The problem of human localization presents unique challenges, particularly in areas where GPS satellite signals may be blocked for a significant period of time. A new approach to localization based on a physical model of human motion is presented. Data is taken from inertial measurement units on the limbs, and gait analysis is used to detect and measure different modes of human motion from this data. This approach produces an accurate form of dead reckoning that is less prone to drift than current inertial navigation systems and more adaptable then current pedometry systems.