Space-time block coding represents a practical and pragmatic method to mitigate the fading effects and to increase the channel capacity, by using spatial diversity. The space-time coding techniques are still preferred in communication systems because of the very simple transmit coding rule and of the relative reduced complexity of the maximum likelihood (ML) receivers with respect to other coding methods (e.g. space-time trellis codes) or other MIMO (Multiple Input Multiple Output) algorithms (e.g., spatial multiplexing). One of the drawbacks when using space-time block codes is the increase in complexity at the ML receiver when increasing the number of transmit antennas and the number of constellation points. This paper deals with this issue and introduces a decoding algorithm applicable only for orthogonal space-time block codes. We show that this decoding algorithm reduces the search space dimension of a regular ML receiver, leading to significant complexity reduction for QAM (Quadrature Amplitude Modulation) constellations.