The paper proposes a time-frequency ridge index algorithm for gearboxes under variable speed conditions， based on Dy? namic Path Planning of Barycenter （DPPB）. This algorithm addresses the challenge of estimating the instantaneous frequency of signals in a high-noise environment. The algorithm builds upon the analysis of the Multi-Path Matching Pursuit （MMP） ridge in? dex algorithm and its limitations under high noise. By adding windows to the ridge set obtained by the MMP algorithm， a ridge barycenter sparse matrix of the signal is constructed. A dynamic path planning function is then designed for the barycenter sparse matrix to index the barycenters on the ridge line. The optimal time-frequency ridge line is calculated based on the values of the ridge line cost function. The similarity coefficient Ra and confidence σRa are used as measures of the ridge extraction effect. Simulations and experiments indicate that the DPPB algorithm can effectively extract the time-frequency ridge of signals in high-noise environ? ments， and it is more reliable and robust than the peak index algorithm and the MMP algorithm under various noise intensities.