The ability to measure 3-D head kinematics in motor vehicle crash conditions is important for assessing head-neck loads as well as brain injuries. A method for obtaining accurate 3-D head kinematics of post mortem human subjects (PMHS) in short duration impact conditions is proposed and validated in this study. The proposed methodology utilizes six accelerometers and three angular rate sensors (6aω configuration) such that an algebraic equation is used to determine angular acceleration with respect to the body-fixed coordinate system, and angular velocity is measured directly rather than numerically integrating the angular acceleration. Head impact tests to validate the method were conducted using the internal nine accelerometer head of the Hybrid III dummy and the proposed 6aω scheme in both low (2.3 m/s) and high (4.0 m/s) speed impact conditions. A rear impact sled test (10.5g and 24 km/h) using the Hybrid III dummy was also conducted to validate the 6aω scheme on a tetrahedron fixture which can be installed on the head of a PMHS. For both test conditions, the 6aω method was compared with a nine accelerometer array sensor package (NAP) as well as a configuration of three accelerometers and three angular rate sensors (3aω), both of which have been commonly used to measure 3-D kinematics of the head for assessment of brain and neck injuries. The ability of each of the three methods (6aω, 3aω, and NAP) to accurately measure 3-D head kinematics was quantified by calculating the normalized root mean squared deviation (NRMSD), which provides an average percent error over time. Results from the head impact tests indicate that angular acceleration obtained from the 6aω scheme was comparable to that determined from the NAP scheme, while angular acceleration derived from the 3aω scheme was not accurate in the high speed head impact condition. Results from the rear impact sled test indicate that all three schemes (NAP, 3aω and 6aω) provide accurate linear acceleration in the body-fixed coordinate system on the head as well as in the global coordinate system, while the 6aω and 3aω scheme produce more accurate results for the angular displacement (rotation) in the global coordinate system than the NAP scheme. Overall the proposed 6aω scheme provides more accurate kinematics in the global coordinate system and a more accurate transformation matrix at each time step, since the error due to numerical integration and numerical differentiation is minimized in the transformation of head kinematics to the global (inertial) coordinate system.