Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces CS-ETS, a novel architecture for EMG-to-Speech synthesis that integrates a Samba-based encoder with two innovative chaos-inspired loss functions: Lyapunov Exponent Regularization (LER) and Multi-Scale Detrended Fluctuation Analysis (MSDFA). By leveraging these methods, CS-ETS achieves a remarkable reduction in parameter count by 40.79% compared to previous models while enhancing performance metrics such as LSD, STOI, and SI-SDR. The study highlights the potential of using chaotic dynamics to improve ETS systems, demonstrating a 13.33% reduction in computation without sacrificing quality.
CS-ETS achieves superior EMG-to-Speech synthesis with 40% fewer parameters while leveraging chaos-inspired techniques for enhanced performance.
We propose a chaos-inspired new architecture for EMG-to-Speech (ETS) synthesis called CS-ETS, which combines a Samba-based encoder with two novel chaos-inspired loss functions -- Lyapunov Exponent Regularization (LER) and Multi-Scale Detrended Fluctuation Analysis (MSDFA). LER is designed based on Lyapunov exponents to capture nonlinear fluctuations and sensitivity to initial conditions. MSDFA exploits detrended fluctuation analysis to quantify fractal-like, long-range temporal chaotic correlation. CS-ETS surpasses prior work with a 40.79\% lower parameter count (32M vs 54.1M) and introduces a new Post-Vocoder Alignment approach that improves LSD by 2.1x, STOI by 4.7x, and SI-SDR by 1.25x. CS-ETS reduces computation by 13.33\% while maintaining improved performance. To the best of our knowledge, for the first time, we show how ETS can be supervised by the subtle non-linear chaotic physics with Samba attention to achieve a significantly smaller model with superior performance.