In current WiFi-based device-free tracking systems, fine-grained Channel State Information (CSI) has aroused great interest among researchers. However, the CSI obtained from commercial WiFi devices contains large amounts of phase error and cannot be used directly. Therefore, we propose a framework that only utilizes the CSI amplitude to achieve device-free trajectory tracking. In our system, we adopt the Spatial-Temporal-Smoothing Multi-Signal Classification (STS-MUSIC) to jointly estimate the Doppler Frequency Shift (DFS) and Angle of Arrival (AoA) of the moving target. Then, we leverage the cross correlation between different subcarriers to eliminate the parameter ambiguity caused by the conjugate operation, and further realize trajectory tracking. Our implementation on commercial WiFi devices shows that our system achieves an average tracking error of 0.75m using only one pair of transceivers.