Abstract: The response methods for acoustic target detection are typically classified into two categories: the simultaneous transmit-receive method and the time-sharing transmit-receive method. The simultaneous transmit-receive method is widely employed due to its real-time capabilities. However, the transmitted signal may also be captured by the receiving transducer after propagating through the shell and water, leading to acoustic crosstalk. This paper focuses on the nonlinear interference present in the acoustic crosstalk path and proposes a convex combination algorithm based on the normalized least mean square (NLMS) filtering structure and the Chebyshev nonlinear filtering structure. The joint parameter is optimized to reduce the error at the upper and lower boundaries, thus achieving rapid convergence and a low steady-state error for nonlinear systems. When processing single-frequency signals and frequency modulation signals, compared with the NLMS algorithm, this proposed algorithm improves the acoustic crosstalk cancellation performance by 6 dB and 8 dB, respectively. In comparison with the Chebyshev nonlinear filtering algorithm, it reduces the number of iterations required to reach steady state by 33% and 20%, respectively. The results indicate that the algorithm demonstrates superior modeling ability and faster convergence speed in addressing the nonlinear characteristics of acoustic crosstalk paths, effectively mitigating the nonlinear issues prevalent in acoustic crosstalk phenomena.