Motor vehicle crashes claim over 40,000 lives and injure over two million people each year in the United States. To reduce the number of injuries and fatalities through vehicle design improvements, it is important to study occupant kinematics and related injury mechanisms during crashes. Occupant motion in crash tests is typically measured with high speed video, spatial scanning, direct field sensing, and inertial sensing. In this work, we present simulation and testing results on inertial sensing of dummy kinematics based on a novel algorithm known as Quaternion Fuzzy Logic Adaptive Signal Processing for Biomechanics (QFLASP-B). This approach uses three angular rates and three accelerations (one gyroscope-accelerometer pair about each axis) per rigid body to compute orientations (roll, pitch and yaw), positions and velocities in the inertial (fixed) reference frame. In QFLASP-B, quaternion errors and gyro biases are calculated and used in an adaptive loop to remove their effects. The Fuzzy Estimator at the core of the algorithm consists of a fuzzification process, an inference mechanism, a Rule Base and a defuzzification process. In this paper, we examine those aspects of the QFLASP-B Fuzzy Estimator critical to accurate kinematics sensing, hardware and software implementations and experimental results compared with traditional approaches.