基於無線感測網路之多人姿態辨識系統

Abstract

本論文以Zigbee無線感測網路發展了一套多人姿態辨識系統。此系統包括人體姿態估測模組、Zigbee無線網路發展板(CC2420DBK board)及電腦端的多人姿態監測軟體。其中人體姿態估測模組由一個三軸加速度計、一個Zigbee晶片及一個8-bit的微控器所組成。人體姿態估測模組可與CC2420DBK board構成Zigbee無線網路，達到同時觀察多人姿態的效果。CC2420DBK board為Zigbee無線感測網路的接收端，透過RS-232介面將Zigbee無線感測網路中接收到的資訊與機器人溝通。多人姿態監測軟體則可以同時檢視三位使用者的姿態，並對各個使用者的姿態作記錄與統計。本文並提出一套結合時域分析與小波轉換的人體姿態估測演算法，實現在人體姿態估測模組的微控器中。透過此演算法分析三軸的加速度訊號，可辨識跌倒、站、坐、躺、上樓、下樓與走路七種人體姿態。本系統經由五位受測者測試，得到辨識率達88%。

In this thesis, a multi-person pose recognition system has been developed. This system includes a human pose detection module, a CC2420DBK board and a multi-person pose monitoring software module. The human pose detection module consists of a triaxial accelerometer, a Zigbee chip and an 8-bit microcontroller. Several human pose detection modules and the CC2420DBK board form a Zigbee wireless network. The CC2420DBK board works as the receiver of the Zigbee wireless sensor network and communicates with the robot onboard computer through RS-232 link. The multi-person pose monitoring software monitors and records activities of multiple users simultaneously. We propose a pose classification algorithm by combining time-domain analysis and wavelet transform analysis. This algorithm has been implemented in the microcontroller of the human pose estimation module. Through the algorithm, the system can classify seven human poses: falling, standing, sitting, lying, walking, going upstairs and going downstairs. A pose recognition rate of 88% has been demonstrated after testing the system by five different users.