The purpose of the study is to develop an efficient method to recognize facial expressions more efficiently, especially for different cultures. Humans interact verbally and non-verbally, but facial expressions play a key role in determining verbal communication also. This lot of information through nonverbal communication is not considered in the previous human-computer interaction. A system is needed, which can detect and understand the intents and emotions as stated by social and cultural pointers. In this paper, we have purposed a method to classify face images among six types of expressions. The method consists of three phases; in the first phase, we applied viola jones to crop only the face from the whole image and generate new images. Then we extracted gradient features using a HOG histogram. Lastly, we have classified image features using SVM and produced promising results. The proposed method shows remarkable results among other state-of-the-art methods. It provides 99.97% accuracy on combined cross-cultural databases.