Application of wavelet transforms in the detection of underground shallow cavities is investigated. Wave propagation is simulated through a transient response analysis on an axisymmetric finite element model. Cavities in a homogeneous half-space and a pavement system of a variety of shapes and embedment depths are considered. The continuous wavelet transform is introduced as a new tool for cavity detection. Effects of different types of cavities on power spectral surfaces (power spectral amplitudes versus frequency and receiver location) and Gaussian wavelet time-frequency maps (wavelet transform coefficients versus time and frequency) are studied. Results show strong energy concentration in power spectral surfaces right in front of a cavity in certain frequency bands. Time and frequency signatures of waves reflected from near and far faces of the cavity can be clearly observed in the wavelet time-frequency maps. These observations are used to locate and estimate the size of the cavity. It is demonstrated that the wavelet transform is a promising analysis tool for cavity detection and characterization.