This paper addresses the problem of online and adaptive gait pattern generation for powered lower-limb exoskeletons (PLLEs), exploiting the motion of sensorized crutches. We conduct a series of experiments with subjects walking with and without crutches to investigate the synergies of walking between upper and lower body segments, by adopting principal component analysis (PCA), We also evaluate the effect of using crutches on the walking synergies, and we demonstrate that upper and lower limb walking synergies undergo slight changes in that case. However, the upper and lower limb synergies remain evident and can be exploited in order to use the motion of crutches as an input to PLLEs to identify a desired motion of the lower limb. We propose a method to use the results of synergy analysis to shape gait parameters in the real-time control of PLLEs. To evaluate the scalability of our approach for real-world applications, we conduct a number of experiments with subjects wearing a PLLE and using sensorized crutches to adaptively change the gait parameters of walking steps, depending on upper body actions.