A new chaos-based technique for modelling the diversity of approximately periodic signals is introduced and exploited, combined with generalized chaotic synchronization phenomena, for the solution of temporal pattern recognition problems