Brain-Computer Interfaces have been proposed for stroke rehabilitation, but a potential problem with this technology is the dependence of high-quality brain signals. The aim of this study was to investigate if attempted hand open motions can be detected from the muscle activity instead. Ten stroke patients performed 63 ± 7 attempted movements while three channels of EMG were recorded. Hudgins time-domain features and linear discriminant analysis were used, and 92 ± 3% of the movement activity was correctly classified. The Spearman correlation between the upper limb Fugl-Meyer score and the classification accuracies was 0.58 (P = 0.08). In conclusion, attempted movements from stroke patients can be detected using EMG.