You say you can use ICs, so the restriction on microcontrollers makes no sense. A microcontroller is just another IC. Since justification for this arbitrary restriction is not given, we can conclude that it is religious in nature and ignore it.
I would use a micro that has both "input capture" and PWM output. There are many that have both these, including the many of the PIC 16. Use the input capture to measure the incoming PWM duty cycle. Depending on how exactly the hardware works, you may need two input capture modules. One to capture a free running timer on the rising edge, and another on the falling edge. Another option is to use a gated timer so that the timer runs when the incoming PWM signal is high. On the falling edge, you grab the timer value and subtract it from the previous one to get a measure of the PWM high time.
No matter how you use the hardware to get the incoming duty cycle, you use that to adjust the output duty cycle to be the same. Since this is done pulse by pulse, the response should be effectively the same as with the slower PWM. The feedback loop should not be able to tell the difference between driving at 400 Hz directly, or at 24 kHz with your converter in there.
Another advantage of doing this is that the motor will probably run a little more efficiently. The motor coils look inductive to the driving circuit. With the higher switching frequency, there is less time between switching for the inductor current to change. The lower ripple current thru the motor causes less I2R loss. The ripple current at the PWM frequency does nothing to move the motor. It just heats the windings.