Myoelectric signals or muscle signals provide an intuitive and rapid interface for controlling technical devices, in particular bionic arm prostheses. However, inferring the intended movement from a surface myoelectric recording is a non-trivial pattern recognition task, especially if myoelectric data stems from low-cost sensors. At the same time, overly complex models are prohibited by strict speed, data parsimonity and robustness requirements. As a compromise between high accuracy and strict requirements we propose to apply Echo State Networks (ESNs), which can be seen as an extension of standard linear regression with 1) a memory and 2) nonlinearity. We find that both features, memory and nonlinearity, independently as well as in conjunction, improve the prediction accuracy on simultaneous movements in two degrees of freedom (hand opening/closing as well as pronation/supination) recorded from four able-bodied participants using a low-cost myoelectric sensor. However, we also find that the model is still not sufficiently resistant to external disturbances such as electrode shift.