Abstract: Mixed LFM signal is everywhere in signal environments in practice, it is important to recognize it for electronic intelligence. The cross-terms of Wigner-Ville distribution resulting from mixed signals reduce its resolution. A new approach of eliminating the cross-terms based on independent component analysis is presented. The independent signal components were extracted via blind source separation. The cross-terms were reduced significantly based on joint diagonalisation of the time-frequency distribution matrices. The reconstructed Wigner-Ville distribution has good resolution. Each LFM signal components are recognized from mixed signals using Wigner-Hough transform. The relation between input SNR and output SNR is computed, the simulation results show that the new approach improves recognition performance and WHT can enhance the output SNR when sampling number is increased.