A much better method is to makd the serial processor simply ignore errant data. If you use a watchdog to catch errors, the processor will 'go blind' during reset and possibly screw up on a byte transmitted in that time, thus leading to another reset, which leats to another reset .......

Erik