The gyroscope was one of the core components of the inertial navigation system (INS), whose residual lifetime distribution prediction of INS was of great importance in operational use, combat readiness test and maintenance decision making. First, the gyroscope was modeled in its degradation process with the drift Brownian motion. In view of the small sample of the gyroscope performance parameters, the Bayesian Bootstrap method was introduced into the degradation model of parameter estimation. This method extended the amount of samples by numerical simulation. As the credibility of Bayesian Bootstrap method estimation fell for the particularly small sample, this paper put forward the improved Bayesian the Bootstrap method. It shortened the confidence interval and effectively improved the prediction accuracy in ensuring the invariance of the expectation and variance point estimation. At last the effectiveness and practicability of the proposed method was proved through the gyroscope life prediction.