Reduced interference distribution (Boan-Jordan Distribution), singular value abstraction and minimum distance classification is studied to recognize the signal. Simulated examples proof that BJ distribution has the approximate time-frequency centering, do better in cross-term surpression and has strong ability of distinguish signal from low SNR background. After the computation of separable measure based on distance, it can be easily find that the singular value do have the advantage of stable, reflecting the characters of matrix with limited dimension. And the character of noise averaged at the rear. The minimum distant classification and separable measure based on distance is adopted to analyze the results of recognition. At last, a sine pulse and white noise is adopted to simulize theoretically, the real signal is used to classify and recognize the targets, and a satisfactory result is got. |