WaveCycleGAN2: Time-domain Neural Post-filter for Speech Waveform Generation
This work addresses the quality gap in speech synthesis for applications requiring high-fidelity audio, though it is incremental as it builds on prior WaveCycleGAN methods.
The paper tackled the problem of aliasing in synthesized speech waveforms, which made them distinguishable from natural speech, by proposing WaveCycleGAN2, a time-domain neural post-filter that eliminates down/up-sampling modules and combines waveform and acoustic parameter discriminators, achieving a mean opinion score comparable to natural speech and state-of-the-art models like WaveNet and WaveGlow while processing at over 150 kHz on an NVIDIA Tesla P100.
WaveCycleGAN has recently been proposed to bridge the gap between natural and synthesized speech waveforms in statistical parametric speech synthesis and provides fast inference with a moving average model rather than an autoregressive model and high-quality speech synthesis with the adversarial training. However, the human ear can still distinguish the processed speech waveforms from natural ones. One possible cause of this distinguishability is the aliasing observed in the processed speech waveform via down/up-sampling modules. To solve the aliasing and provide higher quality speech synthesis, we propose WaveCycleGAN2, which 1) uses generators without down/up-sampling modules and 2) combines discriminators of the waveform domain and acoustic parameter domain. The results show that the proposed method 1) alleviates the aliasing well, 2) is useful for both speech waveforms generated by analysis-and-synthesis and statistical parametric speech synthesis, and 3) achieves a mean opinion score comparable to those of natural speech and speech synthesized by WaveNet (open WaveNet) and WaveGlow while processing speech samples at a rate of more than 150 kHz on an NVIDIA Tesla P100.