Forward Attention in Sequence-to-sequence Acoustic Modelling for Speech Synthesis

TL;DR

This paper introduces a forward attention mechanism with a transition agent for speech synthesis, improving convergence speed, stability, naturalness, and speech speed control in sequence-to-sequence acoustic modeling.

Contribution

It presents a novel forward attention approach with a transition agent that leverages monotonic alignment for enhanced speech synthesis performance.

Findings

Faster convergence compared to baseline methods

Higher stability in attention during training

Improved naturalness and controllability of synthetic speech

Abstract

This paper proposes a forward attention method for the sequenceto- sequence acoustic modeling of speech synthesis. This method is motivated by the nature of the monotonic alignment from phone sequences to acoustic sequences. Only the alignment paths that satisfy the monotonic condition are taken into consideration at each decoder timestep. The modified attention probabilities at each timestep are computed recursively using a forward algorithm. A transition agent for forward attention is further proposed, which helps the attention mechanism to make decisions whether to move forward or stay at each decoder timestep. Experimental results show that the proposed forward attention method achieves faster convergence speed and higher stability than the baseline attention method. Besides, the method of forward attention with transition agent can also help improve the naturalness of synthetic speech and control the speed of synthetic speech effectively.