Fast Griffin Lim based waveform generation strategy for text-to-speech synthesis
- PDF / 1,960,473 Bytes
- 29 Pages / 439.642 x 666.49 pts Page_size
- 107 Downloads / 166 Views
Fast Griffin Lim based waveform generation strategy for text-to-speech synthesis Ankit Sharma1 · Puneet Kumar1 · Vikas Maddukuri2 · Nagasai Madamshetti2 · K. G. Kishore2 · Sahit Sai Sriram Kavuru2 · Balasubramanian Raman1 · Partha Pratim Roy1 Received: 22 October 2019 / Revised: 23 June 2020 / Accepted: 9 July 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract The performance of text-to-speech (TTS) systems heavily depends on spectrogram to waveform generation, also known as the speech reconstruction phase. The time required for the same is known as synthesis delay. In this paper, an approach to reduce speech synthesis delay has been proposed. It aims to enhance the TTS systems for real-time applications such as digital assistants, mobile phones, embedded devices, etc. The proposed approach applies Fast Griffin Lim Algorithm (FGLA) instead Griffin Lim algorithm (GLA) as vocoder in the speech synthesis phase. GLA and FGLA are both iterative, but the convergence rate of FGLA is faster than GLA. The proposed approach is tested on LJSpeech, Blizzard and Tatoeba datasets and the results for FGLA are compared against GLA and neural Generative Adversarial Network (GAN) based vocoder. The performance is evaluated based on synthesis delay and speech quality. A 36.58% reduction in speech synthesis delay has been observed. The quality of the output speech has improved, which is advocated by higher Mean opinion scores (MOS) and faster convergence with FGLA as opposed to GLA. Keywords Tacotron · Vocoder · Text to speech synthesis delay · Dilated convolutional neural network
1 Introduction The conclusive step in a text-to-speech (TTS) system is the generation of speech from the spectrogram representation of the signal. This process is known as the waveform Electronic supplementary material The online version of this article (https://doi.org/10.1007/s11042-020-09321-7) contains supplementary material, which is available to authorized users. Vikas Maddukuri
[email protected] 1
Computer Science and Engineering Department, Indian Institute of Technology, Roorkee, 247667, India
2
Electronics and Communication Engineering Department, Indian Institute of Technology, Roorkee, 247667, India
Multimedia Tools and Applications
reconstruction while the generation of intermediate signal from input text is called the construction process. The waveform generated in the reconstruction process is the time-domain signal obtained from its intermediate spectrogram. The overall performance of a TTS system depends on the waveform processing involved in the reconstruction phase [24]. The main challenge in the TTS systems is to optimize the waveform processing time while maintaining or improving the quality of the generated speech [47]. The TTS systems have emerged as valuable tools for day-to-day applications such as digital assistants, mobile phones, embedded devices, etc. Most of these devices have limited computational capacity and they are sometimes used in offline mode. Reducing the speech synthesis de
Data Loading...
