Action observation (AO) has been recently described as an effective strategy in stroke rehabilitation, since it can shape neural circuit reorganization, promote neural plasticity and motor learning. Very few evidences are available for rehabilitation of patients with Parkinson disease trough AO. The specific aim of this paper is to evaluate the effectiveness of action observation as an adjunctive rehabilitative tool to improve balance, gait and to reduce falls and to investigate the changes in EEG pattern duo to the treatment's effects. Prospective quasi-experimental study, pre- and post-design. Five cognitively intact participants with mild Parkinson's disease and gait disturbance were enrolled in the study. The subjects underwent a rehabilitation programme of action observation for gait and balance recovery for 30 minutes, 3 times a week for 4 weeks. The assessments were performed at the beginning (T0) and at the end of the treatment (T1). The main outcome were: the Hoehn and Yahr scale; Unified Parkinson's Disease Rating Scale (UPDRS) motor section III, Mini Mental State Examination (MMSE), Freezing of gait Questionnaire (FOG-Q), Timed Up&Go Test (TUG), Ten meters walking test (steps and seconds) (10 MWT), Berg Balance Scale (BBS), the 39-items Parkinson's Disease Questionnaire (PDQ-39), Short-Form 12-items Healthy Survey (SF-12) and Functional Independence Measure (FIM). Action observation training was feasible, acceptable, safe, and the participants completed 100% of the prescribed training sessions. The statistical analysis performed by the t -test of the T1 vs. T0 showed statistically significant improvements for UPDRS motor section III P = 0.0082, SF12 MCS P = 0.0007, FOG Q P = 0.0030, PDQ39 P = 0.100 and for P300LD P = 0.0077. No statistically significant improvements for other scales were found. AO is a feasible and safe form of rehabilitative exercise for cognitively intact people with mild PD. Further investigation regarding the long-time effectiveness of robot training is warranted.