According to many authors, crosstalk can be considered still as an open problem, which determinates difficulties in the estimation of several EMG indexes such as spectral ones. In literature a common strategy to avoid crosstalk is using spatial filters, although with some limitations in their performances. Indeed, their effect depends on many factors (type/dimension of electrodes used, physical properties of the tissue and anatomy of the subject), with the complication that a small detection volume is explored. In this work an innovative strategy with the aim to reduce the effects of crosstalk is evaluated experimentally. The presented algorithm, known with the name optimal spatio-temporal filter, is an adaptive method that include different channels filtering them both in time and space, with the aim to maximize the energy of target muscle EMG discarding that of adjacent muscles.