Enhanced Direct Speech-to-Speech Translation Using Self-supervised Pre-training and Data Augmentatio