Improving Ultrasound Tongue Image Reconstruction from Lip Images Using Self-supervised Learning and Attention Mechanism

Abstract

Speech production is a dynamic procedure, which involved multi human organs including the tongue, jaw and lips. Modeling the dynamics of the vocal tract deformation is a fundamental problem to understand the speech, which is the most common way for human daily communication. Researchers employ several sensory streams to describe the process simultaneously, which are incontrovertibly statistically related to other streams. In this paper, we address the following question: given an observable image sequences of lips, can we picture the corresponding tongue motion. We formulated this problem as the self-supervised learning problem, and employ the two stream convolutional network and long-short memory network for the learning task, with the attention mechanism. We evaluate the performance of the proposed method by leveraging the unlabeled lip videos to predict an upcoming ultrasound tongue image sequence. The results show that our model is able to generate images that close to the real ultrasound tongue images, and results in the matching between two imaging modalities. CCS CONCEPTS• Computing methodologies → Machine learning algorithms; Knowledge representation and reasoning; Object recognition.

DAFT-Net: Dual Attention and Fast Tongue Contour Extraction Using Enhanced U-Net Architecture

“…These transformations lead to a vector-based connection notation for cascaded features between x l and g within the intermediary space of R F int . The AG's output merges input elements with attention coefficients through elementwise multiplication, as formulaically represented in Equation (3). In this study, the AG calculates a singular scalar focus value for each pixel vector x l i ∈ R F l , with F l indicating the number of feature maps at layer l.…”

“…Research indicates that tongue contours serve as an invaluable foundation for the quantitative analysis of speech, with data obtained from these contours facilitating the advancement and comprehension of speech models [ 3 , 4 ]. Ultrasonic tongue contour extraction can dynamically capture the tongue’s position across various phonetic expressions and depict the movements responsible for sound transitions during articulation [ 5 ].…”