Facial expression plays an integral part in expressing ones emotions. These expressions can be conveyed in various forms such as happiness, sadness, anger, surprise, fear, disgust, and neutral. So we propose a system which recognizes and classifies these expressions. We have used an image compression method (FMM) to compress the images collected. Then we detect human face in the compressed image using Viola Jones Haar-like object detector. Using the detected face, we extract the facial features that changes with the changing expressions using LBP. Finally we classify the expression using the extraction using k-NN. Presently, FER is applied in a wide variety of environments including robotics, mobile application, digital signs, as a psychiatric tool for verifying the observations made by the psychologist, etc. The existing system use grayscale/ RGB images which consumes a lot of space and requires a lot of computational time. We present a new approach of using compressed images to reduce the space and time required.