Hand gesture interfaces have been proposed as an alternative to the remote controller, and products with such interfaces have appeared in the market. We propose the vision-based unicursal gesture interface (VUGI) as an extension of our unicursal gesture interface (UGI) for TV remotes with touchpads. Since UGI allows users to select an item on a hierarchical menu comfortably, it is expected that VUGI will yield easy-to-use hierarchical menu selection. Moreover, gestures in the air such as VUGI offer an interface area that is larger than that provided by touchpads. Unfortunately, since the user loses track of his/her finger position, it is not easy to input commands continuously using VUGI. To solve this problem, we propose the dynamic detection zone and the detection axes. An experiment confirms that subjects can input VUGI commands continuously.