The excellent overall performance and compliant nature of Dielectric Elastomer Actuators (DEAs) make them ideal candidates for artificial muscles. Natural muscle however is much more than just an actuator, it provides position feedback to the brain that is essential for the body to maintain balance and correct posture. If DEAs are to truly earn the moniker of "artificial muscles" they need to be able to reproduce, if not improve on, this functionality. Self-sensing DEAs are the ideal solution to this problem. This paper presents a system by which the capacitance of a DEA can be sensed while it is being actuated and used for feedback control. This system has been strongly influenced by the desire for portability i.e. designed for use in a battery operated microcontroller based system. It is capable of controlling multiple independent DEAs using a single high voltage power supply. These features are important developments for artificial muscle devices where accuracy and low mass are important e.g. a prosthetic hand or force-feedback surgical tools. A numerical model of the electrical behaviour of the DEA that incorporates arbitrary leakage currents and the impact of arbitrary variable capacitance has been created to model a DEA system. A robust capacitive self-sensing method that uses a slew-rate controlled Pulse Width Modulation (PWM) signal and compensates for the effects of leakage current and variable capacitance is presented. The numerical model is then used to compare the performance of this new method with an earlier method previously published by the authors.