Intelligent Method for Collecting Vital Signals in Versatile Distributed e-Home Healthcare by GUO Ran Thesis Supervisor: Prof. VAI Mang I Master of Science in Electrical and Computer Engineering Over the recent dozen of years, the current healthcare system is facing a severe challenge from sharp increasing aging population. Consequently, this situation brings irreversible increase that the amount of chronic disease is quite widespread. Furthermore, cardiovascular diseases (CVDs) have been classified as one of the main pathological threats to the health of human beings all over the world. There are kinds of vital signals could be utilized for detecting and diagnosing CVDs including electrocardiography (ECG), resonance imaging (MRI), heart sound (HS), echocardiography and sphygmogram (SPG). Compared with other vital signals and considering the aspects of cost, non-invasive, easy acquisition and large signal to noise rate, SPG is convenient, competitive and suitable for large scale usage in e-home healthcare system. Generally, the piezoelectric film pulse sensor is commonly utilized to detect and collect SPG signals. Unfortunately, due to this kind of sensor is deployed on user’s wrist firmly and stably, the acquisitioned data is easily influenced by user’s behavior and hard to ensure the quality of signals. Since the artifact motion will introduce much man-made noise into collected signals and affect further analysis or diagnosis, such noise need to be recognized and removed from normal SPG data. Meanwhile, for e-Health system, a continuous monitor requires a continuous sampling and data transmission. Many low-power transmission methods have been invented and implemented in different areas, such as Bluetooth Low Energy (BLE), near field communication (NFC) and so on. However, low-power technologies are usually designed for tiny data and not suitable for such large vital data all the time. Hence, the bottleneck problems still exist in two aspects. The first one is how to ensure the quality of SPG signal and reduce the human error effect for subsequent analysis on embedded platform. The second one is how to achieve a tradeoff between low-power technology and large vital signal. In another word, the latter could be considered as how to achieve a balance to let limited-resource hardware work compatibly with huge-data-required software. In this thesis, one close-loop, based on BLE technology and intelligent transmission system is proposed. Taking advantage of physiological characterization vector (PCV) method and reverse order Pearson correlation (ROPC) method, this system could detect and distinguish artificial, arrhythmia and normal data efficiently and adjust the transmission speed accordingly. There are 58 subjects including 11 normal people and 47 patients who joined this research activity. In total, 179 records of SPG signal with 1828 cycles were collected from collaborative hospital and our research group. For filtered result and performance evaluation, the true positive rate, false positive rate and accuracy of ROPC were 96.17%, 2.40% and 97.37%, while the same benchmarks of PCV could reach 86.47%, 0.84% and 96.88%, respectively. Moreover, for transferring the same amount of SPG data (PCV mode), the power could be saved as 62.5%, 65.7% and 87.7% compared with RS232-USB, classic Bluetooth and Wi-Fi module.