Time-interleaved analog-to-digital converters (TIADCs) are widely used in multi-Gigabit communication systems due to their high conversion rates. In practice, the offset, gain, and time-skew mismatch of TIADC significantly degrade the performance of the OFDM system and even make the convolutional coding system works worse. In this paper, we propose a two-stage background joint mismatches calibration using Least-square estimation to detect the offset and gain mismatch, and the time-skew mismatch is estimated by golden-section search. This method improves the reliability of the OFDM-TIADC system. The simulation result shows a better performance for the OFDM-TIADC system with / without the coded system.