Framewise WaveGAN: High Speed Adversarial Vocoder in Time Domain with Very Low Computational Complexity

TL;DR

This paper introduces Framewise WaveGAN, a low-complexity GAN vocoder that generates speech waveforms in framewise manner, enabling fast and high-quality speech synthesis on low-power devices with minimal computational resources.

Contribution

The paper proposes a novel GAN vocoder architecture based on recurrent and fully-connected networks for framewise waveform generation, reducing computational complexity significantly.

Findings

Achieves higher speech quality than LPCNet at 1.2 GFLOPS

Enables fast speech synthesis on CPUs and low-power devices

Reduces computational cost compared to traditional GAN vocoders

Abstract

GAN vocoders are currently one of the state-of-the-art methods for building high-quality neural waveform generative models. However, most of their architectures require dozens of billion floating-point operations per second (GFLOPS) to generate speech waveforms in samplewise manner. This makes GAN vocoders still challenging to run on normal CPUs without accelerators or parallel computers. In this work, we propose a new architecture for GAN vocoders that mainly depends on recurrent and fully-connected networks to directly generate the time domain signal in framewise manner. This results in considerable reduction of the computational cost and enables very fast generation on both GPUs and low-complexity CPUs. Experimental results show that our Framewise WaveGAN vocoder achieves significantly higher quality than auto-regressive maximum-likelihood vocoders such as LPCNet at a very low complexity of 1.2 GFLOPS. This makes GAN vocoders more practical on edge and low-power devices.