This mini review summarizes some of the recent advances in machine-learning (ML)-driven chemical and biological sensors. Specific focus is on field-effect-transistor (FET)-based sensors with a description of their structures and detection mechanisms. Key ML techniques are briefly reviewed for an audience not familiar with the basic principles. We mainly discuss two aspects: (1) data analysis based on ML and (2) ML applied to sensor design. In conclusion, the challenges and opportunities for the advancement of ML-based sensors are briefly considered.

Wet etching is more economical than dry etching and provides a uniform etching depth regardless of wafer sizes. Typically, potassium hydroxide (KOH) and tetra-methyl-ammonium hydroxide (TMAH) solutions are widely used for the wet etching of silicon. However, there is a limit to the wet etching process when a material deposited on an unetched surface reacts with an etching solution. To solve this problem, in this study, an apparatus was designed and manufactured to physically block the inflow of etchants on the surface using a rubber O-ring. The proposed apparatus includes a heater and a temperature controller to maintain a constant temperature during etching, and the hydrostatic pressure of the etchant is considered for the thin film structure. A corrugation membrane with a diameter of 800 μm, thickness of 600 nm, and corrugation depth of 3 μm with two corrugations was successfully fabricated using the prepared device.

Image processing is one of the major techniques that are used for computer vision. Nowadays, researchers are using machine learning and deep learning for the aforementioned task. In recent years, digit recognition tasks, i.e., automatic meter recognition approach using electric or water meters, have been studied several times. However, two major issues arise when we talk about previous studies: first, the use of the deep learning technique, which includes a large number of parameters that increase the computational cost and consume more power; and second, recent studies are limited to the detection of digits and not storing or providing detected digits to a database or mobile applications. This paper proposes a system that can detect the digital number of meter readings using a lightweight deep neural network (DNN) for low power consumption and send those digits to an Android mobile application in real-time to store them and make life easy. The proposed lightweight DNN is computationally inexpensive and exhibits accuracy similar to those of conventional DNNs.

The electrocardiogram (ECG) is one of the most extensively employed signals used to diagnose and predict cardiovascular diseases (CVDs). In recent years, several deep learning (DL) models have been proposed to improve detection accuracy. Among these, deep neural networks (DNNs) are the most popular, wherein the features are extracted automatically. Despite the increment in classification accuracy, DL models require exorbitant computational resources and power. This causes the mapping of DNNs to be slow; in addition, the mapping is challenging for a wearable device. Embedded systems have constrained power and memory resources. Therefore full-precision DNNs are not easily deployable on devices. To make the neural network faster and more power-efficient, spiking neural networks (SNNs) have been introduced for fewer operations and less complex hardware resources. However, the conventional SNN has low accuracy and high computational cost. Therefore, this paper proposes a new binarized SNN which modifies the synaptic weights of SNN constraining it to be binary (+1 and -1). In the simulation results, this paper compares the DL models and SNNs and evaluates which model is optimal for ECG classification. Although there is a slight compromise in accuracy, the latter proves to be energy-efficient.

In this study, we successfully developed a highly reliable ultrasonic sediment sensor to detect the sediment levels in sewer pipes in harsh environments. The ultrasonic transducer employed in the ultrasonic sediment sensor was designed so as to possess a simple structure. The developed sensor was carefully optimized by simulating the electromechanical characteristics, radiated sound wave pressures, and directivity via finite element analysis. It was also designed to possess a simple mounting structure minimizing the flow disturbance in a 400-mm sewer pipe; additionally, eight ultrasonic transducers were arranged in a four-channel mode, allowing for measurement of the sediment height in five easy steps. Through experimental evaluations, we verified the performance of the ultrasonic sediment-level sensor and its industrial applicability. The results suggested that although the precision value was notably low at 15mm, the sediment detection performance was adequate; therefore, the developed sensor can potentially be used in industrial applications.

To operate an underwater robot in an environment with fluid flow, it is necessary to recognize the speed and direction of the fluid and implement motion control based on these characteristics. Fish have a lateral line that performs this function. In this study, to develop an artificial lateral line sensor that mimics a fish, we developed a method to measure the flow speed and the incident angle of the fluid using a pressure sensor. Several experiments were conducted, and based on the results, the tendency according to the change in the flow speed and the incident angle of the fluid was confirmed. It is believed that additional research can aid in the development of an artificial lateral line sensor.

The motivation of this paper is to easily analyze the properties of nondispersive infrared gas sensor that has more than two different optical path length and to suggest the criterion and definition of infrared light absorbance in order to minimize the measurement errors. With the output voltage ratios and the normalized derivatives of infrared ray (IR) absorbance, when the normalized derivatives of IR absorbance decreases from 0.28 to 0.10, the lower and higher limits of errors were decreased from -5.62% and 2.39% to -4.27% and 2.78%. When the normalized derivatives of IR absorbance were 0.10, the output voltage could be partitioned into two regions with one exponential equation and the temperature compensation error was less than 5%.

In this paper, we describe the development of a force sensor to measure the tool gripping force of the BT50 spindle. The force sensor for a BT50 must be installed inside the gripping force tester; hence, it must be of an appropriate size and have a rated capacity suitable for measuring the gripping force. So, the structure of the force sensor for BT50 was modeled, the size of the sensing part was determined by structural analysis, and the force sensor was manufactured by attaching a strain gauge. The characteristic test results of the manufactured force sensor, indicated that the nonlinearity error, hysteresis error, and reproducibility errors were each within 0.91%, Therefore it was determined that the manufactured force sensor can be used for checking the spindle tool gripping force.

Human hearing sensitivity is frequency-dependent. The sensitivity is low at both ends of the audible frequency, and the sensitivity is the highest in the middle band at 3000 Hz. The heart sound of a healthy person is concentrated at a low frequency of 200 Hz or less, and despite using a stethoscope, the hearing sensitivity of the human body is low, and the stethoscope sound is low. Amplifying the sound of the stethoscope is not effective in distinguishing heart sounds in noisy environments because it maintains the same signal-tonoise ratio. In this study, a method of enhancing auditory stimulation was developed by applying a method of moving the spectrum of auscultation sounds into a high-frequency region where the human body is highly sensitive to hearing. The spectrum of the auscultation sound was moved up by 500 Hz in the frequency domain, and an inverse fast Fourier transform (FFT) was performed to reconstruct the auscultation sound. The heart sounds reconstructed by moving the spectra were divided into the first heart and second heart sound components, as in the original heart sound, and it was confirmed that the intensity was large in the cochleagram representing auditory stimulation. Therefore, this study suggested that spectral shift is a method to enhance auditory stimulation during auscultation without increasing the intensity of the auscultation sound.

We demonstrated a polymer dielectric with low leakage characteristics through an optimal blade coating method for low-cost and large-scale fabrication of metal-insulator-metal (MIM) capacitors. Cross-linked poly(4-vinylphenol) (C-PVP), which is a typically used polymer dielectric, was coated on a 10 × 10 cm indium-tin-oxide (ITO) deposited glass substrate by changing the deposition temperature (TD) and coating velocity (VC) in the blade coating. During the blade coating, the thickness of the thin c-PVP varied depending on TD and VC owing to the 'Landau-Levich (LL) regime'. The c-PVP-dielectric-based MIM capacitor fabricated in this study showed the lowest leakage current characteristics (10-6 A/cm2 at 1.2MV/cm2, annealing at 200 °C) and uniform electrical characteristics when TD was 30 °C and VC was 5mm/s. In addition, at TD = 30 °C, stable leakage characteristics were confirmed when a different electric field was applied. These results are expected to positively contribute to applications with next-generation electronic devices.