Acoustic-to-articulatory inversion for dysarthric speech: Are pre-trained self-supervised representations favorable?
This work addresses the challenge of mapping acoustic to articulatory features for individuals with dysarthria, an incremental improvement in speech processing for assistive technologies.
The paper tackles acoustic-to-articulatory inversion for dysarthric speech by using pre-trained self-supervised learning representations, achieving relative improvements in Pearson Correlation Coefficient of ~1.81% for healthy controls and ~4.56% for patients over MFCCs in fine-tuned schemes.
Acoustic-to-articulatory inversion (AAI) involves mapping from the acoustic to the articulatory space. Signal-processing features like the MFCCs, have been widely used for the AAI task. For subjects with dysarthric speech, AAI is challenging because of an imprecise and indistinct pronunciation. In this work, we perform AAI for dysarthric speech using representations from pre-trained self-supervised learning (SSL) models. We demonstrate the impact of different pre-trained features on this challenging AAI task, at low-resource conditions. In addition, we also condition x-vectors to the extracted SSL features to train a BLSTM network. In the seen case, we experiment with three AAI training schemes (subject-specific, pooled, and fine-tuned). The results, consistent across training schemes, reveal that DeCoAR, in the fine-tuned scheme, achieves a relative improvement of the Pearson Correlation Coefficient (CC) by ~1.81% and ~4.56% for healthy controls and patients, respectively, over MFCCs. We observe similar average trends for different SSL features in the unseen case. Overall, SSL networks like wav2vec, APC, and DeCoAR, trained with feature reconstruction or future timestep prediction tasks, perform well in predicting dysarthric articulatory trajectories.