The goal of this study is to assess the possibility of accurate on-line instantaneous velocity estimation in swimming. Having an on-line tool, coaches could provide immediate feedback about performance to trainees. More importantly, by on-line monitoring of velocity anomaly in open-water swimming, the safety of events can be significantly improved. We have previously introduced a method, using a wearable IMU, to estimate swimming instantaneous velocity, though information about pool length and a complete lap data were needed to correct the integration drift of IMU signals. In the present study, we used our previous algorithm for cycle’s mean velocity estimation, as a criterion for drift correction in instantaneous velocity estimation without the knowledge about pool length. Using a simple within-cycle linear drift model, the relative error of the algorithm tested on 8 swimmers is 0.1±15.4%. As a result, the instantaneous velocity is available at the end of every cycle.