Text this: Continuous robust sound event classification using time-frequency features and deep learning.