The In the recent years, some studies have focused on powered lower-limb prostheses to enable normal walking gait. However, most proposed prostheses use manual switch to change the locomotion mode, for example from walking to sitting or vice-versa. Intelligent prostheses use micro-processing control for automatic switching by utilizing the advances of signal processing and pattern recognition techniques, where the user's intent could be recognized by analyzing the EMG signals sensed from the lower limb. After the recognition phase, the prosthesis controller controls different prosthesis components, to mimic the natural leg. Introduced in this paper is a comparative study and survey on the recent work on EMG signal analysis is for the purpose of classification and recognition. The paper presents details on the recent work done on each stage of EMG analysis, starting by preprocessing, de-noising, and segmentation through feature extraction till classification. It also presents recent work done using deep learning. The results achieved by the different research groups are summarized at the end of the paper