Children with autism have trouble interacting communally and are more likely to isolate individuals. Many autistic children struggle to learn but basically interested in music activities and respond to that. The proposed research work implements emotion detection in children’s with autism spectrum disorders (ASD) using physiological signals, CNN, and music that match their mood. Here we capture real-time images using a webcam, and by using a convolution neural network, the facial emotion can be predicted, and concurrently, a GSR sensor is directly connected to the user for measuring their sweat gland activity. If both the predicted facial emotion and the GSR sensor are happy, then it will play some interesting audio to them using an AT mega microcontroller and mini-player, which will keep them relaxed and help them to learn interactively.