Improving access to health care services for the medically under-served population is vital ensure that critical illness can be addressed immediately. In scenarios where there a severely lacking of skilled medical staff, basic lung sound classification through digital stethoscope used provide an immediate diagnostic respiratory-related diseases such as chronic obstructive pulmonary. this work, we have developed improved bi-ResNet deep learning architecture, LungBRN, which uses STFT and...

An event-driven system generates samples only when a predefined event is triggered, thus the power consumption tracks input activities leading to significant savings in for wearable sensors. In this brief, we presented an artificial neural network (ANN) core cardiac arrhythmia classifier (CAC). The proposed data alignment mechanism allows seamless cooperation between ANN CAC and clockless analog front-end. Measurement results show that consumes merely 1.3μW dynamic at heart rate of 75bpm...

We demonstrate a new digital stethoscope system, LungSys, for our users to detect adventitious respiratory sounds automatically. LungSys includes commercial and software application installed on an Android mobile tablet. The converts acoustic sound from the users' chest electronic signals transmits tablet through built-in Bluetooth device. Our custom in provides real-time analysis of lung using proposed neural network model bi-ResNet(BRN) identifies any users. Since is based non-invasive...

Reinforcement learning (RL) based dynamic voltage and frequency scaling (DVFS) is an effective approach to balance performance power consumption for video rendering applications. To approximate the "God's eye view" regulation, CPU-GPU should be fine-grained regulated SoC have fully observed by a RL-based DVFS governor. Fine-grained regulation with traditional value-based RL governor suffers action space explosion it impossible observable SoC. address these two issues, on deep recurrent...

A wearable electrocardiogram (ECG) device is an effective tool for managing cardiovascular diseases. This paper presents a low power clinician-like cardiac arrhythmia watchdog (CAW) ECG devices. The CAW based on novel P-QRS-T detection algorithm that makes use of clinical features to identify abnormalities. Implemented in 0.18 μm CMOS process, the consumes 2.66 µW 80 bpm heart rate at 1.2 V supply with area 0.578 mm2. Verified QT database, average sensitivity/positive predictivity P-wave,...

Atrial fibrillation (AF) is a common type of cardiac arrhythmia and silent killer which will affect 12.1 million people in USA 2030 according to CDC. It possible identify AF at early stage. An ECGon-chip for wearable detector needs possess following features: (1) $\mathrm{G} \Omega$ input impedance deal with dry-electrodes that have contact $M level; (2) Good noise performance capture small ECG details; (3) Embedded detection detect sporadic events minimize wireless transmission. The prior...

12-lead electrocardiogram (ECG) delineation is a critical step in diagnosing of various heart diseases. Current practices for ECG typically involve processing each the 12 leads separately using network, which computationally expensive. To solve this issue, 1-12 mapping strategy proposed to directly map one lead network predictions other and then fine-tune boundaries. CNN-BiLSTM autoencoder architecture employed model sequential dependencies signal. Besides, data augmentation mixed losses are...

Malicious manipulation of very large-scale integration physical-layout design is a serious problem in modern integrated circuit design. The database requires highly compressed secured storage medium. In this article, we propose compressive asymmetrical convolutional auto-encoder (ACAE) machine learning framework, CompressKey, which performs layout compression and encryption simultaneously. It utilizes geometric features to eliminate redundancies patterns. We “Divide Merge” technique...

Artifacts in ExG recordings resulting from motion, environment interferences and stimulation can manifest as common mode (CM), differential (DM) or hybrid of CM/DM signals with up to hundreds mV. To mitigate these artifacts minimize information loss, a sensor interface (SI) high input range fast response speed is required. In contrast the use power-hungry dynamic (DR) ADCs, various artifact recovery approaches including adaptive filter [1], track-and-zoom [2], gain control [3] common-mode...

This paper presents the analysis and behavioral modeling of two high-order DAC linearity enhancement techniques constructs a third-order noise-shaping (NS) SAR ADC with full-bit mismatch-shaped capacitor DAC(CDAC) to verify performance. The CDAC is divided into an MSB segment LSB segments, where VMS second-order MES are utilized deal capacitance mismatch respectively. With model, optimizations focused on transfer function (MTF) length made. simulation results obtained system achieve 113.25dB...

This paper presents a dry electrode friendly interface circuit for flexible ECG sensors. The high input impedance and low noise of the enhance signal quality electrodes. Its performance is demonstrated by wireless sensor, which supports 5 days continuous monitoring with 30mAh battery.

The extracted energy of the thermal harvester circuits are affected by input and output power status. A boost buck reconfigurable with store, supply, extract recycle modes proposed in this paper to harvest from TEG at any possible chances. In addition, maximum point tracking zero current switching also customized for reconfigurability. Moreover, Colpitts oscillator Dickson charge pump based startup circuit adopted millivolts startup. Simulation shows that design end-to-end conversion...

A predictive control algorithm, artificial neural-network-based (ANNPC), is proposed. This algorithm adopts both the neural net (NN) information processing mode and a mechanism. It shows great promise for conventional NN optimal control, can deal with complicated nonlinear systems. Simulation comparisons between ANNPC are given.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Multi-modal biomedical time series (MBTS) data offers a holistic view of the physiological state, holding significant importance in various bio-medical applications. Owing to inherent noise and distribution gaps across different modalities, MBTS can be complex model. Various deep learning models have been developed learn representations but still fall short robustness due ignorance modal-to-modal variations. This paper presents multi-scale multi-modal representation (MBSL) network with...

Fiducial points detection serves as the cornerstone of physiological signal processing. Conventional multimodal biomedical sensors require deployment diverse or reconfigurable algorithms to handle different types signals. However, these approaches can be complex and costly in terms hardware, more processing time. To address challenges, a unified adaptive thresholding method enhanced with synthesizer is proposed signals get R-peak, systolic peak spike, respectively. The approach evaluated on...

Spike sorting is an effective approach for analysis of neuron activities. With the increasing number recording channels, online spike seen as a promising solution to relax wireless transmission burden. Deep learning based methods provide superior accuracy but require intensive computation offsetting efficient transmission. To cut down cost, scattering convolution network (SCN) proposed extract features via wavelet transform, then lightweight convolutional neural (CNN) capable spikes...

Binary phase shift-keying (BPSK) modulation shows great potential for wireless data and power transmission via single inductive link. Conventional BPSK suffers from the inefficiency of wasting cycles on bits that remain unchanged. An adaptive (A-BPSK) encoding-decoding mechanism is proposed in this paper to encode unchanged with merely one carrier cycle, greatly enhancing capacity. Applying a Class-E link, an A-BPSK chipset implemented 0.18μm BCD process. At 13.56MHz carrier, post-layout...