As the position transducers commonly used in industry do not inherently measure an instantaneous velocity, signal processing is generally required to improve the accuracy of velocity estimation at each sampling instant. This estimated signal is then used as the velocity feedback for the velocity loop control in servo motor drives. In this paper, an adaptive fuzzy logic-based observer is proposed to estimate velocity from the measured motor position. The proposed observer has a structure similar to that of the conventional state observer except that the state feedback is replaced with a fuzzy logic feedback. The observer has two adaptation mechanisms: the first one is used to vary the fuzzy output, and the second one is used to identify the motor parameters. The experimental results show that the noise in the estimated velocity caused by the quantization of the measured position can be reduced dramatically with the proposed observer. In addition, the observer has good transient responses in both normal and in low speeds.