Abstract: ‍ ‍Ultra-short baseline has the advantages of small array size and easy layout， but there are problems such as poor localization accuracy due to too small array aperture. This paper proposed an ultra-short baseline underwater source localization method based on wideband higher-order subspace， which improved the scalability of the localization method. The method was mainly divided into three steps： 1） Construct the array wideband signal receiving model according to the ultra-short baseline array； 2） Use the higher-order cumulant subspace method to virtually expand the number of array elements and increase the array aperture， so as to improve the accuracy of azimuth and pitch angle estimation under the ultra-short baseline three-dimensional array structure； 3） According to the time relationship between the emitted signal wave and the receiving signal wave， the slant distance of the underwater source is estimated by merging the matched filter algorithm. On this basis， this paper introduced an adaptive Kalman filter model to weaken the noise interference for the localization distortion problem caused by the dynamic attitude adjustment of underwater sources， and proposed a wideband higher-order subspace dynamic source localization method based on adaptive Kalman filter. Experiments show that the method proposed in this paper has better resolution for underwater source localization and better robustness to noise. Compared with the current ultra-short baseline underwater source localization method based on delay difference and phase difference， it has better localization accuracy. And in the dynamic target localization scenario， the method proposed in this paper has better smoothness and fits more closely with the theoretical value， which also reflects that the dynamic localization method in this paper has better accuracy in tracking underwater source trajectories.