Exploring the Integration of Large Language Models into Automatic Speech Recognition Systems: An Empirical Study