Abstract: Conventional blind source separation algorithms have poor suitability for wideband array signals. To address the problem, a novel wideband constant beamwidth blind beamforming algorithm based on time-frequency analysis is proposed. The received array signals are firstly transformed into time-frequency domain, and the single-source points are extracted. Then, the single-source points are clustered into different classes, which are used to obtain the signal steering vector at different frequencies. Finally, the wideband constant beamwidth blind beamforming is achieved by a spatial response variation constraint method without knowledge of the direction of arrival (DOA). The transformation from wideband blind beamforming to the blind separation of convolution mixtures is avoided in the proposed algorithm. Thus, the system parameters of time-domain algorithms will not increase sharply with the length of filter order, and the arbitrary permutation and scaling ambiguity of frequency-domain algorithms are also eliminated. The simulations illustrate that the proposed algorithm can well separate the wideband signals and achieve a higher output signal to interference plus noise ratio (SINR) compared with the time-domain ,frequency-domain and time-frequency-domain algorithms. The practicality of the proposed algorithm is demonstrated by the measured data.