Degradation of a material affects modal properties such as natural frequencies and mode shapes. These parameters show potential for global damage detection. A recent study of the propagation of flexural waves through an elastic solid has revealed the potential to determine localised cracks and in carbon fibre reinforced plastics (CFRP) delaminations. The continuous wavelet transform has provided useful insight into the damage mechanism allowing the signal to be decomposed into a series of time-frequency bands. Estimation of the group velocity of flexural waves for specific frequencies allows crack position to be identified through the analysis of the first few wave peaks. In this region waves reflected from the crack surface are clearly visible. Crack detection ability is a function of the propagating wave frequency, higher frequencies being more sensitive to small cracks. For application purposes this requires high frequency interrogation to provide sufficient resolution; in this study the acoustic range is selected (20 kHz to 1 GHz). This paper studies the identification of cracks in CFRP beams. The study seeks to identify the location of key data in the signal to provide rapid signal processing suitable for real-time application. Owing to the sensitivity of frequency data to structural changes, wavelet scales of damaged structures are correlated with those of the undamaged structure providing a measure of the changes incurred by the sample. Appropriate data reduction is undertaken for application to a neural network. Results demonstrate a high degree of accuracy in the determination of crack magnitude.