This paper implements a personified card memory game playing support system. Personification is defined as follows: (1) Emotion is expressed through a personified face according to situations, (2) a hint is given to a human player through personified facial expression, and (3) the support system sometimes makes an error. The support system consists of three parts; (a) a card memory part, (b) an emotion inference part, and (c) a face expression part. The support system does not necessarily memorize a card position and a card number and makes an error about the card position according to the error rate estimated by the card memory part. The fuzzy inference technique is used in the estimation of the error rate. The emotion inference part estimates the degree of emotions, i.e., happiness, sadness, anger, disgust, surprise and fear, according to the following situations: (i) The support system gives a hint to a human player, (ii) a human player opens a card, (iii) a game trial of an opponent is ended, where the enemy player is considered as a mimic competitor, and (iv) a game is over. Situations are represented by fuzzy sets and the fuzzy inference technique is also used in the emotion inference part. The face expression part expresses a face according to the estimated degree of emotion. This part uses neural network models which are learned by the use of questionnaire data about the relationship between the degree of emotion and the position of each feature in a face.
Simulation experiments of a card memory game are performed by 14 subjects using the presented personified support system. Questionnaire results show the usefulness of personification: Subjects ask a hint easily by the use of facial expressions. In fact, the number of hints to be asked becomes larger in personification than in non-personification. Some problems, however, are pointed out: It is sometimes hard to understand facial expressions. This needs further investigation.
View full abstract