A keyboard gesture instruction generating method includes: loading an instruction condition definition table, where the instruction condition definition table includes a plurality of gesture instructions, and each gesture instruction includes a plurality of instruction conditions, and each instruction condition has a corresponding defined value; reading an operating position image obtained by a sensor; converting the operating position image into an operating position data, where the operating position data includes relative position data of keys which are referenced; recognizing a gesture event according to multiple sets of the successive operating position data; obtaining a plurality of detected values of the gesture event and comparing the detected values with these defined values; if these detected values match with the defined values of all the instruction conditions of one of the gesture instructions, stopping tracking of the gesture event; and executing the matching gesture instruction.