This thesis is concerned with signal processing methods which can be applied to electroencephalogram (EEG) and electromyogram (EMG) signals for identifying interactions between the brain and muscles of human subjects. An experimental framework for assessing methods is developed and neurophysiological data and simulated data are used to illustrate the potential of the proposed methods. The concept of cortico-muscular coherence with time lag (CMCTL) is introduced. A methodology based on CMCTL for discovering temporal relationships between synchronised activities in the brain and muscle is developed. Simulated data are used to demonstrate that under certain conditions the time lag obtained by the method corresponds to the average delay along the involved cortico-muscular conduction pathways. Experimental results show that the method enhances the coherence between cortical and muscle signals, and that time lags which correspond to local maxima of CMCTL provide estimation of delays involved in cortico-muscular coupling. The time delays obtained by the proposed method are more mutually consistent and in a closer agreement with the underlying physiology compared to the delays obtained by some state-of-the-art methods. Two approaches for noise removal based on Wavelet Independent Component Analysis and Sparse Signal Representation are developed. A component selection algorithm is proposed for use in these methods to reconstruct a version of signal which contains relatively higher levels of coherent components with respect to the considered activity. The methods achieve a pronounced enhancement to the cortico-muscular coherence, resulting in up to a three times increase in CMC levels for physiological data.