A method for determining non-contact gesture is applied to a device for the same. The device has an image sensor to detect an image sensing data, an inertial measurement sensor to detect an inertial sensing data, and a processor to determine if an object data is detected and to determine if an inertial event is occurred. The image sensing data in the same and/or adjacent image frame with the inertial events are excluded so that the image sensing data used to determine the gesture are those not influenced by the inertial events. Therefore, the accuracy of determining the gesture is enhanced.