Investigating Gated Recurrent Neural Networks for Speech Synthesis


Samples to support the ICASSP 2016 submission titled above. More details can be found in our paper. If you have any questions, please drop me an email: zhizheng.wu {at} ed.ac.uk (Zhizheng Wu)


This page contains following samples:

  1. Natural: Natural recording speech.
  2. LSTM: Baseline long short-term memory (LSTM) based RNN system.
  3. NIG: LSTM without input gate.
  4. NOG: LSTM without output gate.
  5. NFG: LSTM without forget gate.
  6. NPH: LSTM without peep-hole connections.
  7. GRU: Gated Recurrent Unit.
  8. S-LSTM: Simplified LSTM with only forget gate.


Samples Natural LSTM NIG NOG NFG NPH GRU S-LSTM
1
2
3
4
5
6