Diagnosing excessive building vibration can be problematic because of long transmission paths. The excitation may be far displaced from the occupant thus increasing the number of potential sources. Source identification is further exacerbated by the intermittent behavior of the problematic excitation. Long time vibration records, on the order of many minutes, hours, or even days are sometimes required to capture a problematic event or the pattern of the problematic events. Analysis of the long time records with traditional spectral processing methods is usually not effective as the ensemble averaging clouds the intermittent features that are found objectionable by the occupants. Methods that rely on identifying peaks are also ineffective because a single transient peak is not typically the source of problem vibration levels. The Short Time Fourier Transform (STFT) is well suited to analyzing long records of building vibration because they provide the insight necessary to separate and identify intermittent excitation sources creating problematic building vibrations. This paper will describe the STFT in relation to analyzing long time record building vibration. The processing technique is applied to diagnose the source in a residence where the occupant complained about floor vibration even though no easily identifiable source was found.