Search

Intracortical brain-computer interfaces (iBCIs) translate neural activity into control commands, thereby allowing paralyzed persons to control devices via their brain signals. Recurrent neural networks (RNNs) are widely used as neural decoders because they can learn neural response dynamics from continuous neural activity. Nevertheless, excessively long or short input neural activity for an RNN may decrease its decoding performance. Based on the temporal attention module exploiting relations in features over time, we propose a temporal attention-aware timestep selection (TTS) method that improves the interpretability of the salience of each timestep in an input neural activity. Furthermore, TTS determines the appropriate input neural activity length for accurate neural decoding. Experimental results show that the proposed TTS efficiently selects 28 essential timesteps for RNN-based neural decoders, outperforming state-of-the-art neural decoders on two nonhuman primate datasets (R2=0.76±0.05 for monkey Indy and CC=0.91±0.01 for monkey N). In addition, it reduces the computation time for offline training (reducing 5-12%) and online prediction (reducing 16-18%). When visualizing the attention mechanism in TTS, the preparatory neural activity is consecutively highlighted during arm movement, and the most recent neural activity is highlighted during the resting state in nonhuman primates. Selecting only a few essential timesteps for an RNN-based neural decoder provides sufficient decoding performance and requires only a short computation time.

In order to accurately diagnose the health of high-order statically indeterminate structures, most existing structural health monitoring (SHM) methods require multiple sensors to collect enough information. However, comprehensive data collection from multiple sensors for high degree-of-freedom structures is not typically available in practice. We propose a method that reconciles the two seemingly conflicting difficulties. Takens' embedding theorem is used to augment the dimensions of data collected from a single sensor. Taking advantage of the success of machine learning in image classification, high-dimensional reconstructed attractors were converted into images and fed into a convolutional neural network (CNN). Attractor classification was performed for 10 damage cases of a 3-story shear frame structure. Numerical results show that the inherently high dimension of the CNN model allows the handling of higher dimensional data. Information on both the level and the location of damage was successfully embedded. The same methodology will allow the extraction of data with unsupervised CNN classification to be consistent with real use cases.

The focus of this paper was designing and demonstrating bus structure FBG sensor networks using intensity wavelength division multiplexing (IWDM) techniques and a gated recurrent unit (GRU) algorithm to increase the capability of multiplexing and the ability to detect Bragg wavelengths with greater accuracy. Several Fiber Bragg grating (FBG) sensors are coupled with power ratios of 90:10 and 80:10, respectively in the suggested experimental setup. We used the latest IWDM multiplexing technique for the proposed scheme, as the IWDM system increases the number of sensors and allows us to alleviate the limited operational region drawback of conventional wavelength division multiplexing (WDM). However, IWDM has a crosstalk problem that causes high-sensor signal measurement errors. Thus, we proposed the GRU model to overcome this crosstalk or overlapping problem by converting the spectral detection problem into a regression problem and considered the sequence of spectral features as input. By feeding this sequential spectrum dataset into the GRU model, we trained the GRU system until we achieved optimal efficiency. Consequently, the well-trained GRU model quickly and accurately identifies the Bragg wavelength of each FBG from the overlapping spectra. The Bragg wavelength detection performance of our proposed GRU model is tested or validated using different numbers of FBG sensors, such as 3-FBG, 5-FBG, 7-FBG, and 10-FBG, separately. As a result, the experiment result proves that the well-trained GRU model accurately identifies each FBG Bragg wavelength, and even the number of FBG sensors increase, as well as the spectra of FBGs, which are partially or fully overlapped. Therefore, to boost the detection efficiency, reliability, and to increase the multiplexing capabilities of FBG sensor networks, the proposed sensor system is better than the other previously proposed methods.

A fluidic microchip incorporating a convergent microchannel and a Nafion-nanoporous membrane is proposed for the preconcentration and separation of multi-species samples on a single platform. In the device, sample preconcentration is achieved by means of the ion concentration polarization effect induced at the micro/nano interface under the application of an external electric field, while species separation is achieved by exploiting the different electrophoretic mobilities of the sample components. The experimental results show that the device is capable of detecting C-reactive protein (CRP) with an initial concentration as low as 9.50 × 10(-6) mg/L given a sufficient preconcentration time and driving voltage. In addition, it is shown that a mixed-species sample consisting of three negatively-charged components (bovine serum albumin (BSA), tetramethylrhodamine(TAMRA) isothiocyanate-Dextran and fluorescent polymer beads) can be separated and preconcentrated within 20 min given a driving voltage of 100 V across 1 cm microchannel in length. In general, the present results confirm the feasibility of the device for the immunoassay or detection of various multi-species samples under low concentration in the biochemical and biomedical fields. The novel device can therefore improve the detection limit of traditional medical facilities.