This paper presents a wireless multimodal physiological monitoring application-specific integrated circuit (ASIC) fabricated in the 180 nm process and assembled in an injectable device to measure photoplethysmography (PPG), electrocardiography (ECG), and subcutaneous body temperature of animals. The device miniaturization is achieved by a shared receiver (Rx) coil and inductive link for both wireless power transfer (WPT) and data communication. An optimized folded L-shape Rx coil enhances the quality (Q) factor, coupling with the transmitter (Tx) coil without increasing the device size. To facilitate WPT, the ASIC alternates between recording and low-power sleep modes, achieving 2.54 mu W ultra-low power consumption on average. The ECG analog frontend (AFE) amplifies the ECG signal with an adjustable gain of 45-80 dB and an input-referred noise (IRN) of 0.99 mu VRMS, while the PPG AFE achieves an IRN of 30.8 pARMS. The temperature AFE provides a 27-47 degrees C sensing range with 0.02 degrees C inaccuracy. The AFE outputs are sequentially quantized by a 10-bit analog- to-digital converter (ADC) with 54.7 dB signal-to-noise-and-distortion ratio (SNDR). For reliable data transmission, the ASIC employs a memory-assisted backscatter scheme, storing data in an 8 Mb memory chip and transmitting it when the inductive link is stable. The ASIC functionality has been validated in vivo on anesthetized rats.
This work presents L-Sort, the first localization-based spike sorting accelerator on a miniaturized chip hardware (FPGA). Spike sorting, crucial for neural activity analysis, involves detecting, extracting, and clustering spikes. L-Sort uses geometric data to localize spikes, reducing the need for extensive waveform retention. The system includes a ZCU104 FPGA board and a computer for real-time visualization. Visitors will feed raw Neuropixel data into the FPGA and observe real-time sorted spike trains.
For decades, sequence-to-sequence alignment is a crucial tool in bioinformatics that identifies the similarity between a query and the reference genome sequence. Because one single reference sequence is insufficient to express the diversity of genes, the concept of the reference genome graph has been proposed, which utilizes a graph structure to represent genetic variations and has a better recording of biological information. As results, different sequence-to-graph alignment algorithms have been developed to address the new demand. Although there are several fast methods available, in the regions enriched with genetic variations, the non-heuristic Smith-Waterman method is still the main approach to guarantee finding the optimal solutions at the cost of lengthy processing time. To alleviate this issue, we design a memory-efficient hardware accelerator for variation-enriched sequence-to-graph alignment on an Intel Arria 10 FPGA. According to our experiment, the proposed 128parallel accelerator operates at 200 MHz, and achieves more than 9X speed-ups when compared to the PaSGAL software for the MHC gene dataset in the human genome GRCh38.
Data security has become a critical issue limiting the rapid development of Wireless Body Area Networks (WBANs). Conventional encryption methods are considered not feasible for WBAN scenarios due to the complexity of the algorithms. Physiological signal-based encryption methods are believed to be potential solutions for WBANs due to the natural randomness of the data. However, physiological data has the drawbacks of high correlation and small key space within a short period. In this paper, a novel Electrocardiogram (ECG)based dynamic encryption scheme for WBANs is proposed. By utilizing the similarity of ECG features in various locations, a novel key distribution method based on security templates is proposed to synchronize the key seeds securely. Then, the designed multidimensional chaotic system dynamically generates key sequences by synchronizing key seeds among different nodes to en/decrypt the information in WBANs. The validation and analysis results demonstrate that the scheme has the advantages of large key space, high randomness and robustness, and low latency compared to the existing methods.
This paper shows a circuit concept for low-energy neuromorphic hardware for Spiking Neural Networks (SNNs). Combining continuous time-encoded values with time-discrete circuitry helps to improve energy-efficiency. Simple design blocks for neurons and synapses make a scalable neural network design process possible. The direct connection between neurons and synapses in an analog manner reduces latency and energy consumption. The functionality was proven by a small neural net solving a classification problem on silicon. Measurements show a FOM of 86.9 fJ for one synaptic operation.
Wireless power transfer has emerged as an essential requirement in implantable brain-machine interfaces (BMIs), providing significant advantages over traditional wired interfaces. However, challenges such as poor wireless power transfer efficiency, limited coverage area and transmission distance, safety, and interference with wireless communication hinder the widespread adoption of wireless power transfer in implantable BMIs. This paper demonstrates that two types of flexible electromagnetic lenses (EM-lens), which is an array of relay resonators formed as metamaterials, can significantly improve power transfer efficiency (PTE). The circular split ring resonator (C-SRR) based EM-lens outperforms in ex-vivo experiments with a 12.07 dB PTE enhancement from 0.06% to 1.03%. The square resonator based EM-lens outperforms in air, improving the PTE from 0.66% to 16.58%. Additionally, EM-lenses placed at two different locations in the subcutaneous and epidural regions are simulated in HFSS to determine the optimal placement. The simulations indicate that epidural placement could achieve higher PTE 4.53% but with increased specific absorption rate (SAR) as a trade-off. Finally, measurements of PTE versus location misalignment show that the EM-lens can maintain power coverage up to 88.58% of the lens area; and improve the power transmission distance by a factor of 53.85% compared to a single transmitter configuration.
Closed-loop deep brain stimulation, a critical technology in brain-machine interfaces, offers a promising treatment for neurological disorders. This paper presents a high-precision closed-loop deep brain stimulation device that allows simultaneous stimulation and recording on the same or adjacent electrode sites. To achieve high-precision closed-loop control, this work utilizes a custom neuromodulation integrated circuit incorporating Redundant-Crossfire stimulator and Frequency Shaping & Quick Blanking based front-end interface. Measurement results show that the stimulation precision reaches 9.3 bits with an 8-channel configuration (13.1 bits for a 4-channel configuration). When recording LFPs in the range of 1 - 100 Hz, the total input reference noise of the system is 3.98 mu V. In addition, long-term animal experiments conducted in pigs demonstrate that the system effectively suppresses external interference and artifacts, providing an effective solution for extended DBS applications.
Neural recording is fundamental to advancements in neuroscience and the development of brain-computer interfaces. Central to this process is the neural amplifier, a critical component that directly influences the quality and fidelity of neural signal acquisition. The amplifier's performance determines the noise level contaminating neural signals and ensures adequate amplification for accurate analog-to-digital conversion (ADC). However, conventional neural amplifiers face significant challenges. They typically require a constant biasing current, leading to inefficient power consumption, and are limited by a constrained gain-bandwidth product due to power budget restrictions. This paper addresses these challenges by delving into the essential requirements for neural amplifiers and discussing the limitations of traditional designs. We introduce a dynamic amplifier that not only surpasses the conventional gain-bandwidth trade-off but also effectively mitigates saturation issues caused by stimulation or motion artifacts. Furthermore, the proposed amplifier exhibits significantly lower thermal and 1/f noise compared to traditional static amplifiers. Our simulation results demonstrate that the proposed neural amplifier consumes less than 200nW of power, with a signal-to-noise-and-distortion ratio (SNDR) of 82.6dB at a 5kHz bandwidth. The amplifier also achieves an input-referred noise of 4.54 mu V-rms. Additionally, the noise-efficient factor is 1.1, highlighting the amplifier's noise performance and its suitability for high-fidelity neural signal acquisition.
This work proposes a wireless intracranial pressure (ICP) monitoring system based on a multi-layer printed coil (MLPC) inductive link. The system is composed of an implantable ICP sensor, a portable controller and a data management terminal. The ICP sensor in our previous work were powered by off-board coil, making the power supply the primary obstacle to implant miniaturization. This paper introduces an MLPC structure, and the corresponding model along with an inductive link based on it. The results show that an MLPC of the same size can provide 3x quality factor of a printed spiral coil. PCB-integrated coil design reduces the size of the ICP sensor by 75% compared with previous work. The proposed ICP monitoring system achieves a 17 mm operating distance and a measurement resolution of 0.3 mmHg
We present a customized approach in development of a fully-spiking neural networks (SNNs) tailored for energy-efficient data-driven signal processing in implantable brain neural interfaces. The importance of a customized design lies in its ability to optimize hardware and energy efficiency while maintaining high classification performance. Our approach allows for the customization of key parameters, including quantization resolution of weights and biases, encoding scheme, encoder placement, temporal resolution, neuron type, and internal parameters such as threshold value, resetting process, and refractory period. We demonstrated the efficacy of this customization in improving hardware and energy efficiency through model development, software-based training and testing, and subsequent synthesis using Verilog RTL on FPGAs and ASIC implementation. Performance evaluation using the CHB-MIT dataset showed an average sensitivity of 92.2% and specificity of 97.3% for seizure detection. The synthesis reports provided insights into the memory, computation, and energy requirements for hardware implementation, highlighting the efficiency and effectiveness of our approach. Our results show that SNN models leads to only 1% drop of sensitivity compared with a 32-bit real-value resolution SCNN model, while offering more than 4 times improvement in memory efficiency.
This paper presents a temperature-robust, high energy-efficiency, integrated and low-power transmitter for 2.4 GHz industrial scientific medical (ISM) band. Fabricated in 65 nm CMOS process, the proposed transmitter with low pulse-position modulation (LPPM) comprises a subthreshold-leakage-current-compensation relaxation oscillator (RxO) and a ring oscillator (RO) composed of delay units with different temperature characteristics. Power efficiency and temperature stability of the transmitter are improved by the above technology. Supporting LPPM and on-off keying (OOK) modulation at 2.4 GHz, the transmitter for Wireless body area network (WBAN) achieves an output power of -14.1 dBm with 20 Mbps and 11 Mbps data rate at 0.6 V, respectively. The power consumption of 99.3 mu W and energy per bit (EPB) of 9.03pJ/bit are measured under LPPM. The frequency variation of temperature compensated RxO and RO are -0.55%similar to 2.25% and -0.23%similar to 0.07%, respectively.
Closed-loop neuromodulation systems have gained significant attention due to their ability to adapt stimulation parameters based on real-time brain activity, potentially reducing individual variability and enhancing treatment efficacy compared to open-loop approaches. This study investigates the impact of system latency on the performance of alpha rhythm phase-locked neuromodulation systems. Offline analysis revealed that signal transmission delays significantly affect phase estimation accuracy. A comparison between BrainAmp and OpenBCI EEG devices showed notable differences in latency, with BrainAmp exhibiting a latency of 74.4 +/- 7.9 ms while the embedded OpenBCI system achieved minimal transmission delays (<< 1 ms). Online implementation demonstrated that despite OpenBCI's lower signal quality, its reduced transmission delays resulted in higher accuracy in phase estimation at the 0-moment, with Mean Absolute Circular Error (MACE) values improving by over 0.25 in both eyes-open and eyes-closed conditions. The findings emphasize the importance of minimizing system latencies in closed-loop neuromodulation systems to enhance the effectiveness of phase-locked stimulation. By addressing both algorithmic sophistication and system latency, this research contributes to the development of more precise and efficient neuromodulation techniques for applications in neurological and psychiatric disorders.
In this study, we propose to optimize the irradiated light in the front light structure for lensless fluorescence imaging devices by controlling the optical polarization. We fabricated strip diffraction on a thin-cover glass, which operates with a hybrid filter on the image sensor. The control of the excitation light polarization was coupled to diffractor substrates. The results confirm that the effect of P-polarized light can significantly increase the intensity of irradiance light in front of the lensless imaging device, as observed through the contrast of the fluorescent beads.
This paper introduces a capacitor-less n-type Low Dropout Regulator (LDO) designed to enhance transient response and expand the output voltage and current range, making it suitable for low-power biomedical System-on-Chip (SoC) applications. Utilizing a Recycling Folded Cascode (RFC) configuration to implement a common-mode feedback loop, this design improves the gain and bandwidth of the regulator. Additionally, the buffer used in this design is a Super Source Follower (SSF) with a pull-up capability structure that enhances the gate voltage of the power transistor Furthermore, an output voltage peak detection circuit is incorporated to implement a fast feedback loop, improving transient response and compensating for loop stability. The proposed LDO structure covers input voltages from 1.4V to 2.0V and generates an output voltage range from 1.2V to 1.8V with a dropout voltage of 200mV. This design is implemented using a 65-nm CMOS process with an effective area of 0.0198 mm(2). Simulation results indicate that with an input voltage of 1.8V and an edge time of 400ns, during load current transitions from 1mA to 200mA, the recorded overshoot and undershoot values are 78mV and 54mV, respectively. The proposed LDO achieves a minimal Figure of Merit (FOM).
Drowsiness and fatigue pose significant risks across various industries, causing 15-20% of severe road crashes in the driving industry alone. Wearable devices are a promising approach for detecting drowsiness since they do not require modifications to existing vehicles. Common wearable approaches are based on PPG signals, and EEG-based solutions are also gaining more popularity. However, poor signal quality, user discomfort, and lack of computational capabilities for low-power processing at the edge hinder the development of standalone wearable drowsiness detection systems. This paper introduces a novel drowsiness detection system based on BioGAP, a compact acquisition and processing platform for heterogeneous biosignals powered by the GAP9 parallel ultra-low-power System-onChip. BioGAP is integrated into a comfortable, non-stigmatizing headband and acquires data from 8 fully-dry EEG channels and a PPG sensor on the earlobe. Five subjects performed drowsiness experiments wearing the headband while engaged in a professional car driving simulator. A lightweight Convolutional Neural Network (CNN) enables sensor fusion between EEG and PPG, achieving an average accuracy of 91.1% in detecting drowsy states (+6% increase compared to using a single modality only). The network is lightweight (21.6k parameters) and is deployed on GAP9, demonstrating an inference time of 10.17 ms, energy per inference of 0.36 mJ, and average system power consumption of only 19.6 mW, thereby enabling continuous operation for 14 h in realistic driving conditions when powered by a small 75 mAh battery.
Wearable electrocardiogram (ECG) systems have shown significance in daily health monitoring. Traditional ECG systems are characterized by bulky and stationary equipment as well as high power consumption, which restricts the applications in long-term wearable applications [1]. In this work, we demonstrate a wearable BLE ECG monitoring system on human body, which is based on the design of a low-power highly-integrated physiological signal-acquisition chip. The chip is fabricated in 65 nm CMOS process, incorporating an 8 -channel analog front end (AFE), a 10-bit successive approximation register analog-to-digital converter (SAR-ADC). In addition, the power management is also integrated on the chip, which enables the entire system to dissipate low power from a single battery. Measured results show that the physiological signal-acquisition chip consumes 249 mu W in all, which features: 1) The AFE realizes a gain of 43.3 dB, an input referred noise (IRN) of 9.68 mu Vrms, a bandwidth of 0.9 Hz-7.2kHz, and a common-mode rejection ratio (CMRR) of 82.75dB. 2) The SAR ADC shows an effective number of 9.66 bits with 315nW power. 3) The wearable wireless ECG system takes 18 mW power and 16.8 g weight. Demonstrated on a human body, the system can send the ECG signals to a smartphone through BLE.
Challenging gait, particularly under destabilized or perturbing conditions, adversely impact elderly individuals and those with neurological impairments, increasing their risk of falls and consequent injuries. This study presents a novel powered hip exoskeleton designed to enhance one’s postural stability to unexpected perturbations through autonomous hip modulation based on force feedback from foot-ground interactions. The exoskeleton includes multi-planar actuators for controlled hip flexion/extension and abduction/adduction, systematically responding to dynamic loading changes based on bipedal plantar pressure data and center of pressure (CoP) trajectories. A balance-assisted controller, inspired by human adaptive control strategies, is integrated into the exoskeleton to provide customized support, enhancing posture control under sudden translational perturbation conditions. Preliminary experiments involving a healthy subject wearing the designed hip exoskeleton demonstrated improved balance responses to unexpected perturbations, as evidenced by a significant 24% reduction in the muscle activation of the gluteus medius (GM).
This paper presents a Discrete Wavelet Transform based poolformer network model (DWT-Poolformer) for real-time electrocardiogram (ECG) signal processing on wearable devices. By integrating a FPGA-based CWT module together with advanced quantization techniques, this model addresses the computation and resource constraints challenges faced by the wearable devices. This design achieved optimum balance between computational efficiency and diagnostic accuracy. Tested with the MIT-HIB dataset, this model achieved high accuracy of 99.4 2% with 4 -bit weight resolution and 8 -bit activation resolution while significantly reducing computational load and memory usage. The results highlight the potential of this model for real-time diagnostic applications in AI-powered wearable healthcare devices.
Real-time monitoring of vital signs in wearable devices focuses on accurate photoplethysmogram (PPG) and electrocardiogram (ECG) recording. Motion artifacts, ambient light interference and sensor-skin contact variability affect signal quality significantly, demanding a multi-channel sensor interface chip with a high dynamic range and energy efficiency. A PPG/ECG SoC is proposed for robust signal optimization. Time-division multiplexing, ambient double sampling and DC current compensation together enhance the dynamic range. Fabricated in 0.18 mu m CMOS technology, the chip features a 5.63-pArms direct digitized input-referred noise for PPG readout and 365-nVrms for ECG. A cross-scale dynamic range of 133dB is achieved, providing saturation-free usage for sport wearable devices such as smart rings.
The peripheral nervous system (PNS) facilitates communication between the brain and various organs. Advanced PNS neural interfaces can help in restoring motor functions for patients suffering from spinal cord injuries, amputations, and other conditions. The efficacy of these neural interfaces, and the precision of sensory activity decoding is heavily reliant on artificial intelligence techniques at the backend. Traditional machine learning techniques, such as feature-based classification, generally requires extensive feature engineering and domain expertise, and is often ineffective at handling very high-dimensional data. Additionally, convolutional neural networks (CNNs) tend to perform best when employing floating-point operations, which in-turn can be computationally intensive and less energy-efficient. Addressing these challenges, this paper presents a multiplierless spiking neural network (SNN) that utilizes fewer neurons to achieve higher accuracy. Leveraging the discrete firing characteristics of SNNs, we propose a hardware implementation model, trained on experimentally recorded action potentials from rat peripheral nerves. This demonstrates a significant improvement in Macro F1-score, reaching 0.89, over various CNNs while using 99.8% fewer parameters. It fully decodes three degrees of rat motion (dorsiflexion, plantarflexion, and pricking stimulation), showcasing the potential for efficient and accurate hardware integration. This work highlights a path towards the development of next-generation hardware-assisted neural interfaces.
