Dynamic monitoring of blood pressure (BP) is beneficial to obtain comprehensive cardiovascular information of patients throughout the day. However, the clinical BP measurement method relies on wearing a bulky cuff, which limits the long-term monitoring and control of BP. In this work, a microcavity assisted graphene pressure sensor (MAGPS) for single-vessel local BP monitoring is designed to replace the cuff. The microcavity structure increases the working range of the sensor by gas pressure buffering. Therefore, the MAGPS achieves a wide linear response of 0–1050 kPa and sensitivity of 15.4 kPa⁻¹. The large working range and the microcavity structure enable the sensor to fully meet the requirements of BP detection at the radial artery. A database of 228 BP data (60-s data fragment detected by MAGPS) and 11,804 pulse waves from 9 healthy subjects and 5 hypertensive subjects is built. Finally, the BP was detected and analyzed automatically by combining MAGPS and a two-stage convolutional neural network algorithm. For the BP detection method at local radial artery, the first stage algorithm first determines whether the subject has hypertension by the pulse wave. Then, the second stage algorithm can diagnose systolic and diastolic BP with the accuracy of 93.5% and 97.8% within a 10 mmHg error, respectively. This work demonstrates a new BP detection method based on single vessel, which greatly promotes the efficiency of BP detection.