Building floor accelerations recorded during earthquakes represent valuable data which contribute to the improved understanding of structural dynamic behavior and the state of the art and practice. After recording, all accelerations are considered to be "raw" (or unprocessed, uncorrected, unfiltered), and need to be processed before they can be used. In practice, the most common approach is to apply the conventional high-and low-pass filtering methods. However, some records cannot be corrected in such a manner and remain practically useless. The problems which may occur when conventional filtering methods are applied are: (1) inability to distinguish frequencies of noise and signal, (2) change of noise characteristics in time, and (3) a need to estimate and pre-select the parameters of a chosen filter function. These issues can be avoided if wavelets are used for signal processing. In mathematical terms, wavelets are finite wave-like functions which are able to capture local changes, and are used to transform a signal into a representation which shows its properties in the timefrequency domain. Such transformation is known as a wavelet transform. The continuous wavelet transform (CWT) was applied to recordings obtained from a 3-storey reinforced concrete building shake table test performed at the E-Defense facility in Kobe, Japan. It was demonstrated that the CWT represents a useful tool for the visual interpretation of the energy localization in the time-frequency domain, and that there is a large potential for the CWT application in earthquake engineering.