The energetics of human motion has been intensely studied using experimental and theoretical methods. Knowing the kinetic energy of the human body, and its decomposition into the kinetic energies of the major body segments, has tremendous value in applications ranging from physical therapy, athlete training, soldier performance, worker health and safety, among other uses. Significant challenges thwart our ability to measure segmental kinetic energy in real (non-laboratory) environments such as in the home or workplace, or on the playing/training field. The aim of this research is to address these challenges by advancing the use of an array of miniaturized body-worn inertial measurement units (IMUs) for estimating segmental kinetic energy. As a step towards this goal, this study reports a benchmark experiment that demonstrates the accuracy of IMU-derived estimates of segmental kinetic energy. The study is conducted on a well-characterized mechanical system, a double pendulum that also serves as an apt model for the lower or upper extremities. A two-node IMU array is used to measure the kinematics of each segment as input to the segmental kinetic energy computations. The segments are also instrumented with two high-precision optical encoders that provide the truth data for kinetic energy. The segmental kinetic energies estimated using the IMU array remain within 3.5% and 3.9% of the kinetic energies measured by the optical encoders for the top and bottom segments, respectively, for the freely decaying pendulum oscillations considered. These promising results support the future development of body-worn IMU arrays for real-time estimates of segmental kinetic energy for health, sports and military applications.