Cut point detection system and short recognition system using the same, cut point detection method and cut point detection program
View Patent ↗The MBT information for each frame is extracted from the moving picture compressed data containing the MBT information, an MBT symbol string is created by obtaining the frequency at which a specific MBT is contained from the extracted MBT information for each frame, appending a predetermined MBT symbol to each frame in accordance with its frequency feature, and arranging the MBT symbols according to the frame order, the cut point of the moving picture is discriminated from an array feature of MBT symbols in the MBT symbol string, and further a shot is identified based on an estimation model using the MBT symbol string.
1. A cut point detection system comprising:
means for extracting macroblock type (MBT) information for each frame from moving picture compressed data comprising macroblocks;
means for creating an MBT symbol string comprised of a series of MBT symbols, one for each frame, wherein each MBT symbol represents a frequency feature of specific MBTs in a frame, wherein the frequency feature indicates a frequency at which a specific MBT occurs within the extracted MBT information for each frame, appending a predetermined MBT symbol to each frame in accordance with its frequency feature in order to arrange the MBT symbols according to the frame order; and
means for discriminating the cut point of moving picture horn an array feature of MBT symbols in said MBT symbol string.
2. The cut point detection system according to claim 1 , wherein said MBT symbols include a first MBT symbol indicating that the frequency of forward predictive coded MBT is largest, and a second MBT symbol indicating that the frequency of backward predictive coded MBT is largest, and the cut point of moving picture is discriminated from an array feature of said first MBT symbol and said second MBT symbol in said MBT symbol string.
3. The cut point detection system according to claim 2 , wherein said first MBT symbol indicates that the frequency of forward predictive coded MBT is largest and a total number of backward predictive coded MBT and intra-coded MBT is smaller than a predetermined threshold value.
4. The cut point detection system according to claim 2 , wherein said second MBT symbol indicates that the frequency of backward predictive coded MBT is largest and a total number of forward predictive coded MBT and intra-coded MBT is smaller than a predetermined threshold value.
5. The cut point detection system according to claim 2 , wherein when said two first MBT symbols are adjacent in said MBT symbol string, the cut point occurs immediately before the frame corresponding to the top MBT symbol of them.
6. The cut point detection system according to claim 2 , wherein when said two second MBT symbols are adjacent in said MBT symbol string, the cut point occurs immediately after the frame corresponding to the last MBT symbol of them.
7. The cut point detection system according to claim 2 , wherein when said first MBT symbol and said second MBT symbol are adjacent in frame order in said MBT symbol string, the cut point occurs between two frames corresponding to said two MBT symbols.
8. A shot identification system using the cut point detection system according to claim 1 , further comprising:
means for making a shot for an identification object between two cut points detected by said cut point detection system; and
means for selecting and identifying said shot of identification object of which the probability is largest based on an estimation model using the MBT symbol string created by said cut point detection system.
9. The shot identification system according to claim 8 , wherein said estimation model selects beforehand a specific event shot from the moving picture compressed data containing an event shot that is a shot to be identified, creating said MBT symbol string for each specific event shot, and learning said MBT symbol string as a learning object MBT symbol string for each event shot.
10. The shot identification system according to claim 8 , wherein said estimation model is a hidden Markov model.
11. A cut point detection method including:
a step of computer extracting macroblock type (MBT) information for each frame from moving picture compressed data comprising macroblocks;
a step of computer creating an MBT symbol string comprised of a series of symbols, one for each frame, wherein each MBT symbol represents a frequency feature of specific MBTs in a frame, wherein the frequency feature indicates a frequency at which a specific MBT is occurs within the extracted MBT information for each frame, appending a predetermined MBT symbol to each frame in accordance with its frequency feature in order to arrange the MBT symbols according to the frame order; and
a step of computer discriminating the cut point of moving picture from an array feature of MBT symbols in said MBT symbol string.
12. A non-transitory computer readable storage medium storing instructions for cut point detection, the instructions comprising functionality to:
extract macroblock type (MBT) information, for each frame from moving picture compressed data comprising macroblocks;
create an MBT symbol string comprised of a series of symbols, one for each frame, wherein each MBT symbol represents a frequency feature of specific MBTs in a frame, wherein the frequency feature indicates a frequency at which a specific MBT occurs within the extracted MBT information for each frame, appending a predetermined MBT symbol to each frame in accordance with its frequency feature in order to arrange the MBT symbols according to the frame order; and
discriminate the cut point of moving picture from an array feature of MBT symbols in said MBT symbol string.