登录 EN

添加临时用户

建筑空间形态要素多模态脑认知机制解析研究与设计交互

Multimodal Brain Cognitive Mechanism Analysis and Interaction Design of Architectural Spatial Morphological Elements

作者:曾子欣
  • 学号
    2023******
  • 学位
    硕士
  • 电子邮箱
    zen******.cn
  • 答辩日期
    2025.05.19
  • 导师
    王辉
  • 学科名
    建筑学
  • 页码
    149
  • 保密级别
    公开
  • 培养单位
    000 建筑学院
  • 中文关键词
    建筑空间形态;生成式设计;脑机接口 (BCI);脑电(EEG)
  • 英文关键词
    Architectural spatial morphology;Generative design;Brain-computer interface (BCI);Electroencephalography (EEG)

摘要

本研究探讨建筑空间形态在人脑认知中的生理数据反馈,并基于脑电数据进行自动化的建筑形态生成和效果图像生成,融合了建筑学、计算机科学与脑电(EEG)技术的跨学科方法。已有研究表明,建筑空间的形式与特征会对人产生显著的认知和情绪影响,为从脑机制角度理解建筑设计提供了背景依据。研究采用脑电(EEG)技术和生成式设计相结合的方法,主要进行了两个方面的工作:1)通过实验采集人在认知具有不同特征的建筑空间形态要素时的脑电数据,探究建筑空间形态要素多模态(包括文字和图像两种载体)的认知机制;2)通过数据分析归纳脑电数据对应的要素特征,从而实现基于脑电数据的交互式建筑形态和效果图生成。研究首先确定了建筑空间的形态特征的基本分类,以形状、尺寸、材质、色彩为主要研究要素,通过参数化建模基于不同的建筑空间形态特征构建了数据集,通过实验将被试的主观评价、空间形态特征与脑电(EEG)信号对应起来,建立空间刺激—大脑反应的映射关系。实验结果显示,文字想象相比于图片观看会引起被试更高的活跃度和更强的大脑反应,不同的空间形态特征会引发大脑不同的反应模式,不同被试对于尺度变化、颜色和材质的认知反应相对统一,而在形态组合和材质通透性上有较大的个体差异。通过分析和分类这些不同的脑电特征,能够构建脑电特征与建筑空间形态对应关系的模型,从而辅助基于脑电的生成式设计。基于此,本研究构建了一个机制解析与设计交互的系统框架,将建筑空间形态要素通过数字化方式加以控制与呈现。基于实验分析,以形状、尺寸、材质、色彩四种建筑空间形态要素为基础,本研究选择了体块丰富度、尺度延展度、界面通透度与墙面色彩度四个特征维度,通过参数化设计手段,对建筑空间形态要素进行数字化建模,在此基础上生成不同维度下的空间视觉图像,通过实验进行分析与建模,提取个体在不同设计维度上的脑电数据特征,构建脑电感知与空间变量之间的映射关系模型,并基于形状文法进行参数化设计和建筑形体生成,借助图生图模型完成建筑风格图像的生成与视觉风格的个性化表达,实现从脑电感知获取、空间形态参数设定、智能生成机制构建到设计表达可视化的完整流程,为建筑设计与用户感知之间的交互机制提供了一种新的技术路径与方法论支持。综上,研究主要探索了对建筑空间的脑认知理解,尝试构建了基于脑电数据的交互设计流程,对建筑空间要素的认知和设计的跨领域探索进行了尝试。

This study explores the physiological feedback of human brain cognition in response to architectural spatial forms, and establishes a system that enables automated architectural form generation and visual rendering based on EEG data. It adopts an interdisciplinary approach that integrates architecture, computer science, and electroencephalography (EEG) technology.Previous studies have shown that the form and characteristics of architectural spaces can significantly affect human cognition and emotions. For instance, curved forms are often perceived as more pleasant, and moderately complex spaces tend to elicit more positive aesthetic responses and a higher level of relaxation. These findings, supported by EEG data, offer a neurophysiological foundation for understanding architectural design from a brain mechanism perspective by analyzing EEG features and matching them with cognitive characteristics.This study combines EEG technology with generative design methodologies, focusing on two main aspects: (1) collecting EEG data from participants while they cognitively process architectural spatial features presented in multimodal forms (i.e., text and images), in order to investigate the cognitive mechanisms involved; and (2) analyzing and categorizing EEG data to extract corresponding spatial features, enabling interactive generation of architectural forms and visual renderings based on EEG responses.First, the study classifies key features of architectural spatial form, focusing on shape, scale, material, and color. A dataset is constructed using parametric modeling techniques, where each spatial feature is represented in both textual descriptions and visual imagery. By controlling for other variables, the dataset ensures that variations in individual dimensions—such as shape composition, spatial volume, material texture, or color tone—are isolated. Participants are recruited to visually browse both the text and images of spatial forms. Their EEG activity is recorded as they imagine spaces based on these stimuli, and subjective evaluations are collected via questionnaires. The experiment controls presentation time and requires participants to complete simple cognitive tasks during the viewing process. EEG frequency band power variations—especially those related to cognitive processing (such as changes in alpha, theta, and gamma power)—are analyzed to identify neural responses induced by different spatial attributes. By linking subjective evaluations, spatial features, and EEG signals through computational data analysis, the study establishes a mapping relationship between spatial stimuli and brain responses.Results show that different spatial features evoke distinct neural patterns. Compared with image viewing, text-based spatial imagination triggers more active and stronger brain responses. Among the four spatial features, shape variation induces higher levels of advanced cognitive processing and information handling than scale, color, or material. While participants show relatively consistent neural responses to changes in scale, color, and material, individual differences are more pronounced in response to shape composition and material transparency. Even under the same textual descriptions, significant cognitive differences are observed across individuals. By analyzing and categorizing these EEG features, a model can be constructed that maps EEG patterns to architectural spatial forms, thereby enabling generative design based on brain data for interactive and automated creation.Based on these findings, the study develops a mechanism-based interactive system framework that enables digital control and representation of architectural spatial elements. Grounded in experimental analysis, four design dimensions—massing richness, scale extension, interface transparency, and wall colorfulness—are defined as key parameters. These are incorporated into parametric modeling to visualize spatial variations. Multimodal stimuli are presented to participants repeatedly, and EEG data are collected under specific visual conditions. The data are then processed and used for computational modeling. Through EEG data analysis, individual neural features associated with different design dimensions are extracted and modeled to establish a mapping between brain perception and spatial variables.The system incorporates a rule-based shape grammar for parametric generation of architectural forms, and employs image-to-image models to synthesize architectural styles and visually personalized outputs. Ultimately, it enables intelligent generation of architectural spatial imagery based on EEG-driven inputs.This system realizes a complete pipeline from EEG-based perception, spatial parameter setting, intelligent generative modeling, to visual design expression. It offers a novel technological pathway and methodological framework for exploring the interaction between architectural design and user cognition.In conclusion, this study primarily investigates human cognitive responses to architectural spaces, and explores the possibility of establishing a full workflow for interactive and visual design expression based on EEG data. It provides a cross-disciplinary attempt to bridge the understanding and design of architectural spatial features through neurophysiological insights.