Fig. 3: We compared the effectiveness of decoding algorithms at reconstructing various audio aspects and generating intelligible audio using six different performance metrics.
The top-performing algorithm across all metrics was the LSTM RNN (red stars).
