Ultra-wideband (UWB) wireless communication technologies have been proposed to support high data rate multimedia services in office or residential environments. Due to the low transmission power of UWB, the shadowing effect by moving people can considerably reduce the received signal quality and thus significantly degrade the quality of service (QoS) of on-going transmissions. An open issue is to build a simple model which captures the temporal variation of UWB channels and the packet error rate (PER) due to the people shadowing effect (PSE), which will be a useful tool for upper layer protocol performance analysis and simulation. This paper presents an analytical study of the PSE and the temporal variation of UWB channels induced by the motion of a person. First, we derive the angular power spectral density (APSD) of the indoor UWB channel impulse response (CIR), and the PSE in terms of signal power attenuation. Second, based on a two-dimensional random walk mobility model, the PER variation due to people shadowing is modeled as a finite-state Markov chain (FSMC). The investigation of APSD provides important insights on the spatial propagation characteristics of UWB signals. The proposed packet-level channel model can be conveniently incorporated into analytical frameworks and simulation tools for evaluating upper-layer protocols of UWB networks.