Search

Reservoir computing is a potential neuromorphic paradigm for promoting future disruptive applications in the era of the Internet of Things, owing to its well-known low training cost and compatibility with hardware. It has been successfully implemented by injecting an input signal into a spatially extended reservoir of nonlinear nodes or a temporally extended reservoir of a delayed feedback system to perform temporal information processing. Here we propose a novel nondelay-based reservoir computer using only a single micromechanical resonator with hybrid nonlinear dynamics that removes the usually required delayed feedback loop. The hybrid nonlinear dynamics of the resonator comprise a transient nonlinear response, and a Duffing nonlinear response is first used for reservoir computing. Due to the richness of this nonlinearity, the usually required delayed feedback loop can be omitted. To further simplify and improve the efficiency of reservoir computing, a self-masking process is utilized in our novel reservoir computer. Specifically, we numerically and experimentally demonstrate its excellent performance, and our system achieves a high recognition accuracy of 93% on a handwritten digit recognition benchmark and a normalized mean square error of 0.051 in a nonlinear autoregressive moving average task, which reveals its memory capacity. Furthermore, it also achieves 97.17 ± 1% accuracy on an actual human motion gesture classification task constructed from a six-axis IMU sensor. These remarkable results verify the feasibility of our system and open up a new pathway for the hardware implementation of reservoir computing.

Micromachined resonant accelerometers (MRAs) are very attractive for high-precision measurement applications due to their high sensitivity, frequency output, and large dynamic range. The sensitivity is one of critical metrics in MRA sensors. In order to enhance this metric, a micro-lever is usually used. However, the micro-lever may suffer from the attenuation of effective amplification ratio and increasing the structure complexity. Lowering spring stiffness of suspension structure and enlarging the proof mass are alternative methods, but these methods are restricted in real devices due to the trade-offs for the device sizes, cross-axis sensitivity, robustness and fabrication yield. This paper presents a novel method to improve sensitivity of MRAs by using electrostatic spring softening effect without sacrificing the mechanical robustness or enlarging footprint size of the device. Besides, this method can on-line adjust the sensitivity of MRAs, which is another potential advantage for specific applications. The prototype has been designed, fabricated and characterized in both open-loop and close-loop ways. In agreement with the proposed device model, the experiment results demonstrated that the sensitivity of the prototype device increased from 492.7 Hz/ ${g}$ ( ${g}$ being the gravity acceleration) to 2277 Hz/ ${g}$ by using a sensitivity enhancing bias voltage of 67V in a linear scale range of $\pm 1~{g}$ . Moreover, the prototype’s equivalent input acceleration noise reduces down from 5.5 ug / $\surd $ Hz to 1.6 ug / $\surd $ Hz within 1Hz bandwidth, and the bias-instability is down from 6.1 ug to 2.2 ug .

High-sensitivity MEMS magnetoresistive (MR) sensors have attracted ever-increasing attention due to their ability to detect weak magnetic fields, but the resolution is severely limited by the [Formula: see text] noise. This paper reports a MEMS MR sensor that can effectively suppress the [Formula: see text] noise by modulating the magnetic field signal to the higher frequency region utilizing magnetic tunnel junction vertical motion modulation based on the MEMS piezoelectric cantilever resonator. The magnetic sensitivity can be increased to 2283.3%/mT with a high magnetic gain of 39.3 by integrating the structure-optimized fixed magnetic flux concentrators. The signal-to-noise ratio of the MEMS MR sensor can be improved by three orders of magnitude and the 160 pT/√Hz resolution can be reached. These results make the MEMS MR sensor based on the MTJ motion modulation a competitive candidate in the pico-Tesla magnetic detection.

This paper reports a novel designed electrostatic spring structure that use to significantly enhance the sensitivity of mode-localized resonant accelerometer (ML-RXL) without increasing the device size. Moreover, the sensitivity can be flexibly adjusted by the bias voltage. Finite element method (FEM) simulation results presented demonstrate that the sensitivity of ML-RXL is increased from 59.322 AR/g to 96.034 AR/g with a linear operating range of 0.3g by applying the bias voltage of 30V.

Reservoir computing (RC) is an attractive paradigm of a recurrent neural network (RNN) architecture, owning to the ease of training and existing neuromorphic implementation. Its simulated performance matches other digital algorithms on a series of benchmarking tasks, such as prediction tasks and classification tasks. In this article, we propose a novel RC structure based on the coupled MEMS resonators with the enhanced dynamic richness to optimize the performance of the RC system both on the system level and data set level. Moreover, we first put forward that the dynamic richness of RC comprises linear dynamic richness and nonlinear dynamic richness, which can be enhanced by adding delayed feedbacks and nonlinear nodes, respectively. In order to set forth this point, we compare three typical RC structures, a single-nonlinearity RC structure with single-feedback, a single-nonlinearity RC structure with double-feedbacks, and the couple-nonlinearity RC structure with double-feedbacks. Specifically, four different tasks are enumerated to verify the performance of the three RC structures, and the results show the enhanced dynamic richness by adding delayed feedbacks and nonlinear nodes. These results prove that coupled MEMS resonators offer an interesting platform to implement a complex computing paradigm leveraging their rich dynamical features.

Reservoir computing (RC) is a recently introduced bio-inspired computational framework capable of excellent performances in the temporal data processing, owing to its derivation from the recurrent neural network (RNN). It is well-known for the fast and effective training scheme, as well as the ease of the hardware implementation, but also the problematic sensitivity of its performance to the optimizable architecture parameters. In this article, a particular time-delayed RC with a single clamped–clamped silicon beam resonator that exhibits a classical Duffing nonlinearity is presented and its optimization problem is studied. Specifically, we numerically analyze the nonlinear response of the resonator and find a quasi-linear bifurcation point shift of the driving voltage with the driving frequency sweeping, which is called Bifurcation Point Frequency Modulation (BPFM). Furthermore, we first proposed that this method can be used to find the optimal driving frequency of RC with a Duffing mechanical resonator for a given task, and then put forward a comprehensive optimization process. The high performance of RC presented on four typical tasks proves the feasibility of this optimization method. Finally, we envision the potential application of the method based on the BPFM in our future work to implement the RC with other mechanical oscillators.

Non-contact and non-destructive acceleration measurement is receiving considerable attention due to their low cost, flexibility, and simplicity of implementation, as well as their excellent performance in some emerging applications such as medical electronics applications, vibration monitoring, and some other special scenarios. In this paper, a visual accelerometer system based on laser speckle optical flow detection named Viaxl is proposed. Compared with the conventional non-contact acceleration measurement method based on a laser system, Viaxl has moderate and stable performance with the advantages of low cost and simplicity of implementation. Experiment results demonstrate that Viaxl, which consists of a commercial camera and a low-cost laser pointer, can achieve real-time, non-contact acceleration measurement, and confirm the basic system performance of Viaxl: a measurement nonlinearity better than 1.3%, up to 31 dB signal-to-noise ratio, and 1150 Hz theoretic bandwidth; this demonstrates the huge potential of Viaxl in a wide range of applications, and provides a new possible technical method for non-contact acceleration detection.

Microelectromechanical system (MEMS) resonators have been widely used in the magneto-resistive (MR) sensor for modulating the magnetic flux to enhance the detection limit. However, the manufacturing tolerances in MEMS fabrication processes make it challenging to fabricate the identical resonators with the same vibration frequency, which greatly decreases the detection limit of the MR sensor. To synchronize the MEMS resonators and improve the performance of the MR sensor, the double-ended tuning fork (DETF) based comb-driven MEMS resonators is proposed in this paper, making the system operate at the out-of-phase mode to complete the synchronization. The dynamic behaviour of the resonators is investigated through theoretical analysis, numerical solution based on MATLAB code and Simulink, and experimental verification. The results show that the transverse capacitances in the comb will significantly affect the resonance frequency due to the second-order electrostatic spring constant. It is the first time to observe the phenomenon that the resonant frequency increases with the increase of the bias, and it can also decrease with increasing the bias through adjusting the initial space between the fixed finger and the moving mass, they are different from the model about spring softening and spring hardening. Besides, the proposed DETF based comb-driven resonators can suppress the in-phase and out-of-phase mode through adjusting the driving and sensing ports, and sensing method, meanwhile make the magnetic flux modulation fully synchronized, and maximize the modulation efficiency, and minimize the detection limit. These characteristics are appropriate for the MR sensor, even other devices that need to adjust the resonance frequency and vibration amplitude. Furthermore, the model and the design can also be extended to characteristic the single end tuning fork based MEMS resonator and other MEMS-based MR sensors.

With the advent of IoT era, the demand for position information is increasing constantly. In outdoor cases, satellite navigation solution like Global Positioning System (GPS) is well established and has been widely used. In indoor cases, there are positioning solutions like Wi-Fi, BLE, etc., but their respective defects have limited the practical development. Visible light positioning (VLP) has been a more promising positioning method in indoor cases because of its high accuracy, high cost efficiency, and easy to accomplish compared to other indoor positioning technologies. However, some drawbacks should be overcome during the application of VLP: VLP cannot work when optical signal is blocked or limited by the layout of LEDs, and the positioning accuracy will decrease when multipath reflection is not negligible. In order to solve the two problems, we propose an integrated positioning system based on an unscented Kalman filter, which can integrate the VLP and the inertial navigation data. Miniaturized, mobile and wearable design ensures the system can be easily used in a wide range of application. Experiment results indicate that the accuracy of the system reaches centimeter order, which is better than that of the VLP or inertial navigation alone. The integrated positioning system can provide a more stable, more accurate positioning service with low cost, which makes it a promising candidate for future consumer and industrial indoor positioning applications.

This paper reports a mode-localized resonant accelerometer (ML-RXL) with novel designed micro-lever mechanical coupler to significantly enhance its parametric sensitivity. The experiment results demonstrate the sensitivity of prototype is 32.49 AR/g in the linear operating region of 0.37g, which is one order of magnitude higher than prior work [1] and paves the way for ML-RXL to achieve the sub-µg resolution.

We studied the segregation behavior of multi-size granular particles enclosed in a vertically vibrated container. In the experiment, we applied three kinds of 13X molecular sieve particles. The diameters of these particles are approximately 1:3:6 in the multi-size granular system. Under vibration, the system finally forms a stable hybrid convection mode, and the apparent tendency of the Brazilian nut (BN) effect occurs. The dissipated energy distribution model was introduced, in which the multi-size system was compared with the species competition system. We demonstrated that each kind of particle's top free surface area occupation percentage approximately equals its dissipated energy power percentage in the entire system. The results calculated according to the model well matched the experimental results.

Iron oxide nanoparticle, named synthetic magnetite, as contrast agent has been widely used in clinical MRI. Recently, a new magnetite nanocrystal, called magnetosome, has been found in magnetotactic bacteria. Physicochemical and magnetorelaxo-metric characterization of bacteria magnetosomes and iron oxide nanoparticles are investigated. Bacterial magnetosomes have the larger mean aggregate size, better dispersion and obviously stronger ferromagnetism compared to synthetic magnetites. The samples of several concentrations of magnetic nanoparticles were analyzed using a clinical 3.0 T MR-scanner. The signal decay in the Magnetic Resonance images is found to change proportionally to the nanoparticles concentration. Two kinds of nanoparticles can be though as a negative contrast agent and show slight effects on T1, but strong effects on T2 weighted images. Notice that at the same concentration the signal attenuation of bacterial magnetite samples is more obvious than that of synthetic magnetite samples.

This paper introduces the design of a high magnetic field permanent magnet for Magnetic Birefringence of Vacuum-Q&A Experiment. The magnet is required to have 2.2T maximum magnetic-flux density on the axis of the columnar air-gap (diameter: 27 mm, length: 600 mm) and not more than 0.00001 T magnetic flux leakage at 10 mm away from the round surface of the magnet. Also the magnet should rotate at 10 Hz driven by a motor. All of these requirements have been satisfied. Furthermore, a maximum 2.3 T on the axis of columnar air-gap is reached. The higher magnetic field is reached in two aspects, material and structure. High performance materials, 1J22 and N42M, were adopted and a special structure is designed to shield the leakage flux. Further improvement on this magnet can be achieved with the adoption of more advanced materials.