Speech Audio Synthesis from Tagged MRI and Non-Negative Matrix Factorization via Plastic Transformer

TL;DR

This paper introduces a novel deep learning framework called Plastic Light Transformer (PLT) for converting MRI-derived weighting maps into speech audio, leveraging advanced transformer techniques to improve speech synthesis quality.

Contribution

The work presents the first end-to-end deep learning model using PLT to synthesize speech from MRI-based functional units, incorporating innovative bias and pooling mechanisms for variable input sizes.

Findings

Outperforms conventional models in speech synthesis quality

Effectively models global correlations in matrix inputs

Maintains high realism with limited training data

Abstract

The tongue's intricate 3D structure, comprising localized functional units, plays a crucial role in the production of speech. When measured using tagged MRI, these functional units exhibit cohesive displacements and derived quantities that facilitate the complex process of speech production. Non-negative matrix factorization-based approaches have been shown to estimate the functional units through motion features, yielding a set of building blocks and a corresponding weighting map. Investigating the link between weighting maps and speech acoustics can offer significant insights into the intricate process of speech production. To this end, in this work, we utilize two-dimensional spectrograms as a proxy representation, and develop an end-to-end deep learning framework for translating weighting maps to their corresponding audio waveforms. Our proposed plastic light transformer (PLT) framework is based on directional product relative position bias and single-level spatial pyramid pooling, thus enabling flexible processing of weighting maps with variable size to fixed-size spectrograms, without input information loss or dimension expansion. Additionally, our PLT framework efficiently models the global correlation of wide matrix input. To improve the realism of our generated spectrograms with relatively limited training samples, we apply pair-wise utterance consistency with Maximum Mean Discrepancy constraint and adversarial training. Experimental results on a dataset of 29 subjects speaking two utterances demonstrated that our framework is able to synthesize speech audio waveforms from weighting maps, outperforming conventional convolution and transformer models.