In this study, center of gravity (COG) method is proposed to detect endpoints of isolated words. Common vector approach (CVA) is employed to evaluate the effect of the proposed method in the isolated word recognition. Since the CVA method is sensitive to shifts through the time axis, endpoint detection of words is extremely important. For the comparison purpose, one of the well-known endpoint detection methods is also used together with CVA. The recognition rates obtained by using COG and CVA methods for TI-digit database are greater than those obtained by using other endpoint detection and CVA methods.