Showing papers in "IEEE Transactions on Biomedical Circuits and Systems in 2020"

A Fully Embedded Adaptive Real-Time Hand Gesture Classifier Leveraging HD-sEMG and Deep Learning

Abstract: This paper presents a real-time fine gesture recognition system for multi-articulating hand prosthesis control, using an embedded convolutional neural network (CNN) to classify hand-muscle contractions sensed at the forearm. The sensor consists in a custom non-intrusive, compact, and easy-to-install 32-channel high-density surface electromyography (HDsEMG) electrode array, built on a flexible printed circuit board (PCB) to allow wrapping around the forearm. The sensor provides a low-noise digitization interface with wireless data transmission through an industrial, scientific and medical (ISM) radio link. An original frequency-time-space cross-domain preprocessing method is proposed to enhance gesture-specific data homogeneity and generate reliable muscle activation maps, leading to 98.15% accuracy when using a majority vote over 5 subsequent inferences by the proposed CNN. The obtained real-time gesture recognition, within 100 to 200 ms, and CNN properties show reliable and promising results to improve on the state-of-the-art of commercial hand prostheses. Moreover, edge computing using a specialized embedded artificial intelligence (AI) platform ensures reliable, secure and low latency real-time operation as well as quick and easy access to training, fine-tuning and calibration of the neural network. Co-design of the signal processing, AI algorithms and sensing hardware ensures a reliable and power-efficient embedded gesture recognition system.

Biointegrated and Wirelessly Powered Implantable Brain Devices: A Review

Abstract: Implantable neural interfacing devices have added significantly to neural engineering by introducing the low-frequency oscillations of small populations of neurons known as local field potential as well as high-frequency action potentials of individual neurons. Regardless of the astounding progression as of late, conventional neural modulating system is still incapable to achieve the desired chronic in vivo implantation. The real constraint emerges from mechanical and physical differences between implants and brain tissue that initiates an inflammatory reaction and glial scar formation that reduces the recording and stimulation quality. Furthermore, traditional strategies consisting of rigid and tethered neural devices cause substantial tissue damage and impede the natural behavior of an animal, thus hindering chronic in vivo measurements. Therefore, enabling fully implantable neural devices requires biocompatibility, wireless power/data capability, biointegration using thin and flexible electronics, and chronic recording properties. This article reviews biocompatibility and design approaches for developing biointegrated and wirelessly powered implantable neural devices in animals aimed at long-term neural interfacing and outlines current challenges toward developing the next generation of implantable neural devices.

Robust Real-Time Embedded EMG Recognition Framework Using Temporal Convolutional Networks on a Multicore IoT Processor

Abstract: Hand movement classification via surface electromyographic (sEMG) signal is a well-established approach for advanced Human-Computer Interaction. However, sEMG movement recognition has to deal with the long-term reliability of sEMG-based control, limited by the variability affecting the sEMG signal. Embedded solutions are affected by a recognition accuracy drop over time that makes them unsuitable for reliable gesture controller design. In this paper, we present a complete wearable-class embedded system for robust sEMG-based gesture recognition, based on Temporal Convolutional Networks (TCNs). Firstly, we developed a novel TCN topology (TEMPONet), and we tested our solution on a benchmark dataset (Ninapro), achieving 49.6% average accuracy, 7.8%, better than current State-Of-the-Art (SoA). Moreover, we designed an energy-efficient embedded platform based on GAP8, a novel 8-core IoT processor. Using our embedded platform, we collected a second 20-sessions dataset to validate the system on a setup which is representative of the final deployment. We obtain 93.7% average accuracy with the TCN, comparable with a SoA SVM approach (91.1%). Finally, we profiled the performance of the network implemented on GAP8 by using an 8-bit quantization strategy to fit the memory constraint of the processor. We reach a $\mathbf {4\times }$ lower memory footprint (460 kB) with a performance degradation of only 3% accuracy. We detailed the execution on the GAP8 platform, showing that the quantized network executes a single classification in 12.84 ms with a power envelope of 0.9 mJ, making it suitable for a long-lifetime wearable deployment.

Miniaturised Wireless Power Transfer Systems for Neurostimulation: A Review

Abstract: In neurostimulation, wireless power transfer is an efficient technology to overcome several limitations affecting medical devices currently used in clinical practice. Several methods were developed over the years for wireless power transfer. In this review article, we report and discuss the three most relevant methodologies for extremely miniaturised implantable neurostimulators: ultrasound coupling, inductive coupling and capacitive coupling. For each powering method, the discussion starts describing the physical working principle. In particular, we focus on the challenges given by the miniaturisation of the implanted integrated circuits and the related ad-hoc solutions for wireless power transfer. Then, we present recent developments and progresses in wireless power transfer for biomedical applications. Last, we compare each technique based on key performance indicators to highlight the most relevant and innovative solutions suitable for neurostimulation, with the gaze turned towards miniaturisation.

Deep Neural Network for Respiratory Sound Classification in Wearable Devices Enabled by Patient Specific Model Tuning

Abstract: The primary objective of this paper is to build classification models and strategies to identify breathing sound anomalies (wheeze, crackle) for automated diagnosis of respiratory and pulmonary diseases. In this work we propose a deep CNN-RNN model that classifies respiratory sounds based on Mel-spectrograms. We also implement a patient specific model tuning strategy that first screens respiratory patients and then builds patient specific classification models using limited patient data for reliable anomaly detection. Moreover, we devise a local log quantization strategy for model weights to reduce the memory footprint for deployment in memory constrained systems such as wearable devices. The proposed hybrid CNN-RNN model achieves a score of $\text{66.31}\%$ on four-class classification of breathing cycles for ICBHI’17 scientific challenge respiratory sound database. When the model is re-trained with patient specific data, it produces a score of $\text{71.81}\%$ for leave-one-out validation. The proposed weight quantization technique achieves ${\approx}4 \times$ reduction in total memory cost without loss of performance. The main contribution of the paper is as follows: Firstly, the proposed model is able to achieve state of the art score on the ICBHI’17 dataset. Secondly, deep learning models are shown to successfully learn domain specific knowledge when pre-trained with breathing data and produce significantly superior performance compared to generalized models. Finally, local log quantization of trained weights is shown to be able to reduce the memory requirement significantly. This type of patient-specific re-training strategy can be very useful in developing reliable long-term automated patient monitoring systems particularly in wearable healthcare solutions.

Hardware Implementation of Deep Network Accelerators Towards Healthcare and Biomedical Applications

Abstract: The advent of dedicated Deep Learning (DL) accelerators and neuromorphic processors has brought on new opportunities for applying both Deep and Spiking Neural Network (SNN) algorithms to healthcare and biomedical applications at the edge. This can facilitate the advancement of medical Internet of Things (IoT) systems and Point of Care (PoC) devices. In this paper, we provide a tutorial describing how various technologies including emerging memristive devices, Field Programmable Gate Arrays (FPGAs), and Complementary Metal Oxide Semiconductor (CMOS) can be used to develop efficient DL accelerators to solve a wide variety of diagnostic, pattern recognition, and signal processing problems in healthcare. Furthermore, we explore how spiking neuromorphic processors can complement their DL counterparts for processing biomedical signals. The tutorial is augmented with case studies of the vast literature on neural network and neuromorphic hardware as applied to the healthcare domain. We benchmark various hardware platforms by performing a sensor fusion signal processing task combining electromyography (EMG) signals with computer vision. Comparisons are made between dedicated neuromorphic processors and embedded AI accelerators in terms of inference latency and energy. Finally, we provide our analysis of the field and share a perspective on the advantages, disadvantages, challenges, and opportunities that various accelerators and neuromorphic processors introduce to healthcare and biomedical domains.

A 13.34 μ W Event-Driven Patient-Specific ANN Cardiac Arrhythmia Classifier for Wearable ECG Sensors

Abstract: Artificial neural network (ANN) and its variants are favored algorithm in designing cardiac arrhythmia classifier (CAC) for its high accuracy. However, the implementation of ultralow power ANN-CAC is challenging due to the intensive computations. Moreover, the imbalanced MIT-BIH database limits the ANN-CAC performance. Several novel techniques are proposed to address the challenges in the low power implementation. Firstly, continuous-in-time discrete-in-amplitude (CTDA) signal flow is adopted to reduce the multiplication operations. Secondly, conditional grouping scheme (CGS) in combination with biased training (BT) is proposed to handle the imbalanced training samples for better training convergency and evaluation accuracy. Thirdly, arithmetic unit sharing with customized high-performance multiplier improves the power efficiency. Verified in FPGA and synthesized in 0.18 μm CMOS process, the proposed CTDA ANN-CAC can classify an arrhythmia within 252 μs at 25 MHz clock frequency with average power of 13.34 μW for 75bpm heart rate. Evaluated on MIT-BIH database, it shows over 98% classification accuracy, 97% sensitivity, and 94% positive predictivity.

A Label-Free, Non-Intrusive, and Rapid Monitoring of Bacterial Growth on Solid Medium Using Microwave Biosensor

Abstract: Microwave resonator sensors are attractive for their contactless and label-free capability of monitoring bacterial growth in liquid media. This paper outlines a new label-free microwave biosensor based on a pair of planar split ring resonators for non-invasive monitoring of bacterial growth on a solid agar media. The sensor is comprised of two split ring resonators with slightly different resonant frequencies for differential operation. The transmission coefficient ( S 21) of the sensor is considered as the sensor's response with a designed and measured quality factor above 200 to ensure a high-resolution operation of the biosensor. Two resonant frequencies of 1.95 and 2.11 GHz represent the sensing signal and the reference signal, respectively. The developed sensor demonstrates high performance in monitoring the growth dynamics of Escherichia coli ( E. coli ) on Luria-Bertani (LB) agar with 4 mm thickness. The sensor's resonant amplitude response demonstrated 0.5 dB variation corresponding to the bacterial growth over 48 hours when bacteria were spread on LB agar starting with initial OD600 = 1.5. Moreover, 0.6 dB change in the sensor's response was observed over 96 hours of bacterial growth starting with an initial OD600 = 1.17 spotted on LB agar. The measured results fit well to the curves created using Richards’ bacterial growth model, showing the strength of the sensor as a potential candidate for use in predictive food microbiology systems.

A Trimodal Wireless Implantable Neural Interface System-on-Chip

Abstract: A wireless and battery-less trimodal neural interface system-on-chip (SoC), capable of 16-ch neural recording, 8-ch electrical stimulation, and 16-ch optical stimulation, all integrated on a 5 × 3 mm2 chip fabricated in 0.35-μm standard CMOS process. The trimodal SoC is designed to be inductively powered and communicated. The downlink data telemetry utilizes on–off keying pulse-position modulation (OOK-PPM) of the power carrier to deliver configuration and control commands at 50 kbps. The analog front-end (AFE) provides adjustable mid-band gain of 55–70 dB, low/high cut-off frequencies of 1–100 Hz/10 kHz, and input-referred noise of 3.46 μVrms within 1 Hz-50 kHz band. AFE outputs of every two-channel are digitized by a 50 kS/s 10-bit SAR-ADC, and multiplexed together to form a 6.78 Mbps data stream to be sent out by OOK modulating a 434 MHz RF carrier through a power amplifier (PA) and 6 cm monopole antenna, which form the uplink data telemetry. Optical stimulation has a switched-capacitor based stimulation (SCS) architecture, which can sequentially charge four storage capacitor banks up to 4 V and discharge them in selected μLEDs at instantaneous current levels of up to 24.8 mA on demand. Electrical stimulation is supported by four independently driven stimulating sites at 5-bit controllable current levels in ±(25–775) μA range, while active/passive charge balancing circuits ensure safety. In vivo testing was conducted on four anesthetized rats to verify the functionality of the trimodal SoC.

From Seizure Detection to Smart and Fully Embedded Seizure Prediction Engine: A Review

Abstract: Recent review papers have investigated seizure prediction, creating the possibility of preempting epileptic seizures. Correct seizure prediction can significantly improve the standard of living for the majority of epileptic patients, as the unpredictability of seizures is a major concern for them. Today, the development of algorithms, particularly in the field of machine learning, enables reliable and accurate seizure prediction using desktop computers. However, despite extensive research effort being devoted to developing seizure detection integrated circuits (ICs), dedicated seizure prediction ICs have not been developed yet. We believe that interdisciplinary study of system architecture, analog and digital ICs, and machine learning algorithms can promote the translation of scientific theory to a more realistic intelligent, integrated, and low-power system that can truly improve the standard of living for epileptic patients. This review explores topics ranging from signal acquisition analog circuits to classification algorithms and dedicated digital signal processing circuits for detection and prediction purposes, to provide a comprehensive and useful guideline for the construction, implementation and optimization of wearable and integrated smart seizure prediction systems.

A Noninvasive Glucose Monitoring SoC Based on Single Wavelength Photoplethysmography

Abstract: Conventional glucose monitoring methods for the growing numbers of diabetic patients around the world are invasive, painful, costly and, time-consuming. Complications aroused due to the abnormal blood sugar levels in diabetic patients have created the necessity for continuous noninvasive glucose monitoring. This article presents a wearable system for glucose monitoring based on a single wavelength near-infrared (NIR) Photoplethysmography (PPG) combined with machine-learning regression (MLR). The PPG readout circuit consists of a switched capacitor Transimpedance amplifier with 1 MΩ gain and a 10-Hz switched capacitor LPF. It allows a DC bias current rejection up to 20 μA with an input-referred current noise of 7.3 pA/√Hz. The proposed digital processor eliminates motion artifacts, and baseline drifts from PPG signal, extracts six distinct features and finally predicts the blood glucose level using Support Vector Regression with Fine Gaussian kernel (FGSVR) MLR. A novel piece-wise linear (PWL) approach for the exponential function is proposed to realize the FGSVR on-chip. The overall system is implemented using a 180 nm CMOS process with a chip area of 4.0 mm2 while consuming 1.62 mW. The glucose measurements are performed for 200 subjects with R2 of 0.937. The proposed system accurately predicts the sugar level with a mean absolute relative difference (mARD) of 7.62%.

A Novel Design Approach for Compact Wearable Antennas Based on Metasurfaces

Abstract: This paper presents a novel approach to design compact wearable antennas based on metasurfaces. The behavior of compact metasurfaces is modeled with a composite right-left handed transmission line (CRLH TL). By controlling the dispersion curve, the resonant modes of the compact metasurface can be tuned efficiently. A printed coplanar waveguide (CPW) monopole antenna is used as the feed structure to excite the compact metasurface, which will result in a low profile antenna with low backward radiation. Following this approach, two compact antennas are designed for wearable applications. The first antenna is designed to operate at its first negative mode (−1 mode), which can realize miniaturization, but maintain the broadside radiation as for a normal microstrip antenna. The proposed prototype resonates around 2.65 GHz, with a matching bandwidth of 300 MHz. The total dimensions of the antenna are 39.4 × 33.4 mm2 (0.1 λ02), and its maximum gain is 2.99 dBi. The second antenna targets dual-band operation at 2.45 and 3.65 GHz. A pair of symmetric modes (±1 modes) are used to generate similar radiation patterns in these two bands. The size of the antenna is 55.79 × 52.25 mm2 (0.2 λ02), and the maximum gains are 4.25 and 7.35 dBi in the two bands, respectively. Furthermore, the performance of the antennas is analyzed on the human body. The results show that the proposed antennas are promising candidates for Wireless Body Area Networks (WBAN).

A Review of Algorithm & Hardware Design for AI-Based Biomedical Applications

Abstract: This paper reviews the state of the arts and trends of the AI-Based biomedical processing algorithms and hardware. The algorithms and hardware for different biomedical applications such as ECG, EEG and hearing aid have been reviewed and discussed. For algorithm design, various widely used biomedical signal classification algorithms have been discussed including support vector machine (SVM), back propagation neural network (BPNN), convolutional neural networks (CNN), probabilistic neural networks (PNN), recurrent neural networks (RNN), Short-term Memory Network (LSTM), fuzzy neural network and etc. The pros and cons of the classification algorithms have been analyzed and compared in the context of application scenarios. The research trends of AI-Based biomedical processing algorithms and applications are also discussed. For hardware design, various AI-Based biomedical processors have been reviewed and discussed, including ECG classification processor, EEG classification processor, EMG classification processor and hearing aid processor. Various techniques on architecture and circuit level have been analyzed and compared. The research trends of the AI-Based biomedical processor have also been discussed.

Memristive Circuit Implementation of Biological Nonassociative Learning Mechanism and Its Applications

Abstract: Biological nonassociative learning is one of the simplest forms of unsupervised learning in animals and can be categorized into habituation and sensitization according to mechanism. This paper proposes a memristive circuit that is based on nonassociative learning and can adapt to repeated inputs, reduce power consumption (habituation), and be sensitive to harmful inputs (sensitization). The circuit includes 1) synapse module, 2) neuron module, 3) feedback module. The first module mainly consists of memristors representing synapse weights that vary with corresponding inputs. Memristance is automatically reduced when a harmful stimulus is input, and climbs at the input interval according to the feedback input when repeated stimuli are input. The second module produces spiking voltage when the total input is above the given threshold. The third module can provide feedback voltage according to the frequency and quantity of input stimuli. Simulation results show that the proposed circuit can generate output signals with biological nonassociative learning characteristics, with varying amplitudes depending on the characteristics of input signals. When the frequency and quantity of the input stimuli are high, the degree of habituation and sensitization intensifies. The proposed circuit has good robustness; can reduce the influence of noise, circuit parasitics and circuit aging during nonassociative learning; and simulate the afterimages caused by visual fatigue for application in automatic exposure compensation.

Non-Invasive Real-Time Monitoring of Glucose Level Using Novel Microwave Biosensor Based on Triple-Pole CSRR

Abstract: Planar microwave sensors are considered an attractive choice to noninvasively probe the dielectric attributes of biological tissues due to their low cost, simple fabrication, miniature scale, and minimum risk to human health. This paper develops and measures a novel microwave biosensor for non-invasive real-time monitoring of glucose level. The design comprises a rectangular plexiglass channel integrated on a triple-pole complementary split ring resonator (TP-CSRR). The proposed sensor operates in the centimeter-wave range 1–6 GHz and is manufactured using PCB on top of an FR4 dielectric substrate. The sensor elements are excited via a coupled microstrip transmission-line etched on the bottom side of the substrate. The integrated CSRR-based sensor is used as a near-field probe to non-invasively monitor the glucose level changes in the blood mimicking solutions of clinically relevant concentrations to Type-2 normal diabetes (70–120 mg/dL), by recording the frequency response of the harmonic reflection and transmission resonances. This indicates the sensor's capability of detecting small variations in the dielectric properties of the blood samples that are responsive to the electromagnetic fields. The proposed sensor is verified through practical measurements of the fabricated design. Experimental results obtained using a Vector Network Analyzer (VNA) demonstrate a sensitivity performance of about 6.2 dB/(mg/ml) for the developed triple-pole sensor that significantly outperforms the conventional single-pole and other proposed sensors in the literature in terms of the resonance amplitude resolution.

A Study of Personal Recognition Method Based on EMG Signal

Abstract: With the increasing development of internet, the security of personal information becomes more and more important Thus, variety of personal recognition methods have been introduced to ensure persons’ information security Traditional recognition methods such as Personal Identification Number (PIN), or Identification tag (ID) are vulnerable to hackers Then the biometric technology, which uses the unique physiological characteristics of human body to identify user information has been proposed But the biometrics widely used at present such as human face, fingerprint, iris, and voice can also be forged and falsified The biometric with living body features such as electromyography (EMG) signal is a good method to achieve aliveness detection and prevent the spoofing attacks However, there are few studies on personal recognition based on EMG signal According to the application context, personal recognition system may operate either in identification mode or verification mode In the personal identification mode, the system recognizes an individual by searching the templates of all the users in the database for a match While in the personal verification mode, the system validates a person's identity by comparing the captured features with her or his own template(s) stored in the system database In this paper, both EMG-based personal identification method and EMG-based personal verification method are investigated First, the Myo armband is placed on the right forearm (specifically, the height of the radiohumeral joint) of 21 subjects to collect the surface EMG signal under hand-open gesture Then, two different methods are proposed for EMG-based personal identification, ie, personal identification method based on Discrete Wavelet Transform (DWT) and ExtraTreesClassifier, and personal identification method based on Continuous Wavelet Transform (CWT) and Convolutional Neural Networks (CNN) Experiments with 21 subjects show that the identification accuracy of this two methods can achieve 99206% and 99203% respectively Then based on the identification method using CWT and CNN, transfer learning algorithm is adopted to solve the model update problem when new data is added Finally, an EMG-based personal verification method using CWT and siamese networks is proposed Experiments show that the verification accuracy of this method can achieve 99285%

Deep Learning Approach for Epileptic Focus Localization

Abstract: The task of epileptic focus localization receives great attention due to its role in an effective epileptic surgery. The clinicians highly depend on the intracranial EEG data to make a surgical decision related to epileptic subjects suffering from uncontrollable seizures. This surgery usually aims to remove the epileptogenic region which requires precise characterization of that area using the EEG recordings. In this paper, we propose two methods based on deep learning targeting accurate automatic epileptic focus localization using the non-stationary EEG recordings. Our first proposed method is based on semi-supervised learning, in which a deep convolutional autoencoder is trained and then the pre-trained encoder is used with multi-layer perceptron as a classifier. The goal is to determine the location of the EEG signal that is responsible for the epileptic activity. In the second proposed method, unsupervised learning scheme is implemented by merging deep convolutional variational autoencoder and K-means algorithm for clustering the iEEG signals into two distinct clusters based on the seizure source. The proposed methods automate and integrate the features extraction and classification processes instead of manually extracting the features as done in the previous studies. Dimensionality reduction is achieved using the autoencoder, while the important spatio-temporal features are extracted from the EEG recordings using the convolutional layers. Moreover, we implemented the inference network of the semi-supervised model on FPGA. The results of our experiments demonstrate high classification accuracy and clustering performance in localizing the epileptic focus compared with the state of the art.

A 0.065-mm 3 Monolithically-Integrated Ultrasonic Wireless Sensing Mote for Real-Time Physiological Temperature Monitoring

Abstract: Accurate monitoring of physiological temperature is important for many biomedical applications, including monitoring of core body temperature, detecting tissue pathologies, and evaluating surgical procedures involving thermal treatment such as hyperthermia therapy and tissue ablation. Many of these applications can benefit from replacing external temperature probes with injectable wireless devices. Here we present such a device for real-time in vivo temperature monitoring that relies on “chip-as-system” integration. With an on-chip piezoelectric transducer and measuring only 380 μm × 300 μm × 570 μm, the 0.065-mm3 monolithic device, in the form of a mote, harvests ultrasound energy for power and transmits temperature data through acoustic backscattering. Containing a low-power temperature sensor implemented with a subthreshold oscillator and consuming 0.813 nW at 37 °C, the mote achieves line sensitivity of 0.088 °C/V, temperature error of +0.22/−0.28 °C, and a resolution of 0.0078 °C rms. A long-term measurement with the mote reveals an Allan deviation floor of <138.6 ppm, indicating the feasibility of using the mote for continuous physiological temperature monitoring.

Ultrasensitive Magnetoelectric Sensing System for Pico-Tesla MagnetoMyoGraphy

Abstract: Magnetomyography (MMG) with superconducting quantum interference devices (SQUIDs) enabled the measurement of very weak magnetic fields (femto to pico Tesla) generated from the human skeletal muscles during contraction. However, SQUIDs are bulky, costly, and require working in a temperature-controlled environment, limiting wide-spread clinical use. We introduce a low-profile magnetoelectric (ME) sensor with analog frontend circuitry that has sensitivity to measure pico-Tesla MMG signals at room temperature. It comprises magnetostrictive and piezoelectric materials, FeCoSiB/AlN. Accurate device modelling and simulation are presented to predict device fabrication process comprehensively using the finite element method (FEM) in COMSOL Multiphysics . The fabricated ME chip with its readout circuit was characterized under a dynamic geomagnetic field cancellation technique. The ME sensor experiment validate a very linear response with high sensitivities of up to 378 V/T driven at a resonance frequency of f res = 7.76 kHz. Measurements show the sensor limit of detections of down to 175 pT/√Hz at resonance, which is in the range of MMG signals. Such a small-scale sensor has the potential to monitor chronic movement disorders and improve the end-user acceptance of human–machine interfaces.

An On-Chip Processor for Chronic Neurological Disorders Assistance Using Negative Affectivity Classification

Abstract: Chronic neurological disorders (CND's) are lifelong diseases and cannot be eradicated, but their severe effects can be alleviated by early preemptive measures. CND's, such as Alzheimer's, Autism Spectrum Disorder (ASD), and Amyotrophic Lateral Sclerosis (ALS), are the chronic ailment of the central nervous system that causes the degradation of emotional and cognitive abilities. Long term continuous monitoring with neuro-feedback of human emotions for patients with CND's is crucial in mitigating its harmful effect. This paper presents hardware efficient and dedicated human emotion classification processor for CND's. Scalp EEG is used for the emotion's classification using the valence and arousal scales. A linear support vector machine classifier is used with power spectral density, logarithmic interhemispheric power spectral ratio, and the interhemispheric power spectral difference of eight EEG channel locations suitable for a wearable non-invasive classification system. A look-up-table based logarithmic division unit (LDU) is to represent the division features in machine learning (ML) applications. The implemented LDU minimizes the cost of integer division by 34% for ML applications. The implemented emotion's classification processor achieved an accuracy of 72.96% and 73.14%, respectively, for the valence and arousal classification on multiple publicly available datasets. The 2 x 3mm2 processor is fabricated using a 0.18 μm 1P6M CMOS process with power and energy utilization of 2.04 mW and 16 μJ/classification, respectively, for 8-channel operation.

The Role and Challenges of Body Channel Communication in Wearable Flexible Electronics

Abstract: Flexible electronics are compatible with film substrates that are soft and stretchable, resulting in conformal integration with human body. Integrated with various sensors and communication ICs, wearable flexible electronics are able to effectively track human vital signs without affecting the body activities. Such a wearable flexible system contains a sensor, a front-end amplifier (FEA), an analog-to-digital converter (ADC), a micro-controller unit (MCU), a radio, a power management unit (PMU), where the radio is the design bottleneck due to its high power consumption. Different from conventional wireless communications, body channel communication (BCC) uses the human body surface as the signal transmission medium resulting in less signal loss and low power consumption. However, there are some design challenges in BCC, including body channel model, backward loss, variable contact impedance, stringent spectral mask, crystalless design, body antenna effect, etc. In this paper, we conduct a survey on BCC transceiver, and analyze its potential role and challenges in wearable flexible electronics.

Flexible Body-Conformal Ultrasound Patches for Image-Guided Neuromodulation

Abstract: The paper presents the design and validation of body-conformal active ultrasound patches with integrated imaging and modulation modalities for image-guided neural therapy. A mechanically-flexible linear 64-element array of piezoelectric transducers with a resonance frequency of 5 MHz was designed for nerve localization. A second 8-element array using larger elements was integrated on the wearable probe for low intensity focused ultrasound neuromodulation at a resonance frequency of 1.3 MHz. Full-wave simulations were used to model the flexible arrays and estimate their generated pressure profiles. A focal depth of 10–20 mm was assumed for beamforming and focusing to support modulation of the vagus, tibial, and other nerves. A strain sensor integrated on the probe provides patient-specific feedback information on array curvature for real-time optimization of focusing and image processing. Each patch also includes high voltage (HV) multiplexers, transmit/receive switches, and pre-amplifiers that simplify connectivity and also improve the signal-to-noise ratio (SNR) of the received echo signals by $\sim$ 5 dB. Experimental results from a flexible prototype show a sensitivity of 80 kPa/V with $\sim$ 3 MHz bandwidth for the modulation and 20 kPa/V with $\sim$ 6 MHz bandwidth for the imaging array. An algorithm for accurate and automatic localization of targeted nerves based on using nearby blood vessels (e.g., the carotid artery) as image markers is also presented.

Bladder Volume Monitoring Using Electrical Impedance Tomography With Simultaneous Multi-Tone Tissue Stimulation and DFT-Based Impedance Calculation Inside an FPGA

Abstract: In this article, a novel method for measuring the volume of the urinary bladder non-invasively is presented that relies on the principles dictated by Electrical Impedance Tomography (EIT). The electronic prototype responsible for injecting innocuous electrical currents to the lower abdominal region and measuring the developed voltage levels is fully described, as well as the computational models for resolution of the so-called Forward and Inverse Problems in Imaging. The simultaneous multi-tone injection of current provided by a high performance Field Programmable Gate Array (FPGA), combined with impedance estimation by the Discrete Fourier Transform (DFT) constitutes a novelty in Urodynamics with potential to monitor continuously the intravesical volume of patients in a much faster and comfortable way than traditional transurethral catheterization methods. The resolution of the Inverse Problem is performed by the Gauss-Newton method with Laplacian regularization, allowing to obtain a sectional representation of the volume of urine encompassed by the bladder and surrounding body tissues. Experimentation has been carried out with synthetic phantoms and human subjects with results showing a good correlation between the levels of abdominal admittivity acquired by the EIT system and the volume of ingested water.

An Optimal Electrical Impedance Tomography Drive Pattern for Human-Computer Interaction Applications

Abstract: In this article, we presented an optimal Electrical Impedance Tomography (EIT) drive pattern based on feature selection and model explanation, and proposed a portable EIT system for applications in human-computer interaction for gesture recognition and contact detection, which can reduce the measurement time and realize a performance trade-off between the accuracy and the time response. In our experiment, eleven hand gestures were designed to verify the proposed approach and EIT system. Compared to the traditional eight-electrode method, the optimal electrode drive pattern achieved a recognition accuracy of 97.5% with seven electrodes and the measurement time was reduced by 60%. To illustrate the universality of this method, we performed a contact detection experiment. By setting seven labels on the conductive panel and using optimal electrode drive pattern, the detection accuracy reached 100% with seven electrodes and the measurement time was reduced by 85%.

A 119dB Dynamic Range Charge Counting Light-to-Digital Converter For Wearable PPG/NIRS Monitoring Applications

Abstract: This paper presents a low power, high dynamic range (DR), reconfigurable light-to-digital converter (LDC) for photoplethysmogram (PPG), and near-infrared spectroscopy (NIRS) sensor readouts. The proposed LDC utilizes a current integration and a charge counting operation to directly convert the photocurrent to a digital code, reducing the noise contributors in the system. This LDC consists of a latched comparator, a low-noise current reference, a counter, and a multi-function integrator, which is used in both signal amplification and charge counting based data quantization. Furthermore, a current DAC is used to further increase the DR by canceling the baseline current. The LDC together with LED drivers and auxiliary digital circuitry are implemented in a standard 0.18 μm CMOS process and characterized experimentally. The LDC and LED drivers consume a total power of 196 μW while achieving a maximum 119 dB DR. The charge counting clock, and the pulse repetition frequency of the LED driver can be reconfigured, providing a wide range of power-resolution trade-off. At a minimum power consumption of 87 μW, the LDC still achieves 95 dB DR. The LDC is also validated with on-body PPG and NIRS measurement by using a photodiode (PD) and a silicon photomultiplier (SIPM), respectively.

A 0.0023 mm $^2$ /ch. Delta-Encoded, Time-Division Multiplexed Mixed-Signal ECoG Recording Architecture With Stimulus Artifact Suppression

Abstract: This article demonstrates a scalable, time-division multiplexed biopotential recording front-end capable of real-time differential- and common-mode artifact suppression. A delta-encoded recording architecture exploits the power spectral density (PSD) characteristics of Electrocorticography (ECoG) recordings, combining an 8-bit ADC, and an 8-bit DAC to achieve 14 bits of dynamic range. The flexibility of the digital feedback architecture is leveraged to time-division multiplex 64 differential input channels onto a shared mixed-signal front-end, reducing channel area by 2x compared to the state-of-the-art. The feedback DAC used for delta-encoding also serves to cancel differential artifacts with an off-chip adaptive loop. Analysis of this architecture and measured silicon performance of a 65 nm CMOS test-chip implementation, both on the bench and in-vivo, are included with this paper.

Flexible Electromagnetic Cap for Head Imaging

Abstract: A wideband wearable electromagnetic (EM) head imaging system for brain stroke detection is presented. The proposed system aims at overcoming the challenges of size, rigidity, and complex structures of existing systems. The proposed system is built into a light-weight and compact imaging platform, which integrates a 16-element antenna array into a highly flexible custom-made wearable cap made of a cost-effective and robust room-temperature-vulcanizing (RTV) silicone. The system mitigates the mismatch between the skin and antenna array by introducing a flexible high-permittivity matching layer. The utilized compact antenna demonstrates wideband operational frequency over 0.6-2.5 GHz with a low signal distortion, safe values of SAR, and unidirectional radiations. The system is experimentally validated on realistic head phantoms. The polar sensitivity encoding (PSE) image processing algorithm is utilized to generate 2D images of different testing scenarios. The obtained images of a stroke-like target inside the head phantoms demonstrate the merits and feasibility of the system for preclinical trials.

Body-Area Powering With Human Body-Coupled Power Transmission and Energy Harvesting ICs

Abstract: This paper presents the body-coupled power transmission and ambient energy harvesting ICs. The ICs utilize human body-coupling to deliver power to the entire body, and at the same time, harvest energy from ambient EM waves coupled through the body. The ICs improve the recovered power level by adapting to the varying skin-electrode interface parasitic impedance at both the TX and RX. To maximize the power output from the TX, the dynamic impedance matching is performed amidst environment-induced variations. At the RX, the Detuned Impedance Booster (DIB) and the Bulk Adaptation Rectifier (BAR) are proposed to improve the power recovery and extend the power coverage further. In order to ensure the maximum power extraction despite the loading variations, the Dual-Mode Buck-Boost Converter (DM-BBC) is proposed. The ICs fabricated in 40 nm 1P8M CMOS recover up to 100 μW from the body-coupled power transmission and 2.5 μW from the ambient body-coupled energy harvesting. The ICs achieve the full-body area power delivery, with the power harvested from the ambiance via the body-coupling mechanism independent of placements on the body. Both approaches show power sustainability for wearable electronics all around the human body.

ResOT: Resource-Efficient Oblique Trees for Neural Signal Classification

Abstract: Classifiers that can be implemented on chip with minimal computational and memory resources are essential for edge computing in emerging applications such as medical and IoT devices This paper introduces a machine learning model based on oblique decision trees to enable resource-efficient classification on a neural implant By integrating model compression with probabilistic routing and implementing cost-aware learning, our proposed model could significantly reduce the memory and hardware cost compared to state-of-the-art models, while maintaining the classification accuracy We trained the resource-efficient oblique tree with power-efficient regularization (ResOT-PE) on three neural classification tasks to evaluate the performance, memory, and hardware requirements On seizure detection task, we were able to reduce the model size by 34× and the feature extraction cost by 146× compared to the ensemble of boosted trees, using the intracranial EEG from 10 epilepsy patients In a second experiment, we tested the ResOT-PE model on tremor detection for Parkinson's disease, using the local field potentials from 12 patients implanted with a deep-brain stimulation (DBS) device We achieved a comparable classification performance as the state-of-the-art boosted tree ensemble, while reducing the model size and feature extraction cost by 106× and 68×, respectively We also tested on a 6-class finger movement detection task using ECoG recordings from 9 subjects, reducing the model size by 176× and feature computation cost by 51× The proposed model can enable a low-power and memory-efficient implementation of classifiers for real-time neurological disease detection and motor decoding

A Real-Time Depth of Anesthesia Monitoring System Based on Deep Neural Network With Large EDO Tolerant EEG Analog Front-End

Abstract: In this article, we present a real-time electroencephalogram (EEG) based depth of anesthesia (DoA) monitoring system in conjunction with a deep learning framework, AnesNET. An EEG analog front-end (AFE) that can compensate ±380-mV electrode DC offset using a coarse digital DC servo loop is implemented in the proposed system. The EEG-based MAC, EEGMAC, is introduced as a novel index to accurately predict the DoA, which is designed for applying to patients anesthetized by both volatile and intravenous agents. The proposed deep learning protocol consists of four layers of convolutional neural network and two dense layers. In addition, we optimize the complexity of the deep neural network (DNN) to operate on a microcomputer such as the Raspberry Pi 3, realizing a cost-effective small-size DoA monitoring system. Fabricated in 110-nm CMOS, the prototype AFE consumes 4.33 μW per channel and has the input-referred noise of 0.29 μVrms from 0.5 to 100 Hz with the noise efficiency factor of 2.2. The proposed DNN was evaluated with pre-recorded EEG data from 374 subjects administrated by inhalational anesthetics under surgery, achieving an average squared and absolute errors of 0.048 and 0.05, respectively. The EEGMAC with subjects anesthetized by an intravenous agent also showed a good agreement with the bispectral index value, confirming the proposed DoA index is applicable to both anesthetics. The implemented monitoring system with the Raspberry Pi 3 estimates the EEGMAC within 20 ms, which is about thousand-fold faster than the BIS estimation in literature.