In this paper, we recruited 20 participants preforming user-defined gestures on a touch screen phone for 22 TV remote commands. Totally 440 gestures were recorded, analyzed and paired with think-aloud data for these 22 referents. After analyzing these gestures according to extended taxonomy of surface gestures and agreement measure, we presented a user-defined gesture set for interactive TV remote on touch screen phones. Despite the insight of mental models and analysis of gesture set, our findings indicate that people prefer using single-handed thumb and also prefer eyes-free gestures that need no attention switch under TV viewing scenario. Multi-display is useful in text entry and menu access tasks. Our results will contribute to better gesture design in the field of interaction between TVs and touchable mobile phones.