As shown in the oscilloscope, the pulse is actually 27us, followed by 32 pulse groups with 5us periods and a duty cycle of 40%, how do I capture 27us pulses again and start counting accurately?

I always capture 7us when capturing 2us high voltage, that is, the rising edge of the former to the falling edge of the latter, is it because I interrupt the calculation of pulse width in the function too much time? Is there any way to count exactly