Investigating Deep Learning Architectures for Robust Speech Emotion Recognition: A Study of LSTM, CNN, and CLSTM Model