This paper proposes a parsimonious model based subspace identification (SIM) method for Hammerstein systems under completely unknown periodic disturbances. The existing over-parameterized model based SIM (OPM-like SIM) requires estimating extra parameters and performing a low rank approximation step, which may lead to high variance in parameter estimates, especially with small and noisy data sets. To address this issue, a parsimonious model based SIM is proposed, which avoids estimating extra parameters and the low rank approximation step, thereby improving the accuracy and variance properties of the parameter estimates. The proposed method uses two parsimonious models instead of the over-parameterized model to describe the Hammerstein system. An orthogonal projection based fixed point iteration method is proposed to eliminate the disturbance effects and obtain a consistent parameter estimate. The method is shown to be effective through strict mathematical proofs and simulation examples. Hammerstein systems are block-oriented nonlinear systems consisting of a zero memory static nonlinearity and a linear dynamic system. Many real-world systems can be modeled with this structure, including wireless power transfer systems, heating experimental systems, and hot rolling mill processes. Existing methods for identifying Hammerstein systems include stochastic approximation methods, robust identification methods, least-squares based methods, and subspace identification methods. The SIMs have attracted attention in the control community, with two main types: non-iterative and iterative. Non-iterative SIMs, such as N4SID-SIM, MOESP-SIM, PS-SIM, and least-squares based SIMs, use an over-parameterized model to estimate the parameters of the system. Iterative SIMs, on the other hand, use parsimonious models with fewer parameters to retrieve the original model matrices. The proposed method is shown to be effective in improving the accuracy and variance properties of the parameter estimates for Hammerstein systems under periodic disturbances. The method avoids the need to estimate extra parameters and perform a low rank approximation step, leading to improved results compared to existing OPM-like SIMs. The convergence of the proposed method is established using the contraction mapping theorem, showing that the parameter estimates can converge to the true values under arbitrary nonzero initial conditions.This paper proposes a parsimonious model based subspace identification (SIM) method for Hammerstein systems under completely unknown periodic disturbances. The existing over-parameterized model based SIM (OPM-like SIM) requires estimating extra parameters and performing a low rank approximation step, which may lead to high variance in parameter estimates, especially with small and noisy data sets. To address this issue, a parsimonious model based SIM is proposed, which avoids estimating extra parameters and the low rank approximation step, thereby improving the accuracy and variance properties of the parameter estimates. The proposed method uses two parsimonious models instead of the over-parameterized model to describe the Hammerstein system. An orthogonal projection based fixed point iteration method is proposed to eliminate the disturbance effects and obtain a consistent parameter estimate. The method is shown to be effective through strict mathematical proofs and simulation examples. Hammerstein systems are block-oriented nonlinear systems consisting of a zero memory static nonlinearity and a linear dynamic system. Many real-world systems can be modeled with this structure, including wireless power transfer systems, heating experimental systems, and hot rolling mill processes. Existing methods for identifying Hammerstein systems include stochastic approximation methods, robust identification methods, least-squares based methods, and subspace identification methods. The SIMs have attracted attention in the control community, with two main types: non-iterative and iterative. Non-iterative SIMs, such as N4SID-SIM, MOESP-SIM, PS-SIM, and least-squares based SIMs, use an over-parameterized model to estimate the parameters of the system. Iterative SIMs, on the other hand, use parsimonious models with fewer parameters to retrieve the original model matrices. The proposed method is shown to be effective in improving the accuracy and variance properties of the parameter estimates for Hammerstein systems under periodic disturbances. The method avoids the need to estimate extra parameters and perform a low rank approximation step, leading to improved results compared to existing OPM-like SIMs. The convergence of the proposed method is established using the contraction mapping theorem, showing that the parameter estimates can converge to the true values under arbitrary nonzero initial conditions.