Pre-training with Synthetic Patterns for Audio