Workload in the first two weeks

Set up the environment

Learn the "speech recognition" system architecture, process and principle

Focus on learning acoustic models

Training CNN+LSTM+CTC, the result is not ideal