Time-frequency analysis is combined with array processing to develop a direction of arrival (DOA) estimation method. The array data model is constructed in time-frequency domain by cross time-frequency distribution between the output of a reference sensor and those of two symmetric sub-arrays. Accordingly a subspace method is presented based on the average of two sub-arrays’ time-frequency data vector model instead of the conventional array model, to estimate DOAs of multiple signals. Because the array data is processed both in spatial domain and 2-D time-frequency domain, the proposed method has an ability to select the signal of interesting, and is suitable for non-stationary signal. Additionally, the method is robust to noise and holds an advantage of low computational load. Simulations are conducted to verify the efficiency of the method and comparision is made with other methods. Time-frequency analysis is combined with array processing to develop a direction of arrival (DOA) estimation method. The array data model is constructed in time-frequency domain by cross time-frequency distribution between the output of a reference sensor and those of two symmetric due to the subspace method is presented based on the average of two sub-arrays’ time-frequency data vector model instead of the conventional array model, to estimate DOAs of multiple signals. Because the array data is processed both in the spatial domain and 2-D time-frequency domain, the proposed method has an ability to select the signal of interesting, and is suitable for non-stationary signal. conducted to verify the efficiency of the method and comparision is made with other methods.