Research in speech recognition is expanding tremendously since the last decade. This is due to the needs for humans to interact with machines more naturally and efficiently than those with the current existing input systems (eg. keyboard, mouse and other pointing devices). Many research on speech regnition for several languages such as English and Japanese have been carried out but very few have beeen on Malay. This study attempts to rduce this gap. The study in this thesis is mainly concerned with finding and experimenting methods for implementing a user-friendly and flexible speaker-dependent Malay speech recognition systems. The system is aimed to be used as a man-machine interface for a simple control system for the handicap. A matching process utilizing the "Dynamic Time Warping" algorithm was chosen to be carried out in two stages of the speech recognition. Log area ration parameters were found to be accurately suitable to represent a Bahasa Malaysia speech frame. A "Most Popular Candidate" decision logic was introduced in the first stage of the matching process, while a :Nearest Neighbour" decision logicc was found to be suitable to be used in the second stage. Statistical evaluations on a ptotoype system utilizing the recognition method mentioned above showed that a 96.7% recognition accurancy could be achieved in a normal quiet room.