The invention is a digital video-audio playback device and the playback control method thereof, using user-provided specific time point or key word as playback reference of digital video-audio file. If captions exist in image form, it converts image captions into character contents; or if the captions are stored in character code form, it resolves the needed contents. The character contents resolved/converted can compare with the key word, resolving proper caption corresponding time which can act as start/stop point for setting video-audio playback; or the comparative result can be used to analyze relativity between video-audio file and key word.