This thesis considers how the position of a magnetic levitation train can be found non-optically using signal processing techniques. An overview of magnetic levitation trains is given, including an explanation of the necessity for non-optical position sensing. A least-squares with forgetting factor algorithm is derived, showing its convergence and stability. The input signal is manipulated so that the least-squares algorithm can be applied. Extremum seeking is introduced and utilized for phase estimation of the high frequency signal. When adding noise to the incoming signal, the frequency estimates remain accurate if the forgetting factor is allowed to vary proportional to the slope of the estimate. Additive noise has little effect on the extremum seeking phase estimate. The frequency and phase estimation is combined to reconstruct an estimate of the carrier signal to demodulate the signal and extract the position of the train