The present research focuses on the facial expression of different peoples by using Convolution Neural Network, CNN, Communication is basic need of human. Many existing researches show that fifty to ninety percent of communication is nonverbal. In future the advanced generation of computing like as pervasive computing and artificial intelligence required central multicultural human communication system close to the nature of human. In human computer interaction (HCI) nonverbal communication has too much importance. There are many methods and techniques that are used to recognize the facial expressions and some genetic facial expressions are found in every person and all the different facial expression have their own meanings and people from different culture have their own way to express. These expressions are sadness, fear, happiness, surprise, anger and neutral. So, it is very difficult to recognize different facial expression of the peoples belongs to different culture so in this research work we proposed a technique to recognize facial expression-based CNN. For this we use two different datasets one is JAFFE that is stand for (Japanese facial expressions) and second is TFEID that is stand for (Taiwanese facial expression image database) To recognize facial expression we use CNN and extract common facial features and on basis of similarity measure get unique and common feature then remove common feature to minimize use of memory and use only unique feature to measure the similarity, we detect facial expression accurately with unique expression features. Our proposed methodology identifies the expressions with less memory usage, less time complexity and more accuracy than the existing techniques. By using this technique, we can solve real life problems with better way and detect cross cultural facial expression successfully.