Indoor localization systems aim to accurately and cost-effectively locate targets. While numerous indoor localization solutions based on Wi-Fi, Bluetooth, ZigBee and other technologies have been proposed, they fail to achieve satisfactory performance. The accuracy of these solutions is often affected by obstacles such as shelf, human etc; due to the hight frequency of the signals, which weaken the ability of signals to penetrate obstacles. Moreover, high accuracy comes at the expense of more hardware, labor-expensive deployment. To overcome these limitations, we propose an indoor localization system based on Software Defined Radio (SDR), using FM and DVB-T signals. Our system achieves sub-meter accuracy at low cost with low site survey overhead. Additionally, we investigate the temporal and window effect on different sizes of training data.