The aim of this study was to develop a task set based on personalized material for nostalgic experience, which could detect cognitive ability via a virtual experience system combined with Kinect somatosensory interactive operation applications without the user wearing any accessory input device. Fifty-nine subjects participated in the experiment. The receiver operating characteristic curve of the game system was statistically analyzed for determining the best cutoff-point in the cognitive function assessment. Correlation analysis and regression analysis were used to explore the correlations between the results and the clinical cognitive assessment scales. According to the MoCA scores, the results showed that the accuracy of the system was 86.4% in evaluating mild cognitive impairment. The system seems feasible and was strongly correlated with clinical cognitive assessment scales. We anticipate that daily use of our system could keep track of changes of cognitive function of the elderly in home life.