Automatic timeline generating method specific to film and television videos

The invention discloses an automatic timeline generating method specific to film and television videos. The method comprises the following steps: (1) video scene analysis: i, analyzing sub-scenes through video lens analysis; and ii, clustering the sub-scenes into a large scene through phonetic feature analysis: distinguishing noise from voice in continuously-occurring sub-scenes through zero-crossing rates and short-time energy, comparing the average zero-crossing rates and short-time energy of the continuous sub-scenes determining that two sub-scenes do not belong to one large scene if the difference between the two sub-scenes is within a preset threshold, and clustering the sub-scenes into the large scene; (2) voice timeline splitting in the scene; and (3) splitting of a whole video timeline, wherein the whole video timeline can be obtained by combining timeline data detected in different scenes. By adopting the automatic timeline generating method specific to the film and television videos disclosed by the invention, voice fragments can be extracted automatically, and subtitle timeline is generated.