Text this: Leveraging MFCC and Mel-Spectrogram Representations for Deep Learning-Based Speech Recognition †.