Xinzi Xu
Xinzi Xu has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
- Yongfu Li
Shanghai Jiao Tong University
7 shared publications - Qiao Cai
Shanghai Jiao Tong University
6 shared publications - Yang Zhao
School of Integrated Circuits, Shanghai Jiao Tong University, Shanghai, China
5 shared publications - Guoxing Wang
Shanghai Jiao Tong University
3 shared publications - Yanxing Suo
Shanghai Jiao Tong University
3 shared publications - Yunfang Zhang
School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China
3 shared publications
Latest Publications
- 6 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Circuits and Systems
- Vol. 19 (6) , 1160-1174
Artifact-Tolerant Electrophysiological Sensor Interface With 3.6V/1.8V DM/CM Input Range and 52.3mV pp /${\mu}$s Recovery Using Asynchronous Signal Folding
In the practical implementations of wearable sensors, motion artifacts with large amplitudes often cause signal chain saturation, significantly degrading biopotential signal integrity. Similarly, rapid stimulation artifacts are inevitable during closed-loop brain stimulation therapy, posing additional challenges for real-time signal acquisition. To address motion and stimulation artifacts with amplitudes reaching hundreds of mV while minimizing information loss, a sensor interface with high input range and fast artifacts recovery capability is essential. This paper presents a continuous-time track-and-zoom (CT-TAZ) technique designed to handle large artifacts events without saturation. The proposed system achieves a 3.6V/1.8V differential-mode/common-mode full-scale input range. Fabricated in a 180nm CMOS process, the prototype chip occupies an area of 0.694mm2 and consumes 12/32.6/51.6μW for recordings without/with single-end/ with differential rail-to-rail artifacts. The system demonstrates an average artifacts recovery time of 65.3 μs under 3.6V stimulation artifacts, achieving an average artifacts recovery speed of 52.3mVpp/μs, which is 2.25× larger input range and 3× faster recovery compared to the state-of-the-art.
- 6 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Circuits and Systems
- Vol. 19 (6) , 1160-1174
Artifact-Tolerant Electrophysiological Sensor Interface With 3.6V/1.8V DM/CM Input Range and 52.3mV pp /${\mu}$s Recovery Using Asynchronous Signal Folding
In the practical implementations of wearable sensors, motion artifacts with large amplitudes often cause signal chain saturation, significantly degrading biopotential signal integrity. Similarly, rapid stimulation artifacts are inevitable during closed-loop brain stimulation therapy, posing additional challenges for real-time signal acquisition. To address motion and stimulation artifacts with amplitudes reaching hundreds of mV while minimizing information loss, a sensor interface with high input range and fast artifacts recovery capability is essential. This paper presents a continuous-time track-and-zoom (CT-TAZ) technique designed to handle large artifacts events without saturation. The proposed system achieves a 3.6V/1.8V differential-mode/common-mode full-scale input range. Fabricated in a 180nm CMOS process, the prototype chip occupies an area of 0.694mm2 and consumes 12/32.6/51.6μW for recordings without/with single-end/ with differential rail-to-rail artifacts. The system demonstrates an average artifacts recovery time of 65.3 μs under 3.6V stimulation artifacts, achieving an average artifacts recovery speed of 52.3mVpp/μs, which is 2.25× larger input range and 3× faster recovery compared to the state-of-the-art.
- 18 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 71 (5) , 2119-2131
A Lightweight DRDPG-Based RL DVFS for Video Rendering on CPU-GPU Integrated SoC
Reinforcement learning (RL) based dynamic voltage and frequency scaling (DVFS) is an effective approach to balance performance and power consumption for video rendering applications. To approximate the “God’s eye view” regulation, the CPU-GPU should be fine-grained regulated and the SoC have to be fully observed by a RL-based DVFS governor. Fine-grained regulation with traditional value-based RL governor suffers action space explosion and it is impossible to have a fully observable SoC. To address these two issues, a governor based on deep recurrent deterministic gradient (DRDPG) governor is proposed. The governor is based on Deterministic Policy Gradient (DDPG) algorithm with embedded recurrent neural network (RNN). The DDPG algorithm guarantees fine-grained power regulation without action space explosion and the RNN-FC network topology mitigates the partial observability issue. Evaluated on the Nvidia Jetson NX platform, the proposed DRDPG governor achieves over 19% better regulation efficiency compared with Linux default governors and shows superior regulation efficiency to other RL-based state-of-the-arts. Implemented in a 55-nm CMOS process, the proposed governor draws merely 2.16-mA from a 1.2-V supply at 1-MHz clock occupying a silicon area of 0.075mm $^2$ .
- 18 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 71 (5) , 2119-2131
A Lightweight DRDPG-Based RL DVFS for Video Rendering on CPU-GPU Integrated SoC
Reinforcement learning (RL) based dynamic voltage and frequency scaling (DVFS) is an effective approach to balance performance and power consumption for video rendering applications. To approximate the “God’s eye view” regulation, the CPU-GPU should be fine-grained regulated and the SoC have to be fully observed by a RL-based DVFS governor. Fine-grained regulation with traditional value-based RL governor suffers action space explosion and it is impossible to have a fully observable SoC. To address these two issues, a governor based on deep recurrent deterministic gradient (DRDPG) governor is proposed. The governor is based on Deterministic Policy Gradient (DDPG) algorithm with embedded recurrent neural network (RNN). The DDPG algorithm guarantees fine-grained power regulation without action space explosion and the RNN-FC network topology mitigates the partial observability issue. Evaluated on the Nvidia Jetson NX platform, the proposed DRDPG governor achieves over 19% better regulation efficiency compared with Linux default governors and shows superior regulation efficiency to other RL-based state-of-the-arts. Implemented in a 55-nm CMOS process, the proposed governor draws merely 2.16-mA from a 1.2-V supply at 1-MHz clock occupying a silicon area of 0.075mm $^2$ .
- 19 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
A Lightweight CNN Spike Sorting Method Enhanced by Scattering Convolution Network
Spike sorting is an effective approach for analysis of neuron activities. With the increasing number of recording channels, online spike sorting is seen as a promising solution to relax the wireless transmission burden. Deep learning based methods provide superior spike sorting accuracy but require intensive computation offsetting the efficient transmission. To cut down the computation cost, a scattering convolution network (SCN) is proposed to extract features via wavelet scattering transform, then a lightweight convolutional neural network (CNN) is capable of sorting the spikes accurately. Experiment results demonstrate that the proposed SCN-CNN method achieves over 92% computation reduction and maintains a high classification accuracy of up to 99.97%, showing huge potential for online spike sorting on large-scale neural probes.
- 19 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
A Lightweight CNN Spike Sorting Method Enhanced by Scattering Convolution Network
Spike sorting is an effective approach for analysis of neuron activities. With the increasing number of recording channels, online spike sorting is seen as a promising solution to relax the wireless transmission burden. Deep learning based methods provide superior spike sorting accuracy but require intensive computation offsetting the efficient transmission. To cut down the computation cost, a scattering convolution network (SCN) is proposed to extract features via wavelet scattering transform, then a lightweight convolutional neural network (CNN) is capable of sorting the spikes accurately. Experiment results demonstrate that the proposed SCN-CNN method achieves over 92% computation reduction and maintains a high classification accuracy of up to 99.97%, showing huge potential for online spike sorting on large-scale neural probes.
- 19 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
An Adaptive BPSK Modulated Wireless Inductive Link for Implantable Devices
Binary phase shift-keying (BPSK) modulation shows great potential for wireless data and power transmission via single inductive link. Conventional BPSK modulation suffers from the inefficiency of wasting modulation cycles on bits that remain unchanged. An adaptive BPSK (A-BPSK) encoding-decoding mechanism is proposed in this paper to encode the unchanged bits with merely one carrier cycle, greatly enhancing the transmission capacity. Applying on a Class-E inductive link, an A-BPSK wireless transmission chipset is implemented in a 0.18μm BCD process. At a 13.56MHz carrier, post-layout simulation shows a minimum and maximum data rates of 3.87Mbps and 13.56Mbps, respectively. Statistically, the average data rate can be 6.02Mbps showing 2X over speed up.
- 19 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
An Adaptive BPSK Modulated Wireless Inductive Link for Implantable Devices
Binary phase shift-keying (BPSK) modulation shows great potential for wireless data and power transmission via single inductive link. Conventional BPSK modulation suffers from the inefficiency of wasting modulation cycles on bits that remain unchanged. An adaptive BPSK (A-BPSK) encoding-decoding mechanism is proposed in this paper to encode the unchanged bits with merely one carrier cycle, greatly enhancing the transmission capacity. Applying on a Class-E inductive link, an A-BPSK wireless transmission chipset is implemented in a 0.18μm BCD process. At a 13.56MHz carrier, post-layout simulation shows a minimum and maximum data rates of 3.87Mbps and 13.56Mbps, respectively. Statistically, the average data rate can be 6.02Mbps showing 2X over speed up.
- 19 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 36, 1-5
A Peak Detection Algorithm for Multimodal Physiological Signals Based on U-Net
Fiducial points detection serves as the cornerstone of physiological signal processing. Conventional multimodal biomedical sensors require the deployment of diverse or reconfigurable algorithms to handle different types of signals. However, these approaches can be complex and costly in terms of hardware, or require more processing time. To address these challenges, a unified adaptive thresholding method enhanced with a signal synthesizer is proposed to processing multimodal physiological signals to get R-peak, systolic peak and spike, respectively. The proposed approach is evaluated on MIT-BIH, MIMIC and Quiroga databases, achieving precision of 99.03%, 99.38%, 98.81% and recall rate of 99.34%, 99.63%, 98.14% for R-peak, systolic-peak and spike-potential, respectively. These results demonstrate a promising energy-efficient solution for multimodal real-time monitoring.
- 19 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 36, 1-5
A Peak Detection Algorithm for Multimodal Physiological Signals Based on U-Net
Fiducial points detection serves as the cornerstone of physiological signal processing. Conventional multimodal biomedical sensors require the deployment of diverse or reconfigurable algorithms to handle different types of signals. However, these approaches can be complex and costly in terms of hardware, or require more processing time. To address these challenges, a unified adaptive thresholding method enhanced with a signal synthesizer is proposed to processing multimodal physiological signals to get R-peak, systolic peak and spike, respectively. The proposed approach is evaluated on MIT-BIH, MIMIC and Quiroga databases, achieving precision of 99.03%, 99.38%, 98.81% and recall rate of 99.34%, 99.63%, 98.14% for R-peak, systolic-peak and spike-potential, respectively. These results demonstrate a promising energy-efficient solution for multimodal real-time monitoring.
- 2 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 42 (4) , 1030-1043
CompressKey—Near Lossless Layout Compression and Encryption Using Convolutional Auto-Encoder Model and Expansion-Reduction Pattern Techniques
Malicious manipulation of very large-scale integration physical-layout design is a serious problem in modern integrated circuit design. The physical-layout design database requires a highly compressed secured storage medium. In this article, we propose a secured compressive asymmetrical convolutional auto-encoder (ACAE) machine learning framework, CompressKey, which performs layout compression and encryption simultaneously. It utilizes geometric features to eliminate redundancies in layout patterns. We propose a “Divide and Merge” technique to partition each layer into smaller sizes of unique patterns to address the inconsistency of layout pattern complexity. We also propose “Matrix Expansion” and “Matrix Reduction” techniques on the matrix-based pattern to achieve secured “near lossless” compression on the layouts. We have evaluated CompressKey on 14/28/32 nm open-source ICCAD contest databases and achieved a secured compression ratio of 4.54 with encryption features outperforming$1.22\times $ –$1.59\times $ compared to the state-of-the-art techniques.- 2 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 42 (4) , 1030-1043
CompressKey—Near Lossless Layout Compression and Encryption Using Convolutional Auto-Encoder Model and Expansion-Reduction Pattern Techniques
Malicious manipulation of very large-scale integration physical-layout design is a serious problem in modern integrated circuit design. The physical-layout design database requires a highly compressed secured storage medium. In this article, we propose a secured compressive asymmetrical convolutional auto-encoder (ACAE) machine learning framework, CompressKey, which performs layout compression and encryption simultaneously. It utilizes geometric features to eliminate redundancies in layout patterns. We propose a “Divide and Merge” technique to partition each layer into smaller sizes of unique patterns to address the inconsistency of layout pattern complexity. We also propose “Matrix Expansion” and “Matrix Reduction” techniques on the matrix-based pattern to achieve secured “near lossless” compression on the layouts. We have evaluated CompressKey on 14/28/32 nm open-source ICCAD contest databases and achieved a secured compression ratio of 4.54 with encryption features outperforming$1.22\times $ –$1.59\times $ compared to the state-of-the-art techniques.- 28 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Circuits and Systems
- Vol. 16 (5) , 793-806
A 2.66 µW Clinician-Like Cardiac Arrhythmia Watchdog Based on P-QRS-T for Wearable Applications
Wearable electrocardiogram (ECG) device is an effective tool for managing cardiovascular diseases. This paper presents a low power clinician-like cardiac arrhythmia watchdog (CAW) for wearable ECG devices. The CAW is based on a novel P-QRS-T detection algorithm that makes use of clinical features to identify abnormalities. Implemented in 0.18μm CMOS process, the CAW consumes 2.66µW for 80bpm heart rate at 1.2V supply with an area of 0.578mm2. Verified on QT database, the average sensitivity /positive predictivity for P-wave, QRS complex and T-wave are over 93.39%/88.55%, 99.69%/99.48%, and 97.13%/93.18% respectively, across over 190000 beats. It shows over 99.8% arrhythmia detection accuracy for 43 subjects evaluated on MIT-BIH database.
- 28 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Circuits and Systems
- Vol. 16 (5) , 793-806
A 2.66 µW Clinician-Like Cardiac Arrhythmia Watchdog Based on P-QRS-T for Wearable Applications
Wearable electrocardiogram (ECG) device is an effective tool for managing cardiovascular diseases. This paper presents a low power clinician-like cardiac arrhythmia watchdog (CAW) for wearable ECG devices. The CAW is based on a novel P-QRS-T detection algorithm that makes use of clinical features to identify abnormalities. Implemented in 0.18μm CMOS process, the CAW consumes 2.66µW for 80bpm heart rate at 1.2V supply with an area of 0.578mm2. Verified on QT database, the average sensitivity /positive predictivity for P-wave, QRS complex and T-wave are over 93.39%/88.55%, 99.69%/99.48%, and 97.13%/93.18% respectively, across over 190000 beats. It shows over 99.8% arrhythmia detection accuracy for 43 subjects evaluated on MIT-BIH database.