User-Specific Learning for Recognizing a Singer's Intended Pitch