Presently Electromyography (EMG) signals are widely utilized for clinical/biomedical applications, such as disease prognosis and advanced human machine interface. EMG signals are picked from muscles by invasive process or from surface of skin called surface EMG. However acquired from any of the technique it requires important aspect is how to extract useful information from the cached signal for understanding and relating the signal with its relative physical and biological aspects. The reason for this paper is to present analyze the behavior of EMG signal under different transform domains such as frequency and wavelet domain and to relate the coefficients of these domains with the physical and biological aspects of signals. Furthermore the authors point out how the unwanted signals such as noise and other interfering signals can be removed using the different transforms. This paper gives specialists a decent understanding of EMG signals and its investigation methods. This learning can be helpful for creating automated systems for prognosis and man machine interface development.