Wireless body area networks have recently gained a lot of interest due to multiple possible applications such as wireless health monitoring or wearable computing. Because of the rather simple hardware realizations and the energy efficiency, ultra wideband (UWB) communication has become one promising technology for the use in wireless body area networks (BAN). After pointing out the motivation of this work and highlighting its contribution, a definition of body area networks is presented as well as a brief introduction on UWB. There, the main promises of UWB communications are presented as well as the principles of some typical receiver structures for UWB. Since UWB communication at the human body is a brand new topic, channel measurements at the human body are performed. The frequency range for these measurements is chosen from 2 to 8 GHz. Based on 1100 channel measurements a channel model for the UWB BAN is derived. Using the Akaike information criterion (AIC) it is shown that the channel decays over the time and that the channel taps are log-normal distributed. The channel at the head is of particular interest as most human communication organs such as mouth, ears, and eyes are located there. Therefore, the ear-to-ear link, which can be regarded as a worst case scenario at the head due to the missing line-of-sight component, is considered to specify the impact of the channel on the system design. When considering the ear-to-ear link it is shown by means of theory, simulations, and measurements that the direct transmission through the head is attenuated so much that it is negligible. Therefore, antennas should be designed in a way that they do not radiate into the body but away from it or along its surface. Moreover, it is shown that the channel is robust against distance variations between the antenna and the skin, and that reflections and absorptions are caused by the body. For the ear-to-ear link the antennas should be placed behind the ears to get the smallest channel attenuations. From the measurements it can also be observed that the main energy of the channel impulse response is contained in a very small time window. Thus, non-coherent receivers with a short integration duration can capture almost the whole energy of the channel. Since UWB systems are a secondary spectrum user, the impact of existing wireless services on UWB is investigated as well. Due to the low transmit power not only the in-band but also the out-of-band interferers are harmful for UWB transmission. Based on frequency-domain and time-domain measurements it is shown that interference not close to the UWB device can be handled by using filters. However, this is not sufficient enough if an interferer is in close vicinity of the UWB device. Therefore, the temporal cognitive medium access is presented to avoid the interference from burst-wise transmitting devices. There, the UWB system listens if the channel is occupied by an interferer and it transmits only in case that no other system is active at the same time. For such a temporal cognitive MAC an expression is given to calculate the optimum UWB packet length. Assuming different interference scenarios, the packet lengths are evaluated. Moreover, it is shown that reasonable usable idle times can be achieved, which the UWB device can use for transmission, and strict latency time requirements can be met. ALOHA, 1-persistent CSMA, and non-persistent CSMA are considered as access schemes for the performance evaluation of the temporal cognitive MAC. For evaluation, two different cases are distinguished, with and without bandpass filter at the UWB receiver. It is shown that a UWB device with bandpass filter that uses the temporal cognitive MAC in conjunction with non-persistent CSMA has low packet error rates below 102 for up to about 15 active UWB links. Due to complexity reasons non-coherent receivers are the most promising solution for the use in UWB devices. Hence, the focus in this thesis lies on the energy detector and the transmitted reference receiver, which have both the same performance. Furthermore, the maximum likelihood receivers in the presence of inter-symbol interference are derived for binary pulse position modulation and transmitted reference pulse amplitude modulation assuming partial channel state information. The maximum likelihood receivers in the presence of a co-channel interference are calculated for the transmitted reference PAM as well. A family of maximum likelihood receivers is also derived for the transmitted reference pulse interval amplitude modulation, which is a combination of pulse position modulation and transmitted reference pulse amplitude modulation. The performance of all these receiver structures is evaluated by means of bit error rate simulations. The simulations are performed by using channels with independent and identically-distributed channel taps and exponential decaying channels as well as by using the BAN channel model. For all these receiver families a trade-off between performance and complexity is observed. Assuming a higher level of channel state information the performance improves while the complexity increases. The receiver structures with knowledge of the average power delay profile are recommended for the use in wireless BAN. These receiver structures exhibit for most channels better performance than the ones without channel state information, however, they require only moderately higher complexity. Furthermore, the receivers with knowledge of the average power delay profiler are less sensitive to the chosen integration duration, since the weighting can be regarded as choosing a variable integration duration. Finally, recommendations for a UWB BAN system are given and conclusions are presented.