A sense resistor implies that current is being dropped across the resistor, which equates to power loss. If a very low-value resistor is used (to avoid power loss), it becomes hard to sense the very tiny voltages generated, so then amplification and filtering is needed. On an electrically noisy H-bridge, this becomes challenging.
An unusual way to detect current (pulses) with minimal invasiveness may be to add a turn or two of the H-bridge power wire through a donut toroid inductor, forming a transformer. This is most often used with AC loads. Then rectify and filter this AC voltage generated across the inductor. Since there will be more turns on the inductor (secondary) than the wrap (primary), it will "step up" the primary current to a higher secondary voltage. The higher this (filtered DC) voltage becomes during operation, the more current must have been pulsing in the primary (H-bridge.)
simulate this circuit – Schematic created using CircuitLab
As the duty approaches 100% though, the pulses stop (motor is driven with uninterrupted DC) and this concept won't work any longer; transformers can't operate at DC. There is only one practical way to detect static DC magnetic fields:
(Courtesy Allegro Microsystems)
This would work for DC; a hall-effect sensor. Find a slotted (or "gapped") ferrite ring, or two halves of a ring. Wrap the primary a few turns (more turns = more flux = more sensitivity) and stick the hall-effect sensor into the gap. The toroid will focus essentially all of the magnetic flux into the sensor, vastly increasing sensitivity. Then sample the (rapidly changing) hall output and interpolate in software, or filter in hardware as appropriate. AC to DC input (+/-) would give proportional and linear primary current output as a voltage. See the Allegro link for more details.
