Biometrics plays an important role in security systems. This paper presents an efficient approach of using ECG signal for biometric identification system. A linear prediction coding based algorithm is used to extract features from pre-processed ECG signal where the preprocessing has been done by ensemble empirical mode decomposition (EEMD). Instead of using ECG signal directly, reflection coefficients extracted from the signal using linear predictive coding (LPC) method further used to perform identification task, providing high level of correct identification and compactness. For classification, different distance methods are used on database performing high rate of identification. The proposed method has been tested on MIT-BIH Normal Sinus Database, MIT-BIH Arrhythmia Database, PTB Diagnostic Database and self-recorded ECG signals which provided small data samples with high rate of correct identification.