Fingerprint Identification is one of the well known biometrics. Because of their uniqueness and consistency over time, fingerprints have been used for identification for over a century. Recently they have become automated due to advancements in computing capabilities. The authors propose a method in which original fingerprint image is divided into blocks of multiples of two and mean of every block wavelet coefficients are computed and stored in neural network database. With this proposed algorithm, rotation invariant property is observed with real wavelets which are normally rotation variant. Complex wavelets need to be used to retrieve the fingerprint image where the property of rotation invariant is required. To evaluate the efficiency of the proposed algorithm in fingerprint authentication FAR (False Accept Ratio) and FRR (False Reject Ratio) are measured.