Loss of an upper extremity is a traumatic and irreversible event in the life of an affected person. Often amputees are provided with state of the art prostheses to compensate for the functional deficits caused by the loss and cosmetic replacement of the lost arm. Surveys report though, that half of amputees receiving a prosthesis reject it. This is clear indication that current prosthetic arms need further improvement to better meet the needs of their users. To adequately replace a lost arm a prosthesis should be perceived by its user as a part of his or her body and not as an external robotic tool. One aspect, hindering the integration of state of the art prosthesis into the amputee's body image, is the lack of sensory information transferred from the prosthesis to the user. Another aspect is that prostheses are not intuitively controlled and only one joint can be moved at a time. The work presented here addresses both of these aspects in order to improve future prostheses and their control. To focus on improvement of prostheses according to users' needs, their satisfaction and their suggestions for improvement of current prostheses were collected by means of a survey. This survey also covered the topic of sensory feedback by asking participants about the most relevant information they would like to feel with their prosthesis and how they would like them to be transmitted. Results showed that 80% of the 108 participants were satisfied with their prosthesis but, at the same time, 79% were not absolutely satisfied. Most often asked for improvements include appearance and durability of the cosmetic glove, more dexterity and enhanced grasping capabilities of the prosthetic hand, a more comfortable socket that reduces sweating of the stump and reduced weight of the prosthesis. In general, respondents asked for prosthesis with more degrees of freedom while at the same time demanding a more intuitive and reliable control. Sensory feedback was of importance for 88% of respondents, whereas grip force, proprioceptive information about position and movement as well as first and last contact of a grasped object were sensory information requested most often. Vibration, pressure and electrical stimulation were suggested as suitable means for transmission of these information from the prosthesis to the amputee. To address the need for improved control of prostheses identified in the survey, a fully implantable system to measure electromyogram (EMG) was developed conjointly in an international project team. By measuring multiple EMG signals of muscles that are controlled intuitively, the system aims at controlling simultaneous movement of multiple joints of advanced arm prostheses. Moreover, issues connected to the use of surface EMG as control signal should be overcome by application of implanted electrodes. The present work reports the development of implanted electrodes, in vitro and in vivo evaluation of the whole system in rats, sheep and primates, analysis of signals measured during these animal trials and evaluation of algorithms for prosthesis control, all of which are a significant contribution to the overall system. Electrode development underwent multiple iterative steps. A first electrode design based on a polyimide carrier was successively improved, but mechanical in vivo stability was finally achieved by development of a new silicone electrode. During 56 implantations of these electrodes in rats and sheep only one contact of one silicone electrode broke. The implantation procedure developed for these electrodes provided a low invasive way to securely position electrodes at target muscles in rat and sheep experiments. EMG signals measured with the implanted electrodes yielded a considerable increase in signal quality by reduction of artifacts and noise compared to EMG signals measured at the skin surface. High amplitudes of the EMG signal combined with reduced pick-up of external noise resulted in a signal to noise ratio of 39 dB. Analysis of EMG signals measured during reaching movements in primate experiments demonstrated clear distinctness of arm movements into different directions. Signal features and classifiers evaluated during these investigations were able to reliably discriminate between subsets of movements and demonstrated that even few basic features and simple classifiers yield good classification accuracies on these signals. The first EMG signals measured with the whole measurement system when implanted in sheep demonstrated the function of all involved components. Based on the demonstrated reliability and safe usage of the implantable measurement system, demonstrated in animal trials, further development will focus on achieving an evaluation in humans. Additional steps will include integration of control algorithms into the implant electronics as well as incorporation of sensory feedback from the prosthesis to its user.