Design of Network-on-Chip-Based Restricted Coulomb Energy Neural Network Accelerator on FPGA Device.

Sensor applications in internet of things (IoT) systems, coupled with artificial intelligence (AI) technology, are becoming an increasingly significant part of modern life. For low-latency AI computation in IoT systems, there is a growing preference for edge-based computing over cloud-based alternatives. The restricted coulomb energy neural network (RCE-NN) is a machine learning algorithm well-suited for implementation on edge devices due to its simple learning and recognition scheme. In addition, because the RCE-NN generates neurons as needed, it is easy to adjust the network structure and learn additional data. Therefore, the RCE-NN can provide edge-based real-time processing for various sensor applications. However, previous RCE-NN accelerators have limited scalability when the number of neurons increases. In this paper, we propose a network-on-chip (NoC)-based RCE-NN accelerator and present the results of implementation on a field-programmable gate array (FPGA). NoC is an effective solution for managing massive interconnections. The proposed RCE-NN accelerator utilizes a hierarchical-star (H-star) topology, which efficiently handles a large number of neurons, along with routers specifically designed for the RCE-NN. These approaches result in only a slight decrease in the maximum operating frequency as the number of neurons increases. Consequently, the maximum operating frequency of the proposed RCE-NN accelerator with 512 neurons increased by 126.1% compared to a previous RCE-NN accelerator. This enhancement was verified with two datasets for gas and sign language recognition, achieving accelerations of up to 54.8% in learning time and up to 45.7% in recognition time. The NoC scheme of the proposed RCE-NN accelerator is an appropriate solution to ensure the scalability of the neural network while providing high-performance on-chip learning and recognition.

Hand Gesture Recognition Using FSK Radar Sensors.

Hand gesture recognition, which is one of the fields of human-computer interaction (HCI) research, extracts the user's pattern using sensors. Radio detection and ranging (RADAR) sensors are robust under severe environments and convenient to use for hand gestures. The existing studies mostly adopted continuous-wave (CW) radar, which only shows a good performance at a fixed distance, which is due to its limitation of not seeing the distance. This paper proposes a hand gesture recognition system that utilizes frequency-shift keying (FSK) radar, allowing for a recognition method that can work at the various distances between a radar sensor and a user. The proposed system adopts a convolutional neural network (CNN) model for the recognition. From the experimental results, the proposed recognition system covers the range from 30 cm to 180 cm and shows an accuracy of 93.67% over the entire range.

Chemical Analysis of an Isotopically Labeled Molecule Using Two-Dimensional NMR Spectroscopy at 34 µT.

Low-field nuclear magnetic resonance (NMR) spectroscopy, conducted at or below a few millitesla, provides only limited spectral information due to its inability to resolve chemical shifts. Thus, chemical analysis based on this technique remains challenging. One potential solution to overcome this limitation is the use of isotopically labeled molecules. However, such compounds, particularly their use in two-dimensional (2D) NMR techniques, have rarely been studied. This study presents the results of both experimental and simulated correlation spectroscopy (COSY) on 1-13C-ethanol at 34.38 µT. The strong heteronuclear coupling in this molecule breaks the magnetic equivalence, causing all J-couplings, including homonuclear coupling, to split the 1H spectrum. The obtained COSY spectrum clearly shows the spectral details. Furthermore, we observed that homonuclear coupling between 1H spins generated cross-peaks only when the associated 1H spins were coupled to identical 13C spin states. Our findings demonstrate that a low-field 2D spectrum, even with a moderate spectral line width, can reveal the J-coupling networks of isotopically labeled molecules.

Dual-Emissive Mn-Doped Lead Halide Perovskite Nanocrystals as Background-Suppressed Latent Fingerprint Detection Probes.

Diverse strategies have been developed to visualize latent fingerprints (LFPs) that are undetectable by the naked eye. Among them, fluorescence-based approaches have emerged as an attractive method for enabling high-resolution LFP imaging. However, the use of fluorescent probes for LFP detection remains challenging due to cumbersome processing, low selectivity, and high background interference. Here, we demonstrate highly efficient, sensitive, and background-free LFP detection with dual-color emission arising from manganese (Mn)-doped lead halide perovskite (CsPb(Cl1-yBry)3) nanocrystals (NCs). The resulting bright, fluorescent, solid-state nanopowder (NP) permits the visualization of LFP ridge structures and the resolution of level 1-3 LFP features. The dual-color emission of the Mn-doped perovskite NP provides a simple, robust, and effective route to overcome background interference, thereby increasing the resolution and sensitivity of the LFP detection. The combination of the high quantum efficiency and dual emission of Mn-doped perovskite NP offers great potential for forensic science.

Frequency Limits of Sequential Readout for Sensing AC Magnetic Fields Using Nitrogen-Vacancy Centers in Diamond.

The nitrogen-vacancy (NV) centers in diamond have the ability to sense alternating-current (AC) magnetic fields with high spatial resolution. However, the frequency range of AC sensing protocols based on dynamical decoupling (DD) sequences has not been thoroughly explored experimentally. In this work, we aimed to determine the sensitivity of the ac magnetic field as a function of frequency using the sequential readout method. The upper limit at high frequency is clearly determined by Rabi frequency, in line with the expected effect of finite DD-pulse width. In contrast, the lower frequency limit is primarily governed by the duration of optical repolarization rather than the decoherence time (T2) of NV spins. This becomes particularly crucial when the repetition (dwell) time of the sequential readout is fixed to maintain the acquisition bandwidth. The equation we provide successfully describes the tendency in the frequency dependence. In addition, at the near-optimal frequency of 1 MHz, we reached a maximum sensitivity of 229 pT/Hz by employing the XY4-(4) DD sequence.

FPGA Implementation of Keyword Spotting System Using Depthwise Separable Binarized and Ternarized Neural Networks.

Keyword spotting (KWS) systems are used for human-machine communications in various applications. In many cases, KWS involves a combination of wake-up-word (WUW) recognition for device activation and voice command classification tasks. These tasks present a challenge for embedded systems due to the complexity of deep learning algorithms and the need for optimized networks for each application. In this paper, we propose a depthwise separable binarized/ternarized neural network (DS-BTNN) hardware accelerator capable of performing both WUW recognition and command classification on a single device. The design achieves significant area efficiency by redundantly utilizing bitwise operators in the computation of the binarized neural network (BNN) and ternary neural network (TNN). In a complementary metal-oxide semiconductor (CMOS) 40 nm process environment, the DS-BTNN accelerator demonstrated significant efficiency. Compared with a design approach where BNN and TNN were independently developed and subsequently integrated as two separate modules into the system, our method achieved a 49.3% area reduction while yielding an area of 0.558 mm2. The designed KWS system, which was implemented on a Xilinx UltraScale+ ZCU104 field-programmable gate array (FPGA) board, receives real-time data from the microphone, preprocesses them into a mel spectrogram, and uses this as input to the classifier. Depending on the order, the network operates as a BNN or a TNN for WUW recognition and command classification, respectively. Operating at 170 MHz, our system achieved 97.1% accuracy in BNN-based WUW recognition and 90.5% in TNN-based command classification.

sEMG-Based Hand Gesture Recognition Using Binarized Neural Network.

Recently, human-machine interfaces (HMI) that make life convenient have been studied in many fields. In particular, a hand gesture recognition (HGR) system, which can be implemented as a wearable system, has the advantage that users can easily and intuitively control the device. Among the various sensors used in the HGR system, the surface electromyography (sEMG) sensor is independent of the acquisition environment, easy to wear, and requires a small amount of data. Focusing on these advantages, previous sEMG-based HGR systems used several sensors or complex deep-learning algorithms to achieve high classification accuracy. However, systems that use multiple sensors are bulky, and embedded platforms with complex deep-learning algorithms are difficult to implement. To overcome these limitations, we propose an HGR system using a binarized neural network (BNN), a lightweight convolutional neural network (CNN), with one dry-type sEMG sensor, which is implemented on a field-programmable gate array (FPGA). The proposed HGR system classifies nine dynamic gestures that can be useful in real life rather than static gestures that can be classified relatively easily. Raw sEMG data collected from a dynamic gesture are converted into a spectrogram with information in the time-frequency domain and transferred to the classifier. As a result, the proposed HGR system achieved 95.4% classification accuracy, with a computation time of 14.1 ms and a power consumption of 91.81 mW.

FPGA Implementation of Efficient CFAR Algorithm for Radar Systems.

The constant false-alarm rate (CFAR) algorithm is essential for detecting targets during radar signal processing. It has been improved to accurately detect targets, especially in nonhomogeneous environments, such as multitarget or clutter edge environments. For example, there are sort-based and variable index-based algorithms. However, these algorithms require large amounts of computation, making them difficult to apply in radar applications that require real-time target detection. We propose a new CFAR algorithm that determines the environment of a received signal through a new decision criterion and applies the optimal CFAR algorithms such as the modified variable index (MVI) and automatic censored cell averaging-based ordered data variability (ACCA-ODV). The Monte Carlo simulation results of the proposed CFAR algorithm showed a high detection probability of 93.8% in homogeneous and nonhomogeneous environments based on an SNR of 25 dB. In addition, this paper presents the hardware design, field-programmable gate array (FPGA)-based implementation, and verification results for the practical application of the proposed algorithm. We reduced the hardware complexity by time-sharing sum and square operations and by replacing division operations with multiplication operations when calculating decision parameters. We also developed a low-complexity and high-speed sorter architecture that performs sorting for the partial data in leading and lagging windows. As a result, the implementation used 8260 LUTs and 3823 registers and took 0.6 µs to operate. Compared with the previously proposed FPGA implementation results, it is confirmed that the complexity and operation speed of the proposed CFAR processor are very suitable for real-time implementation.

FMCW Radar Sensors with Improved Range Precision by Reusing the Neural Network.

This paper addresses the challenge of enhancing range precision in radar sensors through supervised learning. However, when the range precision surpasses the range resolution, it leads to a rapid increase in the number of labels, resulting in elevated learning costs. The removal of background noise in indoor environments is also crucial. In response, this study proposes a methodology aiming to increase range precision while mitigating the issue of a growing number of labels in supervised learning. Neural networks learned for a specific section are reused to minimize learning costs and maximize computational efficiency. Formulas and experiments confirmed that identical fractional multiple patterns in the frequency domain can be applied to analyze patterns in other FFT bin positions (representing different target positions). In conclusion, the results suggest that neural networks trained with the same data can be repurposed, enabling efficient hardware implementation.

Method for Improving Range Resolution of Indoor FMCW Radar Systems Using DNN.

Various studies on object detection are being conducted, and in this regard, research on frequency-modulated continuous wave (FMCW) RADAR is being actively conducted. FMCW RADAR requires high-distance resolution to accurately detect objects. However, if the distance resolution is high, a high-modulation bandwidth is required, which has a prohibitively high cost. To address this issue, we propose a two-step algorithm to detect the location of an object through DNN using many low-cost FMCW RADARs. The algorithm first infers the sector by measuring the distance to the object for each FMCW RADAR and then measures the position through the grid according to the inferred sector. This improves the distance resolution beyond the modulation bandwidth. Additionally, to detect multiple targets, we propose a Gaussian filter. Multiple targets are detected through an ordered-statistic constant false-alarm rate (OS-CFAR), and there is an 11% probability that multiple targets cannot be detected. In the lattice structure proposed in this paper, the performance of the proposed algorithm compared to those in existing works was confirmed with respect to the cost function. The difference in performance versus complexity was also confirmed when the proposed algorithm had the same complexity and the same performance, and it was confirmed that there was a performance improvement of up to five-fold compared to those in previous papers. In addition, multi-target detection was shown in this paper. Through MATLAB simulation and actual measurement on a single target, RMSEs were 0.3542 and 0.41002 m, respectively, and through MATLAB simulation and actual measurement on multiple targets, RMSEs were confirmed to be 0.548265 and 0.762542 m, respectively. Through this, it was confirmed that this algorithm works in real RADAR.

High-Speed Continuous Wavelet Transform Processor for Vital Signal Measurement Using Frequency-Modulated Continuous Wave Radar.

This paper proposes a high-speed continuous wavelet transform (CWT) processor to analyze vital signals extracted from a frequency-modulated continuous wave (FMCW) radar sensor. The proposed CWT processor consists of a fast Fourier transform (FFT) module, complex multiplier module, and inverse FFT (IFFT) module. For high-throughput processing, the FFT and IFFT modules are designed with the pipeline FFT architecture of radix-2 single-path delay feedback (R2SDF) and mixed-radix multipath delay commutator (MRMDC) architecture, respectively. In addition, the IFFT module and the complex multiplier module perform a four-channel operation to reduce the processing time from repeated operations. Simultaneously, the MRMDC IFFT module minimizes the circuit area by reducing the number of non-trivial multipliers by using a mixed-radix algorithm. In addition, the proposed CWT processor can support variable lengths of 8, 16, 32, 64, 128, 256, 512, and 1024 to analyze various vital signals. The proposed CWT processor was implemented in a field-programmable gate array (FPGA) device and verified through the measurement of heartbeat and respiration from an FMCW radar sensor. Experimental results showed that the proposed CWT processor can reduce the processing time by 48.4-fold and 40.7-fold compared to MATLAB software with Intel i7 CPU. Moreover, it can be confirmed that the proposed CWT processor can reduce the processing time by 73.3% compared to previous FPGA-based implementations.

FPGA Implementation of an Efficient FFT Processor for FMCW Radar Signal Processing.

This paper presents the design and implementation results of an efficient fast Fourier transform (FFT) processor for frequency-modulated continuous wave (FMCW) radar signal processing. The proposed FFT processor is designed with a memory-based FFT architecture and supports variable lengths from 64 to 4096. Moreover, it is designed with a floating-point operator to prevent the performance degradation of fixed-point operators. FMCW radar signal processing requires windowing operations to increase the target detection rate by reducing clutter side lobes, magnitude calculation operations based on the FFT results to detect the target, and accumulation operations to improve the detection performance of the target. In addition, in some applications such as the measurement of vital signs, the phase of the FFT result has to be calculated. In general, only the FFT is implemented in the hardware, and the other FMCW radar signal processing is performed in the software. The proposed FFT processor implements not only the FFT, but also windowing, accumulation, and magnitude/phase calculations in the hardware. Therefore, compared with a processor implementing only the FFT, the proposed FFT processor uses 1.69 times the hardware resources but achieves an execution time 7.32 times shorter.

Command Recognition Using Binarized Convolutional Neural Network with Voice and Radar Sensors for Human-Vehicle Interaction.

Recently, as technology has advanced, the use of in-vehicle infotainment systems has increased, providing many functions. However, if the driver's attention is diverted to control these systems, it can cause a fatal accident, and thus human-vehicle interaction is becoming more important. Therefore, in this paper, we propose a human-vehicle interaction system to reduce driver distraction during driving. We used voice and continuous-wave radar sensors that require low complexity for application to vehicle environments as resource-constrained platforms. The proposed system applies sensor fusion techniques to improve the limit of single-sensor monitoring. In addition, we used a binarized convolutional neural network algorithm, which significantly reduces the computational workload of the convolutional neural network in command classification. As a result of performance evaluation in noisy and cluttered environments, the proposed system showed a recognition accuracy of 96.4%, an improvement of 7.6% compared to a single voice sensor-based system, and 9.0% compared to a single radar sensor-based system.

Signal Expansion Method in Indoor FMCW Radar Systems for Improving Range Resolution.

As various unmanned autonomous driving technologies such as autonomous vehicles and autonomous driving drones are being developed, research on FMCW radar, a sensor related to these technologies, is actively being conducted. The range resolution, which is a parameter for accurately detecting an object in the FMCW radar system, depends on the modulation bandwidth. Expensive radars have a large modulation bandwidth, use the band above 77 GHz, and are mainly used as in-vehicle radar sensors. However, these high-performance radars have the disadvantage of being expensive and burdensome for use in areas that require precise sensors, such as indoor environment motion detection and autonomous drones. In this paper, the range resolution is improved beyond the limited modulation bandwidth by extending the beat frequency signal in the time domain through the proposed Adaptive Mirror Padding and Phase Correction Padding. The proposed algorithm has similar performance in the existing Zero Padding, Mirror Padding, and Range RMSE, but improved results were confirmed through the ρs indicating the size of the side lobe compared to the main lobe and the accurate detection rate of the OS CFAR. In the case of ρs, it was confirmed that with single targets, Adaptive Mirror Padding was improved by about 3 times and Phase Correct Padding was improved by about 6 times compared to the existing algorithm. The results of the OS CFAR were divided into single targets and multiple targets to confirm the performance. In single targets, Adaptive Mirror Padding improved by about 10% and Phase Correct Padding by about 20% compared to the existing algorithm. In multiple targets, Phase Correct Padding improved by about 20% compared to the existing algorithm. The proposed algorithm was verified through the MATLAB Tool and the actual FMCW radar. As the results were similar in the two experimental environments, it was verified that the algorithm works in real radar as well.

A PAPR Reduction Technique for Fast Touch Sensors Adopting a Multiple Frequency Driving Method on Large Display Panels.

The multiple frequency driving method (MFDM) capacitive touch system (CTS), which drives transmit (TX) electrodes in parallel, has been developed to improve the touch-sensitivity of large touch screens at high speed. However, when driving multiple TX electrodes at the same time, TX signals are merged through the touch panel, which results in increasing the peak-to-average power ratio (PAPR) of combined signals. Due to the high PAPR, the signal is distorted out of the power amplifier's linear range, causing a touch malfunction. The MFDM CTS can avoid this problem by reducing the drive voltage or partially driving the TX electrodes in parallel. However, these methods cause a significant performance drop with respect to signal-to-noise ratio (SNR) in the MFDM systems. This paper proposes a stack method which reduces PAPR effectively without the performance degradation of MFDM and achieves real-time touch sensitivity in large display panels. The proposed method allocates a suitable phase for each TX electrode to reduce the peak power of combined signals. Instead of investigating all of the phases for the total number of TX electrodes, the optimal phase is estimated from the highest frequency to the lowest one and fixed one by one, which can reduce the required time to find a suitable phase considerably. As a result, it enables high-speed sensing of multi-touch on a large touch screen and effectively reduces PAPR to secure high signal-to-noise-ratio (SNR). Through experiments, it was verified that the proposed method in this paper has an SNR of 39.36 dB, achieving a gain of 19.35 and 5.98 dB compared to the existing touch system method and the algorithm used in the communication system, respectively.

FMCW LiDAR System to Reduce Hardware Complexity and Post-Processing Techniques to Improve Distance Resolution.

As the autonomous driving technology develops, research on related sensors is also being actively conducted. One system that is widely used today uses a light source with a wavelength in the 905 nm band for the pulse Light Detection And Ranging (LiDAR) system. This has the disadvantages of being harmful to the human eye and in making digital signal processing difficult at high sampling rates. The Frequency Modulated Continuous Wave (FMCW) LiDAR system has been proposed as an alternative. However, the FMCW LiDAR is formed with a high beat frequency by a method different from that of the FMCW Radar, which causes a hardware burden on the FFT (Fast Fourier Transform) module for interpreting the beat frequency information. In this paper, the FFT module that may occur in the FMCW LiDAR using Digital Down Convert (DDC) technology is extracted at 256 points, which is 25 times smaller than the existing 8192 points, and the beat frequency is 0 to 50 m at 3 cm intervals. As a result of generating and restoring the distance, the performance of 0.03 m Root Mean Square Error (RMSE) compared to the conventional one was confirmed. In this process, the hardware module was directly mounted and verified on the FPGA. In the case of the Simple Threshold-Constant False Alarm Rate (ST-CFAR) provided, the RMSE was measured by generating beat frequencies from 0 to 50 m at 1 cm intervals, and as a result, the result of 0.019 m was confirmed at 0.03 m in the past.

A Frame Detection Method for Real-Time Hand Gesture Recognition Systems Using CW-Radar.

In this paper, a method to detect frames was described that can be used as hand gesture data when configuring a real-time hand gesture recognition system using continuous wave (CW) radar. Detecting valid frames raises accuracy which recognizes gestures. Therefore, it is essential to detect valid frames in the real-time hand gesture recognition system using CW radar. The conventional research on hand gesture recognition systems has not been conducted on detecting valid frames. We took the R-wave on electrocardiogram (ECG) detection as the conventional method. The detection probability of the conventional method was 85.04%. It has a low accuracy to use the hand gesture recognition system. The proposal consists of 2-stages to improve accuracy. We measured the performance of the detection method of hand gestures provided by the detection probability and the recognition probability. By comparing the performance of each detection method, we proposed an optimal detection method. The proposal detects valid frames with an accuracy of 96.88%, 11.84% higher than the accuracy of the conventional method. Also, the recognition probability of the proposal method was 94.21%, which was 3.71% lower than the ideal method.

Wi-Fi Backscatter System With Tag Sensors Using Multi-Antennas for Increased Data Rate and Reliability.

In this paper, we propose tag sensor using multi-antennas in a Wi-Fi backscatter system, which results in an improved data rate or reliability of the signal transmitted from a tag sensor to a reader. The existing power level modulation method, which is proposed to improve data rate in a Wi-Fi backscatter system, has low reliability due to the reduced distance between symbols. To address this problem, we propose a Wi-Fi backscatter system that obtains channel diversity by applying multiple antennas. Two backscatter methods are described for improving the data rate or reliability in the proposed system. In addition, we propose three low complexity demodulation methods to address the high computational complexity problem caused by multiple antennas: (1) SET (subcarrier energy-based threshold) method, (2) TCST (tag's channel state-based threshold) method, and (3) SED (similar Euclidean distance) method. In order to verify the performance of the proposed backscatter method and low complexity demodulation schemes, the 802.11 TGn (task group n) channel model was utilized in simulation. In this paper, the proposed tag sensor structure was compared with existing methods using only sub-channels with a large difference in received CSI (channel state information) values or adopting power-level modulation. The proposed scheme showed about 10 dB better bit error rate (BER) performance and throughput. Also, proposed low complexity demodulation schemes were similar in BER performance with a difference of up to 1 dB and the computational complexity was reduced by up to 60% compared to the existing Euclidean distance method.

IMU Sensor-Based Hand Gesture Recognition for Human-Machine Interfaces.

We propose an efficient hand gesture recognition (HGR) algorithm, which can cope with time-dependent data from an inertial measurement unit (IMU) sensor and support real-time learning for various human-machine interface (HMI) applications. Although the data extracted from IMU sensors are time-dependent, most existing HGR algorithms do not consider this characteristic, which results in the degradation of recognition performance. Because the dynamic time warping (DTW) technique considers the time-dependent characteristic of IMU sensor data, the recognition performance of DTW-based algorithms is better than that of others. However, the DTW technique requires a very complex learning algorithm, which makes it difficult to support real-time learning. To solve this issue, the proposed HGR algorithm is based on a restricted column energy (RCE) neural network, which has a very simple learning scheme in which neurons are activated when necessary. By replacing the metric calculation of the RCE neural network with DTW distance, the proposed algorithm exhibits superior recognition performance for time-dependent sensor data while supporting real-time learning. Our verification results on a field-programmable gate array (FPGA)-based test platform show that the proposed HGR algorithm can achieve a recognition accuracy of 98.6% and supports real-time learning and recognition at an operating frequency of 150 MHz.

SQUID-based ultralow-field MRI of a hyperpolarized material using signal amplification by reversible exchange.

The signal amplification by reversible exchange (SABRE) technique is a very promising method for increasing magnetic resonance (MR) signals. SABRE can play a particularly large role in studies with a low or ultralow magnetic field because they suffer from a low signal-to-noise ratio. In this work, we conducted real-time superconducting quantum interference device (SQUID)-based nuclear magnetic resonance (NMR)/magnetic resonance imaging (MRI) studies in a microtesla-range magnetic field using the SABRE technique after designing a bubble-separated phantom. A maximum enhancement of 2658 for 1H was obtained for pyridine in the SABRE-NMR experiment. A clear SABRE-enhanced MR image of the bubble-separated phantom, in which the para-hydrogen gas was bubbling at only the margin, was successfully obtained at 34.3 µT. The results show that SABRE can be successfully incorporated into an ultralow-field MRI system, which enables new SQUID-based MRI applications. SABRE can shorten the MRI operation time by more than 6 orders of magnitude and establish a firm basis for future low-field MRI applications.