The paper presents health indices (HI) for monitoring and prognostics of machine condition. HI are developed using morphology and entropy based complexity measures of machine vibration signals. The indices are compared with a recently introduced energy based feature and the commonly used statistical measure of signal kurtosis. The procedure of extracting HI is illustrated first using the simulated response of a simple gear model with tooth crack. Next the HI extraction process is applied to the experimental vibration data of a helicopter drivetrain gearbox with a seeded tooth fault. The effectiveness of the extracted HI is compared for gear condition monitoring and prognostics.