The output of the traditional inductive particle sensor is pulse signal of inductance or voltage amplitude, which is scalar signal in nature. The metal particle can be distinguished into magnetic metal or non-magnetic particle by the positive and negative of pulse signal. The particle size can only be estimated under the condition of known particle material. However, for oil containing multiple metal particles, the particle identification method based on the scalar signal may be invalid. To solve this problem, a double lock-in amplifier circuit is utilized to convert the complex particle signal into a pair of DC signals. A signal processing method based on fuzzy membership function is proposed, which can realize the material identification and size estimation of various particles under noise interference. In this paper, an experimental system based on a three-coil sensor is established. The fuzzy membership function is formulated and the system is calibrated by using five kinds of metal particles. Finally, two kinds of particles are selected to evaluate the calibrated system. Results show that the particle material can be identified accurately by this system, and the error of particle size estimation is less than 2% .