Emotion recognition system is a very common yet challenging area where research is still extending its exploration. When it comes to analyzing the emotions that autistic children express through speech, the issue gets more difficult. This difficulty has led us to develop a sentiment analysis system that would identify abusive speech and be implemented for children with autism. By extracting pertinent speech-based features and using a natural language processing strategy to detect abusive words, the suggested method has used a deep learning approach. Standard speech data was used to train the model, and it was then compared to an existing model.