Compared with singular value decomposition, symplectic geometry spectrum is a measure preserving and nonlinear transform. So, it is more suitable for nonlinear dynamics system analysis. A new method to detect determinism in time series based on symplectic geometry spectrum (SGS) is proposed in the present work. Chaos and stochastic process could be recognized by applying the non_parameter Mann_Whitney on the SGS of original data and its surrogate data. The method is first tested on stochastic processes, the Lorenz, Rossler, Mackey_Glass and high dimensional coupling equations. Then it is applied to two data sets of Santa Fe to test its effect on experimental data. Finally, the robust ness of the method is tested on the time series with different data length and different levels of additive noise.