Speech synthesis from short-time Fourier transform magnitude and its application to speech processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

You Do Not Need More Data: Improving End-To-End Speech Recognition by Text-To-Speech Data Augmentation (2020, methodology): “We also investigate the influence of waveform generation methods in TTS on final ASR performance by using the Griffin-Lim algorithm [15] and our modification of LPCNet [16].”
Real-Time Signal Estimation From Modified Short-Time Fourier Transform Magnitude Spectra (2007, methodology): “The signal-to-error ratio (SER) function [10] is used in evaluation SER (7) Since the magnitude spectrogram is given, minimizing the mean square error (3) is equivalent to maximizing the SER function (7).”
Generating Synthetic Audio Data for Attention-Based Speech Recognition Systems (2019, result): “In contrast to that, our approach includes a completely separate end-to-end TTS system with a Grifﬁn & Lim (G&L) vocoder [15] for synthetic wave-form generation instead of synthetic feature generation.”