This thesis presents a real-time hand-gesture recognition algorithm suitable for implementation on low-power microcontroller units (MCU) for wearable devices with wireless transmission capability. The algorithm performs string matching against a set of patterns that symbolize acceleration and motion directions. Experimental results show that our algorithm can recognize and transmit characters in the Graffiti alphabet and gestures with high accuracy while keeping complexity low enough to be feasible to run on an 8-bit microcontroller unit.