Cardiac arrhythmia (CA) is a severe cardiac disorder that results in a significant number of fatalities worldwide each year. Conventional electrocardiography (ECG) devices are often unable to detect arrhythmia symptoms during patients’ hospital visits due to their intermittent nature. This paper presents a wearable ECG processor for cardiac arrhythmia (CA) detection. The processor utilizes a Hilbert transform-based R-peak detection engine for R-peak detection, a Haar discrete wavelet transform (HDWT) unit for feature extraction, and a Hybrid ECG classifier that combines linear methods and Non-Linear Support Vector Machines (NLSVM) classifiers to distinguish between normal and abnormal heartbeats. The processor is fabricated by the CMOS 110 nm process with an area of 1.34 mm² and validated with the MIT_BIH Database. The whole design consumes 4.08 μW with an average classification accuracy of 97.34%.

Electrocardiography (ECG) has emerged as a ubiquitous diagnostic tool for the identification and characterization of diverse cardiovascular pathologies. Wearable health monitoring devices, equipped with on-device biomedical artificial intelligence (AI) processors, have revolutionized the acquisition, analysis, and interpretation of ECG data. However, these systems necessitate AI processors that exhibit flexible configuration, facilitate portability, and demonstrate optimal performance in terms of power consumption and latency for the realization of various functionalities. To address these challenges, this study proposes an instruction-driven convolutional neural network (CNN) processor. This processor incorporates three key features: (1) An instruction-driven CNN processor to support versatile ECG-based application. (2) A Processing element (PE) array design that simultaneously considers parallelism and data reuse. (3) An activation unit based on the CORDIC algorithm, supporting both Tanh and Sigmoid computations. The design has been implemented using 110 nm CMOS process technology, occupying a die area of 1.35 mm² with 12.94 µW power consumption. It has been demonstrated with two typical ECG AI applications, including two-class (i.e., normal/abnormal) classification and five-class classification. The proposed 1-D CNN algorithm performs with a 97.95% accuracy for the two-class classification and 97.9% for the five-class classification, respectively.

In recent years, pulse signals have become increasingly important in healthcare, such as cardiovascular disease diagnosis. However, there exists no commonly acknowledged pulse sensing system that simultaneously features stability, accuracy, convenience, and cost efficiency. To address this problem, we present a novel pulse sensing system for human health assessment. The system consists of a T-structured pulse sensor, a circuit system, and a host. Specifically, our sensor has a higher signal-to-noise ratio compared with similar sensors and can be scaled equally, enabling it to greatly reduce the system complexity and cost. Experiments have shown that the pulse signal measured by our system is not only stable, and high fidelity, but also it is suitable for medical assessment, like arrhythmia detection. Furthermore, our system can also perform accurate measurements under different pressures, akin to the pulse diagnosis technique used in Traditional Chinese Medicine. In conclusion, the T-structured pulse sensing system shows great potential for medical and healthcare applications.

To investigate differences in the blood-pressure-related information contributed by the Cun, Guan, and Chi pulse positions, this study proposes a multi-channel weighting analysis method based on pulse signals collected from both wrists. First, pulse data were acquired using a wrist-worn three-channel pulse acquisition system, and systolic blood pressure (SBP) and diastolic blood pressure (DBP) were estimated for each channel through signal preprocessing and a Bayesian-optimization-based support vector regression (BO-SVR) model. Next, for the three channels on a single wrist, a multiple linear regression model with non-negative, sum-to-one constraints was established to estimate channel weights. For the six channels across both wrists, a weighting method based on the inverse of the global mean absolute error (MAE) was adopted to reduce bias caused by inconsistencies between bilateral reference blood pressures. Experimental results showed that, after optimal three-channel fusion, the MAE of SBP/DBP decreased to 4.15/3.06 mmHg for the left wrist and 4.24/3.73 mmHg for the right wrist. The six-channel analysis showed that the ranking of information contributions from the Cun, Guan, and Chi pulse channels on both wrists in blood pressure estimation was consistent with the relative importance traditionally attributed to the heart, liver, kidneys, lung, and spleen in influencing qi and blood in traditional Chinese pulse diagnosis theory. The results indicate that substantial differences exist in the information contributions of different pulse positions to blood pressure estimation. The proposed method provides useful references for explaining multi-channel blood pressure estimation models, optimizing the number and placement of PVDF sensors, and supporting objective research on traditional Chinese pulse diagnosis.