To accurately capture clinically relevant episodes with body sensor networks (BSNs), multi-sensor fusion is essential for extracting intrinsic physiological and contextual information. Due to the heterogeneous nature of the sensors compounded by the mixture of signals across different sensor channels, this process can be practically difficult. The purpose of this paper is to describe the use of source separation for BSN based on independent component analysis (ICA). We demonstrate how this can be used in practical BSN experiments when the number of sensing channels is limited.