A method for estimating gait parameters (shank, thigh, and stance leg angles) from a single, in situ, scalar acceleration measurement is presented. A method for minimizing the impact of errors due to unpredictable variations in muscle actuation and acceleration measurement biases is developed. This is done by determining the most probable gait progression by minimization of a cost function that reflects the size of errors in the gait parameters. In addition, a model for gait patterns that takes into account their variations due to walking speed is introduced and used. The approach is tested on data collected from subjects in a gait study. The approach can estimate limb angles with errors less than 6 deg (one standard deviation) and, thus, is suitable for many envisioned gait monitoring applications in nonlaboratory settings.