If the opposing force is cyclic with rotor position such that the pulsed or steady state current in each position becomes sinusoidal, then it is possible to sense DC current or emission level of transient current pulse width during commutation with a Narrow band AM radio but this will become unreliable if speed or load is changed.
Direct current sensing would be more accurate for predicting/deconvoluting sine position but only if load is strongly opposing gravity or some sine position spring force on the rotor belt.
Otherwise your assumptions are invalid. Stepper current rise time depends on L/(RdsOn+DCR)=Tau and not the rotor mechanical load that might vary with sine position. However the time duration and oscillation levels may vary to reach the next step which means the amplitude of the frequency that corresponds to the step motion pulse width , which is a fraction of step interval or large harmonic of step rate but well below step risetime. This intermediate bandwidth might possibly be correlated with position but skewed by noise of many factors.
Start by scoping DC current and Spectrum Analyzer noise in wide band mode using a time sweep rather the frequency sweep and tune until you get a strong cyclic sinusoid with rotation and expect you need > 40dB SNR at all times to resolve 50:1.
You may need to detect surge edge to enable a sample & hold to capture the signal of interest accelerating the sinusoidal load so as to ignore the noise, which is greater but contains unimportant signal. The greatest current (10x) is when voltage is applied when the stepper has not reached significant speed yet L/R has been exceeded. Then motion /velocity-induced BEMF reduces the current until it stops.
The best rotor position sensing uses a Home or 0 deg or position 000 optical sensor and then up/down counter on quadrature steps or micro steps. This gives the best chance of a calibrated position sense.