Gait analysis can convey important information about one's physical and cognitive condition. Wearable inertial sensor systems can be used to continuously and unobtrusively assess gait during everyday activities in uncontrolled environments. An important step in the development of such systems is the processing and analysis of the sensor data. This paper presents a symbol-based method used to detect the phases of gait and convey important dynamic information from accelerometer signals. The addition of expert knowledge substitutes the need for supervised learning techniques, rendering the system easy to interpret and easy to improve incrementally. The proposed method is compared to an approach based on peak detection. A new symbol-based symmetry index is created and compared to a traditional temporal symmetry index and a symmetry measure based on cross correlation. The symbol-based symmetry index exemplifies how the proposed method can extract more information from the acceleration signal than previous approaches.