High-resolution dynamic speech imaging with joint low-rank and sparsity constraints

Area of Science:

• Medical Imaging
• Speech Science
• Biomedical Engineering

Background:

• Dynamic speech imaging requires high spatiotemporal resolution and full vocal tract coverage.
• Sparse sampling techniques offer potential for improved imaging speed and coverage.

Purpose of the Study:

• To develop a dynamic speech imaging method with high spatiotemporal resolution and full vocal tract coverage.
• To leverage sparse sampling advancements for improved articulatory motion imaging.

Main Methods:

• Developed a novel imaging method exploiting low-rank and sparsity of dynamic speech images.
• Implemented a data acquisition strategy with high temporal frame rate spiral navigators.
• Utilized image reconstruction with joint low-rank and sparsity constraints on sparsely sampled data.

Main Results:

• Achieved 102 frames per second (fps) for single-slice imaging at 2.2 × 2.2 × 6.5 mm(3) resolution.
• Enabled 12.8 fps for an eight-slice protocol covering the entire vocal tract with identical resolution.
• Demonstrated practical utility in phonetic investigations.

Conclusions:

• High spatiotemporal resolution dynamic speech imaging with full vocal tract coverage is feasible.
• Low-rank and sparsity constraints are effective for dynamic speech imaging.
• The developed method advances the study of articulatory motion during speech.