ABSTRACT
This paper proposes a novel framework for deriving causal neural-muscular-articulatory latency models from synchronized multimodal recordings, integrating variable-lag causal inference, neural-muscular latency estimation, and articulatory prediction networks.
PAPER · PDF
Loading PDF…
Key findings
Proposes a variable-lag causal inference module based on time-varying Granger causality.
Develops a neural-muscular latency estimator using attention-based alignment between EEG and EMG signals.
Models the transformation from muscle activity to electromagnetic articulography (EMA) kinematics.
Limitations & open questions
The framework's validation is pending through perturbation experiments and biomechanical simulation comparisons.