Search

In this paper, we investigate the use of the spontaneous speech of dysarthric people for training an automatic speech recognition (ASR) model for them. Although the spontaneous speech of dysarthric people can be collected relatively easily compared to script-reading speech, which is obtained by having them read a prepared script, labeling the spontaneous speech of dysarthric people is very difficult and costly. For training an ASR model using unlabeled speech data, pseudo-labeling and self-supervised feature learning have been studied as effective approaches; however, the effectiveness of these approaches has not been clear when they are applied to the unlabeled dysarthric speech. In addition, pseudo-labeling may not be effective since the pseudo-labels of dysarthric speech include many errors and are not reliable. In this paper, we evaluate the above two approaches for the dysarthric speech recognition, and we propose a multi-task learning approach, which combines these approaches to train an ASR model that is robust against the errors in the pseudo-labels. Experimental results using Japanese and English datasets demonstrated that all approaches are effective, but among them, the proposed multi-task learning approach showed the best performance.

Magnetoelectric multipoles are parity and time-reversal odd structures emerging in multiferroic and exotic ordered phases. Here, the discovery of nonreciprocal linear dichroism in Pb(TiO)Cu4(PO4)4 enables a fast visualization of magnetic quadrupole domains using simple linear polarization microscopy.

It is known that magnetic excitations in magnetoelectric multiferroics can be induced by the oscillation of electric polarization, called electromagnons, which often cause nonreciprocal optical phenomena. Energy diagrams of the excitations in the magnetoelectric multiferroic Pb(TiO)Cu_{4}(PO_{4})_{4} were obtained in wide ranges of magnetic fields up to 50 T and frequencies below 2 THz. Some of the observed resonance modes were reproduced qualitatively by a numerical analysis based on the generalized spin-wave theory. Moreover, we found that electron spin resonance (ESR) signals on some of the modes can be modulated by applying an electric field. This result implies the presence of nonreciprocal linear dichroism on the ESR signals in Pb(TiO)Cu_{4}(PO_{4})_{4}.

Elucidating the various behavioral and ecological uses of animal habitats is the basis for the conservation and management of animal species. Therefore, tracking the movement of animals is necessary. Biotelemetry is used for tracking the movement of animals. By mounting a radio telemetry receiver and antenna on a drone, the time and labor required for surveying animals can be reduced. In addition, it is easy to track difficult-to-reach areas such as rice paddies and forests, and the environment is not invaded by the survey. We think that this drone radio telemetry will be the best method for tracking the movement of small amphibians, such as frogs. However, in order to put the method to practical use, the accuracy of the system needs to be verified. Approximately 26 ha of area in Sogabe, Kameoka City, Kyoto Prefecture, Japan was investigated in this study. We selected and validated the location where frogs are likely to enter farmlands. The location where the detection of movement is expected to be stable are 5 cm deep areas in the soil, gaps in masonry, and under plastic bags, whereas areas in which the detection is likely to be unstable are areas deeper than 5 cm in the soil, covered concrete channels, and grass. By calculating the geographic center, the location of the nanotag could be estimated with an accuracy of less than 16 m. We successfully showed that the drone radio telemetry system used in this study is capable of detecting and tracking the movement of animals with high spatial and temporal resolutions. However, we suggest that the detection of movement may be interrupted depending on the location of the target animal and more than three detections are needed to guarantee the accuracy of the estimation.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.