We present an application of a microelectromechanical systems (MEMS)-based inertial measurement unit (IMU) to thigh angle estimation for the control of a semi-active knee prosthesis. By using inertial based measurement devices, we can move all of our sensors to one location inside the prosthetic leg. These sensors are challenging to use, however, because error in their measurements will accumulate over time. We show that by applying the Kalman filter to an IMU consisting of a dual-axis accelerometer and a singleaxis gyroscope, we can decrease the integral of squared error of our thigh angle estimates by 70%.