This paper proposes a track before detect algorithm based on particle filter to address the problem of continuous-wave signal tracking under the background of underwater acoustic doubly-spread channels. It focuses on improving the performance of the weak signal tracking under nearby multipath signal interference. First of all, a likelihood function which describes multipath signal interference based on the nearby signal predicted state is proposed. Secondly, the measurement set of velocity is introduced in the measurement space, and the analytical formula of the velocity-correlation envelope joint likelihood function is derived. Then the multi-measurement input signal track before detect model is constructed based on the joint likelihood function. Finally the robust signal tracking results are obtained.The results of numerical simulation and pool experiment show that under typical underwater acoustic time-varying doubly-spread channels, the proposed algorithm has better performance of detecting and tracking multipath signal.