高血压是一种常见的且容易被人们忽视的慢性疾病，是导致心血管疾病的主要危险因素之一。目前常用的袖带式电子血压计只能反映当前时刻的血压值，无法体现血压的波动性。持续的血压监测可以更全面地评估患者的血压状态，及时发现异常波动，为病情判断和医疗干预提供更准确的依据。目前的血压监测手段依赖于动态血压监测设备，然而该设备在使用过程中具有不便性，且会给患者带来不适。 本研究提出了一种新型的连续血压监测设备及其相关算法，旨在通过使用无袖带的方式准确、可靠、便捷、连续的实现血压监测。本研究的算法内容分为无袖带血压估计算法和袖带改进算法两部分。 无袖带血压估计算法给出了一套完整的数据预处理流程与深度学习算法流程。其核心在于采用了脉搏波与心电图两种易于采集的生理信号作为模型输入，提出了一种基于Transformer框架的深度学习模型TABP，可以实现连续的血压波形估计。在模型的训练过程中，针对波形估计这一任务，有针对性的设计了损失函数，调整了网络结构。相比于单一血压值的估计，完整的波形估计不仅包含了每个心动周期的血压信息情况，还反映出了心肌收缩力、外周血管阻力、心脏瓣膜等身体状况。本研究基于MIMIC III数据集展开实验，通过一系列的消融实验确定了模型超参数、输入数据长度以及输入通道的组合方案。最优模型在MIMIC III数据集上得到的波形估计误差为1.801mmHg，估计的收缩压和舒张压的平均误差±标准差分别为1.19±2.13mmHg和0.11±1.61mmHg，相关系数分别为0.994和0.987，均能通过BHS标准和AAMI标准的检验。 新型血压监测装置通过无袖带的方式实现血压监测，但使用时仍需要袖带测量方式进行校准。本研究提出了基于幅度系数法与比值特征法的改进式袖带电子血压计算法；提出了动态充气阈值的判断算法，在不影响检测准确度的情况下，控制袖带的最大充气阈值，不仅可以有效地提升血压测量速度，还可以减轻袖带测量过程对患者造成的不适感。 本研究涉及的算法已经进行了相应的部署，目前已有产品原型机，设备计划量产。新型的连续血压监测设备可以解决目前动态血压监测设备的诸多问题，为血压监测提供新的解决方案。

Hypertension is a common and easily overlooked chronic disease and is one of the major risk factors for cardiovascular disease. Currently, commonly used electronic sphygmomanometers can only reflect the current moment's blood pressure's value, which cannot reflect the fluctuation of blood pressure. Continuous blood pressure monitoring can more comprehensively assess the patient's blood pressure status, detect abnormal fluctuations in time, and provide a more accurate basis for condition judgment and medical intervention. The current means of blood pressure monitoring relies on ambulatory blood pressure monitoring equipment, which is inconvenient to use and can cause discomfort to patients.In this thesis, a novel continuous blood pressure monitoring device and its related algorithm are proposed, aiming to monitor blood pressure accurately, reliably, conveniently, and continuously by using a cuffless approach. The algorithmic content of this study is divided into two parts: the cuffless blood pressure estimation algorithm and the cuff improvement algorithm.The cuffless blood pressure estimation algorithm gives a data preprocessing process and deep learning algorithm process. The core lies in the adoption of two easy-to-acquire physiological signals, pulse wave and electrocardiogram, as model inputs, and a deep learning model TABP based on the self-attention mechanism is proposed, which can realize continuous blood pressure waveform estimation. During the training process of the model, the loss function is purposefully designed and the network structure is adjusted for the task of waveform estimation. Compared with the estimation of a single blood pressure value, the complete waveform estimation not only contains the blood pressure information situation of each cardiac cycle, but also reflects the myocardial contractility, peripheral vascular resistance, heart valves and other physical conditions. In this study, experiments were conducted based on the MIMIC III dataset, and a series of ablation experiments were conducted to determine the combination scheme of model hyperparameters, input data length, and input channels.The optimal model yielded a waveform estimation error of 1.801 mmHg on the MIMIC III dataset, and the mean error$\pm$standard deviation of the estimated systolic and diastolic blood pressures were 1.19$\pm $2.13 mmHg and 0.11$\pm $1.61 mmHg, which were able to pass the tests of the BHS criterion and the AAMI criterion, respectively.The new blood pressure monitoring device realizes blood pressure monitoring by means of a cuffless method, but it still needs a cuff measurement method for calibration when used. In this study, we propose an improved electronic sphygmomanometer algorithm based on the magnitude coefficient method and the ratio feature method; we propose a judgment algorithm for the dynamic inflation threshold, which controls the maximum inflation threshold of the cuff without affecting the detection accuracy, which can not only effectively improve the speed of the blood pressure measurement, but also alleviate the discomfort that the cuff measurement process causes to the patient.The algorithms involved in this thesis have been deployed accordingly, and a product prototype is currently available, and the device is planned for mass production. The new continuous blood pressure monitoring device can solve many problems of the current ambulatory blood pressure monitoring devices and provide a new solution for blood pressure monitoring.