睡眠障碍受到越来越多的关注,且自动化睡眠分期的准确性、泛化性受到了越来越多的挑战。然而,公开的睡眠数据十分有限,睡眠分期任务实际上更近似于一种小样本场景;同时由于睡眠特征的个体差异普遍存在,现有的机器学习模型很难保证准确...睡眠障碍受到越来越多的关注,且自动化睡眠分期的准确性、泛化性受到了越来越多的挑战。然而,公开的睡眠数据十分有限,睡眠分期任务实际上更近似于一种小样本场景;同时由于睡眠特征的个体差异普遍存在,现有的机器学习模型很难保证准确判读未参与训练的新受试者的数据。为了实现对新受试者睡眠数据的精准分期,现有研究通常需要额外采集、标注新受试者的大量数据,并对模型进行个性化微调。基于此,借鉴迁移学习中基于缩放-偏移的权重迁移思想,提出一种元迁移睡眠分期模型MTSL(Meta Transfer Sleep Learner),设计了一种新的元迁移学习框架:训练阶段包括预训练与元迁移训练两步,其中元迁移训练时使用大量的元任务进行训练;而在测试阶段仅使用极少的新受试者数据进行微调,模型就能轻松适应新受试者的特征分布,大幅减少对新受试者进行准确睡眠分期的成本。在两个公开的睡眠数据集上的实验结果表明,MTSL模型在单数据集、跨数据集两种条件下都能取得更高的准确率和F1分数,这表明MTSL更适合小样本场景下的睡眠分期任务。展开更多
Autoregressive （AR） power spectral density estimate method was used to analyze the electroencephalogram （EEG） signals in eyes-open and eyes-closed states. From the topographical distributions of delta, theta, alph...Autoregressive （AR） power spectral density estimate method was used to analyze the electroencephalogram （EEG） signals in eyes-open and eyes-closed states. From the topographical distributions of delta, theta, alpha, and beta power spectrum, these two states can be clearly discriminated. In these two states, frontal areas were activated in delta power, both frontal and occipital areas were activated in theta band, and occipital areas were activated in alpha and beta bands. These four bands had significantly higher power in frontal, parietal, and occipital areas when eyes were close. The results also implied that the optimum order of AR model could be more suitable for estimating EEG power spectrum of different states.展开更多
Most biological tissues are supple and elastic,while current electronic devices fabricated by semiconductors and metals are usually stiff and brittle. As a result,implanted electronic devices can irritate and damage s...Most biological tissues are supple and elastic,while current electronic devices fabricated by semiconductors and metals are usually stiff and brittle. As a result,implanted electronic devices can irritate and damage surrounding tissues,causing immune reaction and scarring. In this work,we develop stretchable microelectrode arrays,with the development of a novel soft lithography technology,which are designed and fabricated with a polymer/stretchable metal/polymer sandwich structure. With the great deformability of stretch,compression,bend and twisting,while preserving electrical property,this technology overcomes the fundamental mismatch of mechanical properties between biological tissues and electronic devices,and provides highly-compliant,conformal and stretchable bio-electronic interfaces. Here we also describe the following three applications of the stretchable electrode arrays:a. monitoring intracranial electroencephalography (EEG);b. stimulating peripheral nerves to drive muscles;c. monitoring epicardial electrocardiography (ECG). Stretchable microelectrode arrays create a promising field in biomedical applications for its better modulus match with biological tissues and robust mechanical and electrical properties. They allow for construction of electronic integrated circuits spread over on complex and dynamic curved surfaces,providing a much friendlier bio-electronic interface for diagnosis,treatment and intelligent bio-control.展开更多
针对脑电检测的研究需要,研发了微弱EEG脑电信号采集专用芯片系统。该芯片使用斩波稳定去噪声技术,首先利用2 k Hz的斩波频率对100 Hz以下的EEG信号进行频域隔离,然后利用RRL纹波抑制环路反馈进行调制后位于chopper频率处的主要由失调...针对脑电检测的研究需要,研发了微弱EEG脑电信号采集专用芯片系统。该芯片使用斩波稳定去噪声技术,首先利用2 k Hz的斩波频率对100 Hz以下的EEG信号进行频域隔离,然后利用RRL纹波抑制环路反馈进行调制后位于chopper频率处的主要由失调与低频1/f闪烁噪声引起的纹波电压的抑制;单级斩波放大使用电流复用、亚阈值跨导增强技术对来自EEG传感器的低频（〈100 Hz）小信号（5~100μV）进行40 d B增益的稳定中频放大;级联S/H电路进行去累积毛刺滤波,配合前面斩波技术,达到整体低噪声性能;VGA/LPF通过改变输入、反馈/负载电容,分别进行增益/带宽的数字调整。EEG-DSP加速芯片实现对多通道采集的控制以及信号处理编码。设计使用SMICRF 180 nm混合工艺,使用Cadence的Spectre软件进行功能前/后仿真,使用Caliber工具进行DRC/LVS的版图验收。最后,对设计芯片进行实际测试,结果表明放大芯片关键性能为：8.1μW/通道、面积6.3 mm2/8通道、0.8μVrms（BW=100 Hz）等效输入噪声;该款自主研发的脑电斩波放大芯片性能达到国内外前列的水平,可进行正确的脑电EEG采集,可应用于可穿戴领域、对后续脑电数据分析具有重要的使用价值。展开更多
脑电信号获取过程中,工频噪声干扰现象往往会使所获取的信息产生多种多形态瞬时结构波形,这种现象影响到DIVA(Directions Into Velocities of Articulators)模型对语音的正常处理.为此,本文提出了一种面向特征提取的脑电信号结构自适应...脑电信号获取过程中,工频噪声干扰现象往往会使所获取的信息产生多种多形态瞬时结构波形,这种现象影响到DIVA(Directions Into Velocities of Articulators)模型对语音的正常处理.为此,本文提出了一种面向特征提取的脑电信号结构自适应稀疏分解模型,并在此基础上,通过采用匹配追踪算法求解最佳原子、使用过完备原子库中原子表示原始脑电信号等方法,实现了信号去噪的目的,效果好于传统的小波变换去噪方法.仿真实验表明,本文提出的方法提高了DIVA模型语音发音的精度.展开更多
文摘睡眠障碍受到越来越多的关注,且自动化睡眠分期的准确性、泛化性受到了越来越多的挑战。然而,公开的睡眠数据十分有限,睡眠分期任务实际上更近似于一种小样本场景;同时由于睡眠特征的个体差异普遍存在,现有的机器学习模型很难保证准确判读未参与训练的新受试者的数据。为了实现对新受试者睡眠数据的精准分期,现有研究通常需要额外采集、标注新受试者的大量数据,并对模型进行个性化微调。基于此,借鉴迁移学习中基于缩放-偏移的权重迁移思想,提出一种元迁移睡眠分期模型MTSL(Meta Transfer Sleep Learner),设计了一种新的元迁移学习框架:训练阶段包括预训练与元迁移训练两步,其中元迁移训练时使用大量的元任务进行训练;而在测试阶段仅使用极少的新受试者数据进行微调,模型就能轻松适应新受试者的特征分布,大幅减少对新受试者进行准确睡眠分期的成本。在两个公开的睡眠数据集上的实验结果表明,MTSL模型在单数据集、跨数据集两种条件下都能取得更高的准确率和F1分数,这表明MTSL更适合小样本场景下的睡眠分期任务。
文摘Autoregressive （AR） power spectral density estimate method was used to analyze the electroencephalogram （EEG） signals in eyes-open and eyes-closed states. From the topographical distributions of delta, theta, alpha, and beta power spectrum, these two states can be clearly discriminated. In these two states, frontal areas were activated in delta power, both frontal and occipital areas were activated in theta band, and occipital areas were activated in alpha and beta bands. These four bands had significantly higher power in frontal, parietal, and occipital areas when eyes were close. The results also implied that the optimum order of AR model could be more suitable for estimating EEG power spectrum of different states.
基金National Natural Science Foundation of China (No. 61102042)Youth Innovation Foundation of Chinese Academy of SciencesShenzhen"Peacock Plan"to Z.Y.
文摘Most biological tissues are supple and elastic,while current electronic devices fabricated by semiconductors and metals are usually stiff and brittle. As a result,implanted electronic devices can irritate and damage surrounding tissues,causing immune reaction and scarring. In this work,we develop stretchable microelectrode arrays,with the development of a novel soft lithography technology,which are designed and fabricated with a polymer/stretchable metal/polymer sandwich structure. With the great deformability of stretch,compression,bend and twisting,while preserving electrical property,this technology overcomes the fundamental mismatch of mechanical properties between biological tissues and electronic devices,and provides highly-compliant,conformal and stretchable bio-electronic interfaces. Here we also describe the following three applications of the stretchable electrode arrays:a. monitoring intracranial electroencephalography (EEG);b. stimulating peripheral nerves to drive muscles;c. monitoring epicardial electrocardiography (ECG). Stretchable microelectrode arrays create a promising field in biomedical applications for its better modulus match with biological tissues and robust mechanical and electrical properties. They allow for construction of electronic integrated circuits spread over on complex and dynamic curved surfaces,providing a much friendlier bio-electronic interface for diagnosis,treatment and intelligent bio-control.
文摘脑电信号获取过程中,工频噪声干扰现象往往会使所获取的信息产生多种多形态瞬时结构波形,这种现象影响到DIVA(Directions Into Velocities of Articulators)模型对语音的正常处理.为此,本文提出了一种面向特征提取的脑电信号结构自适应稀疏分解模型,并在此基础上,通过采用匹配追踪算法求解最佳原子、使用过完备原子库中原子表示原始脑电信号等方法,实现了信号去噪的目的,效果好于传统的小波变换去噪方法.仿真实验表明,本文提出的方法提高了DIVA模型语音发音的精度.
