Search

Acoustic features extracted from speech can help with the diagnosis of neurological diseases and monitoring of symptoms over time. Temporal segmentation of audio signals into individual words is an important pre-processing step needed prior to extracting acoustic features. Machine learning techniques could be used to automate speech segmentation via automatic speech recognition (ASR) and sequence to sequence alignment. While state-of-the-art ASR models achieve good performance on healthy speech, their performance significantly drops when evaluated on dysarthric speech. Fine-tuning ASR models on impaired speech can improve performance in dysarthric individuals, but it requires representative clinical data, which is difficult to collect and may raise privacy concerns. This study explores the feasibility of using two augmentation methods to increase ASR performance on dysarthric speech: 1) healthy individuals varying their speaking rate and loudness (as is often used in assessments of pathological speech); 2) synthetic speech with variations in speaking rate and accent (to ensure more diverse vocal representations and fairness). Experimental evaluations showed that fine-tuning a pre-trained ASR model with data from these two sources outperformed a model fine-tuned only on real clinical data and matched the performance of a model fine-tuned on the combination of real clinical data and synthetic speech. When evaluated on held-out acoustic data from 24 individuals with various neurological diseases, the best performing model achieved an average word error rate of 5.7% and a mean correct count accuracy of 94.4%. In segmenting the data into individual words, a mean intersection-over-union of 89.2% was obtained against manual parsing (ground truth). It can be concluded that emulated and synthetic augmentations can significantly reduce the need for real clinical data of dysarthric speech when fine-tuning ASR models and, in turn, for speech segmentation.

The Three Gorges hydropower station in China creates a large reservoir by diverting water from the Yangtze River, increasing the risk of geological disasters, especially massive landslides along the reservoir shoreline. To mitigate these risks, improving geological monitoring and early warning systems is crucial. Interferometric Synthetic Aperture Radar (InSAR) is widely used to monitor reservoir bank landslides. However, its potential in early warning systems is limited due to temporal resolution constraints, preventing timely warnings. To address this, we propose integrating daily hydrological data (precipitation and water level observations) with historical InSAR deformation sequences using our deep learning-based multivariate united state estimation network, “MUSEnet.” This approach generates customized daily landslide deformation products for high-risk areas, greatly enhancing early warning capabilities by providing timely and accurate information on landslide occurrence and magnitude. We validated our method using 161 Sentinel-1 A images of the Xinpu landslide in the Three Gorges Reservoir area. Through statistical analysis, we identified different degrees of influence from rainfall and reservoir water level on the deformation of the Xinpu landslide at various locations. Additionally, we observed distinct lag times between deformation and corresponding rainfall and reservoir water level events. By utilizing deep learning, our method estimates nonlinear states by considering hysteresis and intelligently accounts for the impact of rainfall and reservoir water level, resulting in more accurate estimations compared to traditional models.

In order to fully extract the temporal and spatial features contained in motor imagery electroencephalography (EEG) signals for effective identification of motor imagery, a three-dimensional capsule network (3D-CapsNet) EEG signal recognition model is proposed, which can integrate the MI-EEG temporal dimension, channel spatial dimension and the intrinsic relationship between features to maximize the feature representation capability. Firstly, a multi-layer 3D convolution module is used to extract features in the time and inter-channel space dimensions as the low-level features. Secondly, advanced spatial features are also obtained through capsule network integration. Finally, dynamic routing connections and squash functions are applied for classification. The experimental analysis is conducted on the BCI competition IV dataset 2a. The proposed model performs well on all the subjects’ datasets, such that the average accuracy and average Kappa value of 9 subjects are 84.028% and 0.789, respectively. The experimental results confirm effectiveness of the proposed method. Additionally, accuracy of the four-class classification is improved, and the impact of individual variability is overcome to a certain extent.

Wireless capsule endoscopy (WCE) is a modern, non-invasive method of gastrointestinal examination that can significantly reduce mortality and morbidity. One of the current challenges in WCE is the precise localization of the capsule. An accurate path loss propagation model can be used to find the exact distance from the surface to the capsule inside the abdominal cavity. Unfortunately, there are no standardized In-to-On-Body channel models describing the signal propagation at ultra-high frequencies that are used in the most commercially available WCE systems. This study addresses the gap by conducting an experimental validation of a new propagation model for WCE applications at 2.45 GHz. The results were confirmed by conducting two separate in-vivo trials on porcine animals under general anesthesia. The performance of the model as well as the corresponding ranging errors were evaluated when it was used as an inverse solution for distance estimation to an ingested transmitter. The main advantage of the model is its theoretical basis, which can help further generalize the findings for similar communication scenarios. The obtained ranging error was smaller than one centimeter, suggesting that it can be used for accurate range-based positioning of implanted transmitters.

A flexible first-order OAM modes generation in ring-core fibers (RCFs) based on a mechanically-induced long-period fiber grating (MLPFG) is proposed. By directly fusion splicing a standard single-mode fiber (SMF) and RCFs, the anti-resonant reflection guidance mechanism (ARRGM) is introduced in the RCF, leading to the guided-core modes, so-called anti-resonant modes (AR modes), which have a Gaussian field distribution and are confined in low-refractive index central core of RCFs. A mechanical long-period fiber grating is used to couple the AR modes into the OAM modes propagated in the high-index ring core of the RCF. Results show that the generated OAM modes are the combination of two orthogonal linearly polarized (LP) modes with ±π/2 phase shift in the ring core. The mode conversion efficiency of OAM+1 and OAM−1 can reach 95% and 94%, respectively, indicating promising applications in sensing and communication of the RCFs.

Meibomian glands dysfunction (MGD) is the main cause of dry eyes. Biological parameters of meibomian gland (MG) such as height, tortuosity and the degree of atrophy are closely related to its function. However, Thus, an effective quantitative diagnostic tool is needed for clinical diagnosis. Automatic quantification of MGs' morphological features could be a challenging task and play an important role in MGD diagnosis and classification. Our main objective is to develop an artificial intelligence (AI) system for evaluating MGs' morphology and explore the relationship between the morphological parameters and functions. We proposed a novel MGs extraction method based on convolutional neural network (CNN) with enhanced mini U-Net. A prospective study was conducted, 120 subjects were included and taken meibography. The training and validation sets encompassed 60 subjects; and the test set consisted of other 60 subjects with comprehensive examinations for ocular surface disease index questionnaire (OSDI), tear meniscus height (TMH), tear break-up time (TBUT), corneal fluorescein staining (CFS), lid margin score, and meibum expressibility score. The algorithm effectively extracted MGs from meibography even with this small training sample. As a result, while the intersection over union (IoU) achieved 0.9077, the repeatability was 100%. The processing time for each image was 100ms. Using this method, the investigators identified a significant and linear correlation between MG morphology and clinical parameters. This study provided a new method for quantification of MGs' morphological features obtained by meibography, which has advantages in reducing analysis time, improving diagnostic efficiency, and assisting ophthalmologists with limited clinical expertise.

The magneto-refractive properties of Er/Yb co-doped silica fiber (EYDF) are important in magnetic field sensing. Here, we proposed a theoretical calculation model of the EYDF using density functional theory (DFT) and demonstrated it by an interferometric digital hologram magneto-refractive measurement system. The calculation results show that the fiber material, doped with Er/Yb atoms, has a large net spin and asymmetric spin distribution, which enables the elements to play an important role in the magnetic characteristics of the material. The experimental results show that the refractive index of the fiber decreases linearly with the increase of the magnetic field. The sensitivity of EYDF is 3.8279 × 10-5 RIU/Gs and is two orders of magnitude higher than that of single-mode fiber which is confirmed by the theoretical calculation. It demonstrated that the EYDF has advantages and potential of high sensitivity, miniaturization, and easy integration for magnetic field detection in harsh environments.

Image caption is to understand and describe the visual content, which is expected to be applied in automatic news reporting in future. In recent years, there has been an increasing interest in an Encoder-Decoder framework for image caption: the encoder takes the responsibility for visual semantic comprehension and the decoder is designed for sentence generation. In the Encoder-Decoder framework the translation is based on the correspondence between image feature vectors and caption vectors. Attention mechanism makes sense for a more accurate correspondence. However, the attention model works with the decoder, and the focused content changes dynamically with the generated word. It results that in many cases the salient contents are not described in the caption, or the objects described are not the salient ones. To improve the precision of image caption, to bridge the gap between image understanding and sentence generation in the Encoder-Decoder framework, and to align visual information and semantic information better, we propose a concept of visual keyword as a gang board between seeing and saying. This paper presents an image dataset derived from MSCOCO as the first collection of visual keywords: Image Visual Keyword Dataset (IVKD). Also, a Visual Semantic Attention Model(VSAM) is proposed to obtain visual keywords for generating the annotation. In VSAM, the object-level visual features are extracted by an object detector after pre-training on IVKD. Then the object features are fed in an Optimized Pointer Network(OPN) to generate visual keywords. The experiments show that the precision of visual keyword generation reaches 91.7% by the proposed model VSAM.

Magnetostrictive thin-film coated Surface Acoustic Wave (SAW) devices were promising for sensing magnetic field owing to their superior features as micro-size, fast response, and high sensitivity originated from the magnetostrictive effect. However, the magnetostriction nature in magnetostrictive thin-film causes significantly mechanical fatigue in service, deteriorating the sensor performances. In this work, the fatigue phenomenon in magnetostrictive coating was underlined by characterizing the prepared FeCo thin-film coated magnetic device cyclically. Obvious shedding was observed in FeCo coating after cyclic testing and the magnetic-sensitivity decreases significantly. One of the reasons is the weak adhesion of FeCo thin-film towards the substrate. As an available way allowing enhancement of adhesion, a Cr thin-film was employed as the transition-layer to weaken the mechanical fatigue. However, it accompanied by the issue of the reduced magnetostrictive coefficient and the obstruction in magnetostrain-tranfer to piezoelectric substrate. As a result, the slump in sensitivity was observed. To address such issues, a design of dotted-pattern with Cr transition layer was employed to build the SAW based magnetic-device. High magnetic-sensitivity and excellent long-term stability were achieved because of the release of coercive force in FeCo dots and enhancement of the FeCo adhesion to the substrate.

The Coronavirus pandemic has created complex challenges and adverse circumstances. This research identifies public sentiment amidst problematic socioeconomic consequences of the lockdown, and explores ensuing four potential public sentiment associated scenarios. The severity and brutality of COVID-19 have led to the development of extreme feelings, and emotional and mental healthcare challenges. This research focuses on emotional consequences - the presence of extreme fear, confusion and volatile sentiments, mixed along with trust and anticipation. It is necessary to gauge dominant public sentiment trends for effective decisions and policies. This study analyzes public sentiment using Twitter Data, time-aligned to the COVID-19 reopening debate, to identify dominant sentiment trends associated with the push to reopen the economy. Present research uses textual analytics methodologies to analyze public sentiment support for two potential divergent scenarios - an early opening and a delayed opening, and consequences of each. Present research concludes on the basis of textual data analytics, including textual data visualization and statistical validation, that tweets data from American Twitter users shows more positive sentiment support, than negative, for reopening the US economy. This research develops a novel sentiment polarity based public sentiment scenarios (PSS) framework, which will remain useful for future crises analysis, well beyond COVID-19. With additional validation, this research stream could present valuable time sensitive opportunities for state governments, the federal government, corporations and societal leaders to guide local and regional communities, and the nation into a successful new normal future.