A new method based on the theory of set-membership estimation for time-domain identification of helicopter flight dynamics model has been developed. The parameters in most helicopter flight dynamics models are generally coupled, and it is difficult to obtain an explicit-function relationship between an identification model and the parameters to be identified. An indirect-identification algorithm should be established. The concept of generalized noise is introduced to solve the problems using the set-membership-identification method to construct a helicopter flight dynamics model. The identification of a six-degree-of-freedom linear flight dynamics model of the BO-105 helicopter in hover condition based on the new method is implemented. A comparative study of the new method with the maximum-likelihood method is also performed. The obtained results show that the new method increases the identification accuracy, efficiency, and robustness to noise by comparison with a conventional statistic-based identification method.