The issue of aliasing interference in tunnel vibroseis signals is caused by the limited instantaneous frequency accuracy of wavelet-based time-frequency analysis methods. To address this issue, a novel adaptive time-varying weight-based method for effective signal recognition and extraction in the time-frequency domain was proposed. The characteristic differences between sweep signals and aliasing interference in the time-frequency domain were analyzed. Time-varying gain coefficients and continuity judgment coefficients were utilized, combined with a time-window function, to achieve precise recognition and extraction of effective signals. The results show that this method effectively distinguishes sweep signals from interference components. The wavelet extraction resolution of vibroseis signals is significantly improved, and the recognition capability of weak reflection signals is enhanced. The effectiveness and practicality of the method were demonstrated through validation using 3D elastic wave numerical simulations and underground experimental data. It is concluded that this method not only overcomes the limitations of traditional time-frequency analysis methods in instantaneous frequency calculation and interference suppression but also further improves the accuracy of weak reflection signal recognition, thereby enhancing the precision and reliability of geological ahead-prospecting.