The development of intelligent ambulatory monitoring systems and smart living environments is important when considering the aging of society and its implications. This work concerns the use of human motion analysis as a tool for supporting elderly life. Movement recognition has so far been achieved through some form of template matching after manual segmentation or modeling of important features. However, previous works have failed to generalize movement and have only been able to recognize few predetermined activities. To cope with those limitations, this work suggests a new “motion language” approach. To demonstrate the viability and usefulness of this methodology, the concept of “motion primitives” was used to quantitatively analyze gait unsteadiness, which relates to physical condition and cognitive performance. The variability of stride time and temporal walk symmetry between the two feet were measured. Accelerometers were chosen as motion sensors since they offer desirable features in monitoring human movements such as response to both movement frequency and intensity, miniaturization and low power consumption. This study shows that a motion language methodology is capable of quantitatively measuring temporal gait characteristics and providing tools for continuous, unobtrusive, home-based gait analysis.