Deep Recurrent Neural Network-Based Autoencoders for Acoustic Novelty Detection

Figure 1

LSTM unit, comprising the input, output, and forget gates and the memory cell.