An underfloor accelerometer sensor network can be used to track occupants in an indoor environment using measurements of floor vibration induced by occupant footsteps. To achieve occupant tracking, each footstep impact location must first be estimated. This paper proposes a new energy-based algorithm for footstep impact localization. Compared to existing energy-based algorithms, the new algorithm achieves higher localization accuracy and removes a previously required calibration step (removal of the need to estimate floor-dependent parameters). Furthermore, the algorithm uses a much smaller data sampling rate compared to time of flight/arrival localization methods, which greatly reduces data and data-processing time. The new algorithm is a two-step location estimator: the first step is a coarse location estimate, with the second step as a fine location search through a nonlinear minimization problem. The performance of the proposed algorithm is evaluated using a single occupant walking experiment on an instrumented floor inside an operational smart building. This paper also demonstrates that higher localization accuracy is obtained using an additional Kalman filtering scheme.