This paper explores the possibility to adopt neural oscillators for pathological tremor attenuation. The objective is to suppress the tremor of a single joint of upper limb via functional electrical stimulation (FES). A biologically inspired neural oscillator is developed, which generates the anti-tremor rhythmic stimulation patterns to stimulate a pair of antagonist muscles. Surface electromyographic (EMG) signal is used to entrain the neural oscillator reciprocally and shape the stimulation pattern adaptively. The neural oscillator serves as an adaptive feedforward controller, which is combined with a feedback regulator. Simulation study is performed on musculoskeletal models of wrist joint and elbow joint separately, and some promising results are presented.