Video Information Retrieval (1)

• Three basic approaches:
  • Speech recognition of audio track, optical recognition of captions and transcript
  • Segment video into individual shots, extract key frames from each shot, index key frame as QBIC
  • A combination of the above 2 approaches
• Young,, et al, "Video Mail Retrieval Using Voice," 1996
• 1994, allow voice retrieval by specific users with 35 predefined keywords
• 1995, allow non-specific users for voice retrieval
• 1996, non-specific users, non-specific keywords
• The key technologies is simplified speech recognition, only 45 syllables are recognized, not (thousands of) words

前一張

下一張

回到第一張

瀏覽圖案部份