Abstract: In this paper, the applicability of biomimetic pattern recognition to informa-tion processing of abstract objects is studied, and then a DOA estimation method based on biomimetic pattern recognition is advanced. In this method, the output signals of an-tenna array are collected in practical conditions and expressed as feature vectors. These feature vectors are studied with the method of high dimensional geometry and the principle of biomimetic pattern recognition. By using the feature vectors as training samples ANN models are constructed. In our experiments, when SNE, is 20dB or OdB, the correct estima-tion rate is 100%; when SNR is -20dB, the correct estimation rate is 83%. The experimental results show that the proposed method has the great advantage of preferably robust and fast computation.