
Recently, artificial intelligence has been integrating vast amounts of information through deep learning to create new information. Couldn`t we use emotion recognition AI to help people with aphasia? The purpose of this study is to analyze what emotions aphasics show when they attempt to speak but are unable to express them adequately. 6 aphasic subjects (Goodglass (1972) fluency assessment “word findings” items on a 7-point scale 1-3: 3 subjects, 4-5: one, 6-7: 2) and 6 healthy elderly subjects. A web camera on a laptop computer was used to view the subject`s speech, and “Kokoro Sensor ver. 1.6.0.0” (CAC corporation) was used to analyze the subject`s emotions. Trends in three groups of aphasic subjects were analyzed based on the results of (1) overview (undetected, less detected, and detected emotional values), (2) valence (positive and negative facial expressions), and (3) seven emotions (anger, contempt, disgust, fear, joy, sadness, and surprise). On the other hand, differences were examined in the summary data of healthy elderly and aphasic subjects. The groups that were able to evoke some and a little evocatively showed more of the seven emotions when they tried to evoke them and tried. The group that could evoke to some extent calmed down when they were able to say what they wanted to say. The group that could evoke a little did not calm down when they could not say what they wanted to say. The group that could hardly evoke any words showed only a little joy and surprise. On the other hand, there was a significant difference between the aphasia group and the normal group in the detection of less emotion (p=0.015) and emotion detection (p=0.002) in the summary. Aphasics were found to have less emotional expression than normal subjects. The ability to evoke emotions was found to lead to positive emotions such as joy and calm emotions. On the other hand, when they had something, they wanted to say and were frustrated because they could not say it, their emotions became more complicated.