Without speech, we cannot communicate with people. While interacting with others, we can understand their emotions and feelings. Nevertheless, it is not right place to connect human with machine. In this paper, we are trying to do is make the machine to comprehend the emotions of human while interacting with them. Here, the resolution is based on the common scenarios. To delving the feelings and emotions from human voice is new and it is a challenging process. The main problem is to find the exact feeling concession from the speech dataset, for this reason, various proofs are recognized after discourse and apt decision with proper arrangement in models. The important difficulty is to get the perfect data for emotion extraction, and it is the main era in artificial intelligence. To overcome that here, we have implemented convolution neural network (CNN) models on datasets such as TESS, CREMA-D, and RAVDNESS by appending speckle noises to the particular datasets for emotion detection. © 2024 Elsevier B.V., All rights reserved.