A new joint distribution model was proposed by using the conditional probability method, and a new expression for the joint probability density function of wave height period was given, in order to accurately describe the joint distribution of wave high and period for short-term wind seas. Only two parameters were contained in the expression, the significant wave height and the mean cross zero period, which facilitated engineering applications. Taking the Pierson-Moscowitz (P-M) spectrum, the Bretschneider spectrum and the measured wave spectrum as the target spectrum, the short-term irregular waves were modelled as a sum of cosinoidal wave components. An empirical distribution of wave high and period was obtained through statistical analysis. The empirical joint distribution was used as the benchmark, and the proposed joint distribution models and five joint distribution models in the existing literature were compared and analyzed. The root-mean-square error was used to evaluate the degree of matching between all models and the empirical distribution. The results of the root-mean-square error indicate that the proposed model is closest to the empirical distribution in the all models. The difference between the models is mainly reflected in the distribution of period. Among all the models, the distribution of period derived from the proposed model is also closest to the empirical distribution. The proposed joint distribution is expressed by an explicit closed-form formula, it can be easily extended to non-Gaussian wave cases according to transform Gaussian methods such as Hermite transform and Rychlic transform.