| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To highly accurately classify non-compressed or compressed moving image data into various classes (shot classes) at a low cost by using features as a moving image or features of audio accompanied with the moving image as needed. <P>SOLUTION: The non-compressed or compressed moving image data and audio data accompanied therewith are divided into shots by a shot dividing part 1 and inputted to a motion information analytic part 2 and an audio analytic part 3. In the motion information analytic part 2, a motion strength I<SB>S</SB>of each shot is detected and in the audio analytic part 3, an audio power P<SB>S</SB>and an audio class C<SB>S</SB>are detected. An action class discriminating part 4, a dramatic class discriminating part 5 and a conversation class discriminating part 6 utilize a shot length L<SB>S</SB>, the motion strength I<SB>S</SB>and the audio power P<SB>S</SB>as well as the audio class C<SB>S</SB>, as needed, to discriminate an action class, a dramatic class, a conversation class and a general class. <P>COPYRIGHT: (C)2004,JPO&NCIPI</p> |
