Statistical Research on Human Posture Characteristics Based on RFID

In recent years, human posture recognition has grown in various fields, such as medical pathology examinations and elderly status monitoring. This study focuses on Radio Frequency Identification (RFID) statistics of human posture features. It constructs a controllable simplified human arm model using the ESP32 chip a mechanical arm. Antenna tags are positioned at specific locations (such as the back of the hand or the arm) to collect Received Signal Strength Indicator (RSSI) data, and the collected data serves as the ideal posture dataset. This paper analyzes statistical features under three different arm postures: raising the forearm, lateral elevation and waving, utilizing filtering and feature extraction techniques. Following MATLAB data processing, the probability density functions of actual human arm postures and simplified human arm model postures are compared using Euclidean distance and modified Sigmoid functions. The results indicate that the similarities between raising the forearm, front elevation, lateral elevation, waving, and drawing a triangle are approximately 95%, 94%, and 92%, respectively. The average similarity is 93%. An experimental environment was developed in this study. It has proven the effectiveness of the sampling process and feature extraction methods, providing a valuable reference for human posture recognition and other related fields.