Nowadays components made of unidirectional composite materials are largely diffused in many engineering fields, such as automotive, railways, marine and aerospace. Main drawback of this class of materials lies in their low out-of-plane properties making them very sensible to impulsive loads such as impact events. After a collision with an external object, composites structures could be affected by damage, sometimes not visible from visual inspections (barely visible damage) hence with detrimental consequences on structure resistance and strength. Therefore, it is fundamental in terms of safety to continuously assess the healthy state of structures during their life and determine whether an impact event has occurred and if it caused damage or not. This work proposes a baseline free methodology to determine the coordinates of very high velocity impact on complex structures and evaluate if damage has occurred during the impact by only acquiring signal during the impact event. The technique overcomes the common limitations of previous technique presented in literature, i.e. a priori knowledge of mechanical properties, vibration response analysis, wave propagation direction dependency, sensor locations. The routine developed is based first on the estimation of the power of the acoustic emission generated by impact events, at sensors location, then the power information through the entire structure is reconstructed exploiting radial basis function network. The actual impact estimation is finally obtained using a weighted method. Furthermore, damage assessment is conducted with a novel method based on Hilbert-Huang transform and mode decomposition. Experimental tests were performed on a generic carbon fibre reinforced polymers blade specimen with a complex stacking sequence and embedded sensors. Two test configurations at different velocities were considered: one at 90 m s-1 and one at 190 m s-1. Before and afterwards the actual impact tests, the blade was excited as well with a modal hammer (pre and post impact). The results from the impact analysis highlighted the validity and reliability of the proposed method, with a high level of accuracy in terms of impact localisation estimation, and qualitative integrity state was effectively evaluated.