DPP-TTS: Diversifying prosodic features of speech via determinantal point processes
This work addresses the challenge of producing varied and natural-sounding speech in TTS systems, which is incremental as it builds on existing methods to improve prosody diversity.
The paper tackled the problem of generating speech with diverse prosody in neural text-to-speech models, which often lack perceptual diversity and naturalness when using high sampling temperatures, and proposed DPP-TTS to simultaneously enhance diversity within and among samples, showing improved prosodic diversification over baselines in side-by-side comparison tests.
With the rapid advancement in deep generative models, recent neural Text-To-Speech(TTS) models have succeeded in synthesizing human-like speech. There have been some efforts to generate speech with various prosody beyond monotonous prosody patterns. However, previous works have several limitations. First, typical TTS models depend on the scaled sampling temperature for boosting the diversity of prosody. Speech samples generated at high sampling temperatures often lack perceptual prosodic diversity, which can adversely affect the naturalness of the speech. Second, the diversity among samples is neglected since the sampling procedure often focuses on a single speech sample rather than multiple ones. In this paper, we propose DPP-TTS: a text-to-speech model based on Determinantal Point Processes (DPPs) with a prosody diversifying module. Our TTS model is capable of generating speech samples that simultaneously consider perceptual diversity in each sample and among multiple samples. We demonstrate that DPP-TTS generates speech samples with more diversified prosody than baselines in the side-by-side comparison test considering the naturalness of speech at the same time.