本文提出一种跨被试的深度神经网络识别方法,应对运动想象脑电信号的非线性、非平稳特性.该方法首先计算协方差矩阵均值,将不同被试者样本集的协方差对齐至单位矩阵,提升样本的被试间泛化性.然后,将对齐后的样本输入至卷积神经网络中,...本文提出一种跨被试的深度神经网络识别方法,应对运动想象脑电信号的非线性、非平稳特性.该方法首先计算协方差矩阵均值,将不同被试者样本集的协方差对齐至单位矩阵,提升样本的被试间泛化性.然后,将对齐后的样本输入至卷积神经网络中,通过留一被试交叉验证法,构建跨被试的运动想象脑电信号识别方法.在BCI Competition IV dataset 2b公开数据集上进行实验,结果表明,新的方法在该数据集上取得了高的识别性能,且测试场景中的时间复杂度与现有方法相同.展开更多
功能磁共振成像(functional Magnetic Resonance Imaging,fMRI)研究面临的主要挑战之一是不同被试者fMRI数据的异质性。一方面,多被试数据分析对于确定所生成结果跨被试的通用性和有效性至关重要。另一方面,分析多被试者fMRI数据需要在...功能磁共振成像(functional Magnetic Resonance Imaging,fMRI)研究面临的主要挑战之一是不同被试者fMRI数据的异质性。一方面,多被试数据分析对于确定所生成结果跨被试的通用性和有效性至关重要。另一方面,分析多被试者fMRI数据需要在不同被试者的神经活动之间进行准确的解剖和功能校准,以提升最终结果的性能。然而,现有大多数功能校准研究都采用浅层模型来处理多被试者间的复杂关系,这严重束缚了多被试信息的建模能力。为此,提出了一种基于多视图自编码器的功能校准(Multi-view Auto-encoder Functional Alignment,MAFA)方法。具体地,该方法通过重构不同被试者的响应空间来学习节点嵌入,捕获不同被试者之间共享的特征表示,从而创建一个公共的响应空间。此外,通过引入自训练聚类目标,利用高置信度节点作为软标签来监督图聚类过程。在4个数据集上的实验结果表明,相比其他多被试者脑影像功能校准方法,所提方法在解码精度方面取得了最佳效果。展开更多
体压分布是度量座椅静态乘坐舒适性的重要指标。目前通过试验获取真人体压分布的方式不能用在设计早期来预测座椅舒适性。本研究实现了一种能够在设计早期应用的体压分布仿真和座椅舒适度评价方法。提出了标准被试的概念,并根据本研究...体压分布是度量座椅静态乘坐舒适性的重要指标。目前通过试验获取真人体压分布的方式不能用在设计早期来预测座椅舒适性。本研究实现了一种能够在设计早期应用的体压分布仿真和座椅舒适度评价方法。提出了标准被试的概念,并根据本研究特点和要求,面向中国人群,通过抽样得到95百分位身高男子、50百分位身高男女和5百分位身高女子的人体尺寸,进而建立对应的人体几何和生物力学模型,作为体压分布仿真的标准被试。面向汽车座椅体压分布仿真和舒适性评价的需要,研究了准确计算体压分布的座椅、人体方面的因素和体压分布仿真方法。通过小规模实验建立了座椅舒适度区间评价标准和头颈部舒适度评价标准。结合某SUV座椅开发进行了体压分布仿真分析和舒适性评价,并通过实验验证了上述仿真方法和评价标准的有效性和准确性。The body pressure distribution is an important index in evaluating seat static comfort. At present the interface pressure acquiring method of real person testing is unable to be used in the early stage of design for seat comfort evaluation. In this study a method for body pressure distribution simulation and comfort evaluation was realized which can be used in the early stage of design. The standard test-subject concept was proposed. According to the status and requirements of this study, and by sampling the Chinese anthropometry database the anthropometric dimensions of percentile-95 male, percentile-50 male and female, and percentile-5 female individuals were obtained. Based on these data the corresponding geometric and biomechanical models were established, which were used as the standard test-subjects of body pressure simulation. To meet the requirements of body pressure distribution simulation and comfort evaluation of automobile seats, the factors of the seat and human body in body pressure distribution were studied, as well as the methods for accurate body pressure distribution simulation. The interval-evaluation standard of body pressure and the head-neck comfort evaluation standard were established by small-scale experiments. The body pressure distribution simulation and comfort evaluation were implemented in the development of a specific SUV seat, which validated the effectiveness and accuracy of the above simulation methods and evaluation standards by tests.展开更多
基于快速序列视觉呈现(rapid serial visual presentation,RSVP)范式的目标图像检索借助于人脑在看到目标图像时产生的事件相关电位(event-related potentials,ERP)来完成复杂目标图像检索。在应用RSVP范式进行复杂目标图像检索时存在...基于快速序列视觉呈现(rapid serial visual presentation,RSVP)范式的目标图像检索借助于人脑在看到目标图像时产生的事件相关电位(event-related potentials,ERP)来完成复杂目标图像检索。在应用RSVP范式进行复杂目标图像检索时存在跨时段甚至跨被试的问题。对此,本文提出了一种面向跨被试RSVP的多特征低维子空间嵌入的ERP检测方法,首先采用迁移学习方法中的欧式空间对齐对不同被试的数据进行对齐,其次将来自不同空间的特征分别进行有监督降维、重构。最终采用留一被试法作为检验方法、平衡准确率作为评价指标,在PhysioNetRSVP数据集以及清华RSVP数据集下共计14个长度分段中,有12个长度分段达到最优分类结果。结果表明本文提出的多特征低维子空间嵌入方法能够有效提升ERP检测时的稳定性。展开更多
文摘本文提出一种跨被试的深度神经网络识别方法,应对运动想象脑电信号的非线性、非平稳特性.该方法首先计算协方差矩阵均值,将不同被试者样本集的协方差对齐至单位矩阵,提升样本的被试间泛化性.然后,将对齐后的样本输入至卷积神经网络中,通过留一被试交叉验证法,构建跨被试的运动想象脑电信号识别方法.在BCI Competition IV dataset 2b公开数据集上进行实验,结果表明,新的方法在该数据集上取得了高的识别性能,且测试场景中的时间复杂度与现有方法相同.
文摘体压分布是度量座椅静态乘坐舒适性的重要指标。目前通过试验获取真人体压分布的方式不能用在设计早期来预测座椅舒适性。本研究实现了一种能够在设计早期应用的体压分布仿真和座椅舒适度评价方法。提出了标准被试的概念,并根据本研究特点和要求,面向中国人群,通过抽样得到95百分位身高男子、50百分位身高男女和5百分位身高女子的人体尺寸,进而建立对应的人体几何和生物力学模型,作为体压分布仿真的标准被试。面向汽车座椅体压分布仿真和舒适性评价的需要,研究了准确计算体压分布的座椅、人体方面的因素和体压分布仿真方法。通过小规模实验建立了座椅舒适度区间评价标准和头颈部舒适度评价标准。结合某SUV座椅开发进行了体压分布仿真分析和舒适性评价,并通过实验验证了上述仿真方法和评价标准的有效性和准确性。The body pressure distribution is an important index in evaluating seat static comfort. At present the interface pressure acquiring method of real person testing is unable to be used in the early stage of design for seat comfort evaluation. In this study a method for body pressure distribution simulation and comfort evaluation was realized which can be used in the early stage of design. The standard test-subject concept was proposed. According to the status and requirements of this study, and by sampling the Chinese anthropometry database the anthropometric dimensions of percentile-95 male, percentile-50 male and female, and percentile-5 female individuals were obtained. Based on these data the corresponding geometric and biomechanical models were established, which were used as the standard test-subjects of body pressure simulation. To meet the requirements of body pressure distribution simulation and comfort evaluation of automobile seats, the factors of the seat and human body in body pressure distribution were studied, as well as the methods for accurate body pressure distribution simulation. The interval-evaluation standard of body pressure and the head-neck comfort evaluation standard were established by small-scale experiments. The body pressure distribution simulation and comfort evaluation were implemented in the development of a specific SUV seat, which validated the effectiveness and accuracy of the above simulation methods and evaluation standards by tests.
文摘基于快速序列视觉呈现(rapid serial visual presentation,RSVP)范式的目标图像检索借助于人脑在看到目标图像时产生的事件相关电位(event-related potentials,ERP)来完成复杂目标图像检索。在应用RSVP范式进行复杂目标图像检索时存在跨时段甚至跨被试的问题。对此,本文提出了一种面向跨被试RSVP的多特征低维子空间嵌入的ERP检测方法,首先采用迁移学习方法中的欧式空间对齐对不同被试的数据进行对齐,其次将来自不同空间的特征分别进行有监督降维、重构。最终采用留一被试法作为检验方法、平衡准确率作为评价指标,在PhysioNetRSVP数据集以及清华RSVP数据集下共计14个长度分段中,有12个长度分段达到最优分类结果。结果表明本文提出的多特征低维子空间嵌入方法能够有效提升ERP检测时的稳定性。
