A differential ranging-based extended Kalman filter algorithm (DREKF) is proposed to address the limitations in positioning accuracy and weak stability of UWB systems caused by various errors in indoor environments. The algorithm improves the communication protocol of the UWB positioning system by introducing differential ranging, thus allowing the error model of each base station to be updated during each positioning instance to reduce ranging errors. The improved extended Kalman filter algorithm is applied to locate the tag and thereby enhance both positioning accuracy and stability. Experimental results show that the DREKF algorithm achieves a horizontal positioning accuracy of 3 cm in line-of-sight conditions and 4 cm in non-line-of-sight conditions. Compared with a traditional positioning algorithm, the proposed algorithm significantly improves positioning accuracy and stability, thus satisfying the requirements for high-precision indoor positioning.