随着国民生活水平的提高，人们越来越关注自己的身体健康，将医疗诊断与现代技术相结合的可穿戴医疗、智慧医疗也受到越来越多的关注和研究。中医脉诊具有实时、无创、全面、个性化等特点，非常适合发展可穿戴智慧医疗。实现可穿戴智慧脉诊存在两个方面的难题，一是脉诊过程的数字化问题，另一个是脉象智能识别问题。本文针对这两个问题，构建了一套可穿戴式脉诊数字化软硬件系统以及一套符合中医脉诊理论的脉象识别算法模型。在脉诊数字化方面，本文以一种柔性压电驻极体压力传感器为脉象感知器件，该传感器具有柔软轻薄、可在高强度静压力下感知微弱动态压力信号、灵敏度高、面积小、可靠性与可重复性高等特点，非常适用于可穿戴式脉象传感。将三个该传感器并排成一个传感器阵列，使其感知人体寸口处寸、关、尺三个部位的脉象，再利用三组小型气泵、气阀配合三个气囊压在对应的脉象传感器之上实现加压，以模拟中医师按压取脉的过程，实现中医脉诊寸口处的“三部九侯”取脉。整套硬件系统的电路部分以STM32芯片为控制中枢，联接传感和加压器件，构成可穿戴脉诊仪。同时，设计和开发了上位机操作软件，用于脉象信号的接收、保存和可穿戴脉诊仪的自定义控制。进一步地，设计了一套包含系统矫正、标准化脉象采集、脉象信号前处理、脉象信号后处理的脉象信号调理框架，用于为脉象识别提供可靠的数字化脉象信号，其中涉及的算法分别写进STM32芯片和上位机软件系统中。在脉象识别方面，基于临床脉诊数据，利用可穿戴脉诊仪样机采集了117例包含迟、数、浮、沉、虚、实脉六种中医脉象类别的数据集，利用该数据集实验并分析了基于机器学习模型的脉象识别方法的局限性。随后，提出了“三压力脉图”的概念，不同于过往脉象识别研究中的“单脉象图”，“三压力脉图”具备更多更丰富的中医脉象信息。在“三压力脉图”的基础上，依据中医脉诊理论构建了一套脉象识别算法模型，在临床数据集上对迟、数、浮、沉、虚、实脉与平脉实现了100\%的识别准确率，并有复合脉识别能力。通过招募志愿者开展人体寸口脉诊实验，对可穿戴脉诊仪软硬件系统功能进行试验，初步验证了系统的有效性。

With the improvement of people‘s living standard, people are paying more and more attention to their physical health. Wearable medicine and smart medicine, which combine medical diagnosis with modern technology, are also attracting more and more attention and research. TCM pulse diagnosis has the characteristics of real-time, non-invasive, comprehensive and personalized, which is very suitable for the development of wearable smart medicine. There are two problems in the realization of wearable smart pulse diagnosis, one is the digitization of pulse diagnosis process, the other is the problem of intelligent pulse recognition. Aiming at these two problems, this paper constructs a set of wearable digital software and hardware system for pulse diagnosis and a set of pulse recognition algorithm model conforming to the theory of traditional Chinese pulse diagnosis. In terms of digitization of pulse diagnosis, this paper takes a flexible piezoelectric electret pressure sensor as a pulse sensing device. The sensor is soft and thin, can sense weak dynamic pressure signal under high static pressure, has high sensitivity, small area, high reliability and repeatability, and is very suitable for wearable pulse sensing. The three sensors are side by side into a sensor array, so that it can sense the pulse of the three parts of the human body wrist Cun, Guan and Chi, and then three groups of small air pump, air valve and three air bags are pressed on the corresponding pulse sensor to achieve pressure, so as to simulate the process of TCM doctor pressing the pulse, to achieve the "three regions and nine divisions" pulse taking at the pulse diagnosis of TCM. The circuit part of the whole hardware system takes STM32 chip as the control center, and connects the sensor and pressure device to form the wearable pulse diagnosis instrument. At the same time, the upper computer software operating system is designed and developed, which is used for receiving and preserving pulse signal and self-defining control of wearable pulse diagnosis instrument. Furthermore, a set of pulse signal conditioning framework including system correction, standardized pulse signal acquisition, pulse signal pre-processing, and pulse signal post-processing is designed to provide reliable digital pulse signal for pulse recognition. The algorithms involved are respectively written into the STM32 chip and the upper computer software system.In pulse recognition, a dataset of 117 cases containing six TCM pulse categories: slow, rapid, floating, deep, deficient, and excessive pulse is collected using a prototype wearable pulse diagnostic device based on clinical pulse data, and the limitations of pulse recognition methods based on machine learning models are experimented and analyzed using this dataset. Subsequently, the concept of "triple pressure pulse map" is proposed, which is different from the "single pulse map" in previous pulse recognition studies, and has more and richer information of TCM pulse. The "triple pressure pulse map" has more and richer TCM pulse information. Based on the "triple pressure pulse map", a pulse recognition algorithm model is built based on the theory of Chinese medicine pulse diagnosis, which achieves an accuracy rate of 100% for slow, rapid, floating, deep, deficient, excessive and normal pulses in the clinical data set, and has the ability to recognize compound pulses. The effectiveness of the system was initially verified by recruiting volunteers to carry out human wrist pulse diagnosis experiments and testing the functions of the wearable pulse diagnostic instrument software and hardware system.