| Abstract [eng] |
Children with autism spectrum disorder often face communication difficulties, therefore it is important to seek solutions that could enhance their communication. The combination of reinforcement learning methods and the Picture Exchange Communication System (hereinafter - PECS) provides a way to individualize communication systems in order to improve children’s social skills according to their individual needs. The aim of this thesis is to propose a reinforcement learning model that would help children with autism spectrum disorder communicate more effectively using PECS cards and emotion recognition. During the research, the „CardFinder+ Sense“ algorithm was developed and investigated, which is capable of learning to suggest the correct PECS card. The obtained results showed that, in certain cases, the integration of emotion data into the algorithm improved the model’s effectiveness. |
