近年来，随着脑机接口（BCI）相关研究的不断推进，BCI 在各个领域得到越来越广泛的应用。BCI 技术可将从大脑皮层采集到的脑电信号转换为可被计算机识别的信号，通过对其提取可用于模式识别的特征信号，并将其转化为具体的指令，用于辅助设备的控制操作。不同的应用方向下对 BCI 技术提出了不同的要求，促使了 BCI 技术的不断革新。清华大学有非侵入式 BCI 技术积累，如睡眠脑电分期方法、单导脑电信号准确性分析、基于 BCI 的目标识别范式和算法等。本文基于非侵入式 BCI 在睡眠监测和目标识别两方向的应用，根据 BCI 信号检测的基本原理，设计对应的应用技术方案，并进行了初步的实验用于效果评估。本论文主要完成了以下内容： 针对睡眠监测开展研究工作。脑电作为睡眠最相关的生理指标，利用 BCI 测得的脑电可以进行睡眠监测。脑电头环虽然是一种睡眠监测的新方式，但其在睡眠监测时存在舒适度不佳问题而影响睡眠监测的效果。本文针对脑电头环在睡眠监测中存在的问题，通过对睡眠监测方法和设备硬件进行改进，平衡监测生理指标数量和被试生理心理负荷程度。通过对比实验发现，设备改进有效提升了舒适度评分，能尽量接近智能穿戴手表舒适度水平，有效提升了被试睡眠质量。基于 BCI 的睡眠监测技术发现，设备改进后被试者的睡眠脑电波形出现特征波形，频谱图分布出现了主要频率特征，睡眠阶段特征更明显，被试者睡眠质量有效提升。 针对行动障碍人群开展基于 BCI 的目标识别恢复人机交互能力的研究。本论文基于稳态视觉诱发（SSVEP）进行目标识别，用任务相关成分分析（TRCA）算法进行SSVEP 目标识别模拟的研究。通过训练数据集创建空间滤波器与计算信号模板，检测发现 1s 时间窗下的识别效果最好，识别平均信息传输速率（ITR）为 108.05 bits/min，信息传输速率较高，同时用户的控制负担较轻。再设计结合增强现实（AR）图像识别的候选目标生成方式，可摆脱预设目标和固定摄像头的限制，实时视野动态生成候选目标，提高了目标选择的多样性。设计脑电采集电极与 AR 眼镜结合的组件，佩戴较舒适轻便。综合的技术方案可应用于多个人群和场景，有一定的潜在应用价值。

In recent years, with the continuous advancement of related research on brain-computer interface (BCI), its unique technical charm has made it more and more widely used in various fields. BCI technology can convert the EEG signals collected from the cerebral cortex into signals that can be recognized by computers, extract characteristic signals that can be used for pattern recognition, and convert them into specific instructions for auxiliary equipment control operations . Different application directions put forward different requirements for BCI technology, which promotes the continuous innovation of BCI technology. Tsinghua University has accumulated non-invasive BCI technology, such as sleep EEG staging method, single-lead EEG accuracy analysis, BCI-based target recognition paradigm and algorithm, etc. Based on the application of non-invasive BCI in sleep monitoring and target recognition, according to the basic principle of BCI signal detection, the corresponding application technical scheme is designed, and preliminary experiments are carried out for effect evaluation. This paper mainly completes the following contents:Research work on sleep monitoring. EEG is the most relevant physiological indicator of sleep, and the EEG measured by BCI can be used for sleep monitoring. Although the EEG headband is a new way of sleep monitoring, it has the problem of poor comfort during sleep monitoring, which affects the effect of sleep monitoring. Aiming at the problems existing in the sleep monitoring of the EEG headband, this paper aims to balance the number of physiological indicators monitored and the degree of physiological and psychological load of the subjects by improving the hardware and sleep monitoring methods of the equipment. Through the comfort comparison experiment, it is found that the improvement of the equipment has effectively improved the comfort score, which can be as close as possible to the comfort level of the smart wearable watch, and effectively improve the sleep quality of the subjects. The BCI-based sleep monitoring technology found that after the equipment was improved, the subjects' sleep EEG waveforms showed characteristic waveforms, the spectrogram distribution showed main frequency characteristics, and the sleep stage characteristics were more obvious, and the subjects' sleep quality was effectively improved. Research on BCI-based target recognition to restore human-computer interaction ability for people with mobility impairments. In this paper, target recognition is based on steady-state visual evoked (SSVEP),and task-related component analysis (TRCA) algorithm is used to simulate the target recognition of SSVEP. Through the training data set to create a spatial filter and calculate a signal template, it is found that the recognition effect under the 1s time window is the best, and the average information transmission rate (ITR)of the recognition is 108.05 bits/min, which is a high information transmission rate, and at the same time the user's control burden lighter. The candidate target generation method combined with augmented reality (AR) image recognition is redesigned， which can get rid of the limitation of preset targets and fixed cameras , and dynamically generate candidate targets in real-time field of view, which improves the diversity of target selection. The components that combine EEG acquisition electrodes and AR glasses are designed to be more comfortable and lightweight to wear. The comprehensive technical solution can be applied to multiple groups of people and scenarios, and has certain potential application value.