Speech-based machine learning (ML) has been heralded as a promising solution for tracking prosodic and spectrotemporal patterns in real-life that are indicative of emotional changes, providing a valuable window into ones cognitive and mental state. Y
et, the scarcity of labelled data in ambulatory studies prevents the reliable training of ML models, which usually rely on data-hungry distribution-based learning. Leveraging the abundance of labelled speech data from acted emotions, this paper proposes a few-shot learning approach for automatically recognizing emotion in spontaneous speech from a small number of labelled samples. Few-shot learning is implemented via a metric learning approach through a siamese neural network, which models the relative distance between samples rather than relying on learning absolute patterns of the corresponding distributions of each emotion. Results indicate the feasibility of the proposed metric learning in recognizing emotions from spontaneous speech in four datasets, even with a small amount of labelled samples. They further demonstrate superior performance of the proposed metric learning compared to commonly used adaptation methods, including network fine-tuning and adversarial learning. Findings from this work provide a foundation for the ambulatory tracking of human emotion in spontaneous speech contributing to the real-life assessment of mental health degradation.
We develop a theoretical description of the Raman spectroscopy in the spin-phonon coupled Kitaev system and show that it can provide intriguing observable signatures of fractionalized excitations characteristic of the underlying spin liquid phase. In
particular, we obtain the explicit form of the phonon modes and construct the coupling Hamiltonians based on $D_{3d}$ symmetry. We then systematically compute the Raman intensity and show that the spin-phonon coupling renormalizes phonon propagators and generates the salient Fano linshape. We find that the temperature evolution of the Fano lineshape displays two crossovers, and the low temperature crossover shows pronounced magnetic field dependence. We thus identify the observable effect of the Majorana fermions and the $Z_2$ gauge fluxes encoded in the Fano lineshape. Our results explain several phonon Raman scattering experiments in the candidate material $alpha$-RuCl$_3$.
Here we present a study of the phonon dynamics in the honeycomb Kitaev spin model at finite temperatures. We show that the fractionalized spin excitations of the Kitaev spin liquid, the itinerant Majorana fermions and static $Z_2$ fluxes, have distin
ct effects on the phonon dynamics, which makes the phonon dynamics a promising tool for exploring the Kitaev spin liquid candidate materials. In particular, we will focus on the signature of the fractionalized excitations in the thermodynamic behaviour of the sound attenuation and the phonon Hall viscosity: The former describes the phonon decay into the fractionalized excitations, and the later is the leading order time reversal symmetry breaking effect on the acoustic phonon. We find that the angular dependence of the attenuation coefficient and its magnitude are modified by the thermal excitation of the $Z_2$ fluxes. The strength of this effect strongly depends on the relative magnitude of the sound velocity and the Fermi velocity characterizing the low-energy Majorana fermions. We also show that the Hall viscosity is strongly suppressed by the increase of the density of the $Z_2$ fluxes at finite temperatures. All these observations reflect the effects of the emergent disorder on the Majorana fermions introduced by the $Z_2$ fluxes. Our analysis is based on the complementary analytical calculations in the low-temperature zero-flux sector, and numerical calculations in the inhomogeneous flux sectors at intermediate and high temperatures with stratified Monte Carlo (strMC) method.
