An intangible writing interface allows a user to write in the mid-air with a finger.Some challenging issues would arise because all written characters would be connected in a single motion trajectory of the finger called a one-stroke finger gesture.The invisibility and space limitation of the valid writing area in the mid air usually cause a user to write characters with an irregular text layout that may contain arbitrarily connected and highly superimposed characters.Targeting at the numeral recognition in an intangible writing interface, we tackle the challenging issue of segmenting and recognizing the arbitrarily connected and superimposed numerals in the one-stroke finger gestures.To our best knowledge, no existing work has addressed this problem before.Our solution comprises (1) a simple but effective method to spot any specific key numeral within arbitrary one-stroke finger gestures and (2) a graph path finding algorithm to simultaneously segment and recognize the connected numerals by finding the best scheme to concatenate the spotted key numerals.The performance evaluation shows that the proposed solution achieves satisfactory recognition accuracy, even without exploiting sophisticated features for characterizing handwritten numerals.