Original 24-channel speech
Reconstructed 24-channel Speech, using H-inverse (forward) model
Reconstructed 24-channel Speech, using G-inverse (backward) model