An accelerometer is installed in most current mobile phones, such as iPhones, Android-powered devices, and video game controllers for the Wii or PS3, which enables easy and intuitive operations. Therefore, many gesture-based user interfaces that use accelerometers are expected to appear in the future. Gesture recognition systems with an accelerometer generally have to be models constructed with a user's gesture data before use, and they need to recognize any unknown gestures by comparing them with an output of the recognition result and feedback delays since the recognition process generally starts after the gesture has finished, which may cause users to retry gestures and thus degrade the interface usability. We propose an early stages gesture recognition method that sequentially calculates the distance between the input and training data, and outputs recognition results only when one output candidate has a stronger likelihood than the others. Gestures are recognized in the early stages of a given motion without deteriorating the level of accuracy, which improves the interface usability. Our evaluation results indicated that the recognition accuracy approached 1.00 and the recognition results were output 1, 000 msec on average before a gesture had finished.