Delamination and decay are common structural defects in old glued laminated timber (glulam) buildings, which, if left undetected, could cause severe structural damage. This paper presents a new damage detection method for glulam inspection based on moment analysis and wavelet transform (WT) of impact acoustic signals. Acoustic signals were collected from a glulam arch section removed from service through impact testing at various locations. The presence and positions of internal defects were preliminarily determined by applying time centroid and frequency centroid of the first moment. Acoustic signals were then decomposed by wavelet packet transform (WPT) and the energy of the sub-bands was calculated as characteristics of the response signals. The sub-bands of 0–375 Hz and 375–750 Hz were identified as the most discriminative features that are associated with decay and delamination and therefore are indicative of the presence of delamination or decay defects. A defect diagnosis algorithm was tested for its ability to identify internal decay and delamination in glulam. The results show that depth of delamination in a glulam member can be determined with reasonable accuracy.