic extraction module is directly given to the original lecture input video before summarization and analyzed the outcome with the proposed module after the summarization video.&#xa0;Here, the performance measures like Accuracy, Precision, Recall, F1-score, and Negative predictive value (NPV) used for OCR gained 92.9%, 91.0%, 92.5%, 91.7%, 90.3%, and 7.5%; semantic summarization achieved 92.8%, 91.5%, 92.2%, 91.8%, 90.2%, and 7.8%; DL methods acquired 92.6%, 91.3%, 92.5%, 91.9% and 90.8%, and 7.5%; Fuzzy-based DCKN with video summarization obtained 92.3%, 91.8%, 92.9%, 92.3%, 90.3%, and 7.1%; and Fuzzy-based DCKN without video summarization observed 91.9%, 90.86%, 91.87%, 91.36%, 89.57%, and 8.12%.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fuzzy-based DCKN: Fuzzy-based deep convolutional kronecker network for semantic analysis of summarized video

  • 期刊论文

  • 发表日期：

    2025年2月12日

查看全文
  • Preet Chandan Kaur,
  • Leena Ragha

摘要

Video summarization is a method of deducing the content of video content for generating a summary in video format. The generated summary should have the significant segments of raw video. Recently, the content of video has been rapidly increasing, thus automatic video summarization is beneficial for individuals who want to keep time and learn more in a specific time. However, online courses do not fully control the content since it has diverse open challenges in video indexing, customization requirements, and summarization for particular courses. To overcome this gap, a semantic analysis of a summarized video named Fuzzy-based Deep Convolutional Kronecker Network (Fuzzy-based DCKN) is proposed. The input lecture audio and video are carried out on video shot segmentation is done by the YCbCr space color model. Afterwards, segment the audio and video from every slot using the Honey Badger Based Bald Eagle Algorithm (HBBEA), and the features are extracted from this phase. Then, select the important segments by Deep Residual Network (DRN). Furthermore, combine the relevant audio and video to obtain the summarized video. Here, the summarized video is given to frame extraction individually, where the text content extraction and residual image extraction are accomplished by optical character recognition (OCR) and Deep Residual Network (DRN) respectively. The score generation is done with the Deep Belief Network (DBN). Hence, semantic summarization is achieved by Fuzzy-based DCKN along with summarized video. Finally, the proposed semantic extraction module is directly given to the original lecture input video before summarization and analyzed the outcome with the proposed module after the summarization video. Here, the performance measures like Accuracy, Precision, Recall, F1-score, and Negative predictive value (NPV) used for OCR gained 92.9%, 91.0%, 92.5%, 91.7%, 90.3%, and 7.5%; semantic summarization achieved 92.8%, 91.5%, 92.2%, 91.8%, 90.2%, and 7.8%; DL methods acquired 92.6%, 91.3%, 92.5%, 91.9% and 90.8%, and 7.5%; Fuzzy-based DCKN with video summarization obtained 92.3%, 91.8%, 92.9%, 92.3%, 90.3%, and 7.1%; and Fuzzy-based DCKN without video summarization observed 91.9%, 90.86%, 91.87%, 91.36%, 89.57%, and 8.12%.