Muscle activity detection is important for clinical investigations leading to the identification of neuromuscular disorders. Myoelectric signal recorded via electrodes placed at skin surface can reveal important muscle excitation information about underlying limb movement. However, a primary difficulty in the detection of muscle activity period from myoelectric signals lies in the inherent variability of these signals and the noise added during the collection process. In the literature, the double threshold detector has been commonly used for detection of the muscle activity periods from myoelectric signals. In this study, we propose a new scheme based on the log-likelihood ratio test to detect muscle activity periods accurately. This scheme uses energy information contained in the myoelectric signal, which increases with the start of the activity. We demonstrate the viability of energy detection scheme via successful detection performed on synthetic as well as clinical myoelectric signals.