Text-to-speech system for low-resource language using cross-lingual transfer learning and data augmentation

Abstract

Deep learning techniques are currently being applied in automated text-to-speech (TTS) systems, resulting in significant improvements in performance. However, these methods require large amounts of text-speech paired data for model training, and collecting this data is costly. Therefore, in this paper, we propose a single-speaker TTS system containing both a spectrogram prediction network and a neural vocoder for the target language, using only 30 min of target language text-speech paired data for training. We evaluate three approaches for training the spectrogram prediction models of our TTS system, which produce mel-spectrograms from the input phoneme sequence: (1) cross-lingual transfer learning, (2) data augmentation, and (3) a combination of the previous two methods. In the cross-lingual transfer learning method, we used two high-resource language datasets, English (24 h) and Japanese (10 h). We also used 30 min of target language data for training in all three approaches, and for generating the augmented data used for training in methods 2 and 3. We found that using both cross-lingual transfer learning and augmented data during training resulted in the most natural synthesized target speech output. We also compare single-speaker and multi-speaker training methods, using sequential and simultaneous training, respectively. The multi-speaker models were found to be more effective for constructing a single-speaker, low-resource TTS model. In addition, we trained two Parallel WaveGAN (PWG) neural vocoders, one using 13 h of our augmented data with 30 min of target language data and one using the entire 12 h of the original target language dataset. Our subjective AB preference test indicated that the neural vocoder trained with augmented data achieved almost the same perceived speech quality as the vocoder trained with the entire target language dataset. Overall, we found that our proposed TTS system consisting of a spectrogram prediction network and a PWG neural vocoder was able to achieve reasonable performance using only 30 min of target language training data. We also found that by using 3 h of target language data, for training the model and for generating augmented data, our proposed TTS model was able to achieve performance very similar to that of the baseline model, which was trained with 12 h of target language data.

A Machine Speech Chain Approach for Dynamically Adaptive Lombard TTS in Static and Dynamic Noise Environments

“…Clean Lombard speech is a speech under the Lombard effect but without noise in the audio. In our experiment, the clean Lombard speech is a synthetic Lombard speech generated by modifying the prosody of normal speech (intensity, pitch, duration) into Lombard speech using SoX audio manipulation toolkit [41], [42]. Noises were not included in the resulting audio.…”

Towards Yoruba-Speaking Google Maps Navigation

“…Deep learning however relies heavily on a substantial quantity of training data [34], [35] such that it was stated in [33], [36] that DNN is not a suitable technique for TTS in low-resource languages. In [37] however, techniques such as monolingual transfer learning, cross-lingual transfer learning, multi-speaker models, multilingual models, and data augmentation have been proposed as means of augmenting TTS for low-resource languages.…”

Code-Mixed Text-to-Speech Synthesis Under Low-Resource Constraints

“…Cross-lingual transfer learning and data augmentation approach for low resource TTS were proposed in (Byambadorj et al, 2021). The spectrogram prediction network was trained using crosslingual transfer learning (TL) from high resource language, data augmentation by varying parameters like pitch and speed, and a combination of two approaches.…”