The development of adaptive observer techniques for nonlinear systems in the output canonical form is proposed applying additional impulsive feedbacks in the observer equations. The stability of new impulsive adaptive observers is investigated. It is shown that under some conditions the proposed impulsive feedbacks can improve the rate of the observer convergence or relax the requirement on persistency of excitation, which is frequently needed to ensure convergence of the parameter estimates. The proposed feasibility conditions include positivity of dwell time (boundedness of impulse frequency) and solvability of some Lyapunov-like hybrid matrix inequalities. The results are illustrated by simulation for three examples (including a single link flexible joint robot example).