A speech recognition method, medium, and system. The method includes detecting an energy change of each frame making up signals including speech and non-speech signals, and identifying a speech segment corresponding to frames that include only speech signals from among the frames based inclusive of the detected energy change.