Nowadays, in every sector like mental health, physical health and education, communication plays a major role. The proper communication gives the accurate way of presentation for easy understanding. Proposed system experiments with facial emotion recognition that provides effective communication between system and people. The current system experimented on Cohn kanade (CK+ ) dataset for evaluation of all the seven facial emotion expressions which includes anger, contempt, disgust, fear, happiness, sadness, surprise conditions. Due to rapid growth in artificial intelligence tools, the regular interaction between people's communication is reduced day by day which leads to many mental health conditions which impacts on their performance The usage of CK+ dataset includes variational emotions for effective facial emotion detection. The existing deep learning models proven that this dataset provides multiple variants of emotions for accurate analysis of facial emotions. The proposed system, Light—weight fine—tuned Convolutional neural network (LW-fine-tuned CNN) model experimented with less number of layer for feature detection and produced an accuracy of 97.46 where as traditional feature extraction techniques includes local binary patterns and local directional patters and deep learning models (Alexnet, resnet and Mobile net) obtained an accuracy at the maximum of ninety five percentage. Along with accuracy, the precision, recall measures also limited up to 88% and 87%. To improve the performance, the proposed system fine-tuned the existing VGG-16 model and improved precision and recall from 7–10%. The proposed model addressed the technical challenges for effective emotion detection in the aspects of posing, lighting and expressions. The obtained results effectively identified the facial emotions.