Abstract: An adaptive receiver scheme employing the training sequences of flexible length is proposed for DS-UWB systems. By fully exploiting the NLMS algorithms robust behavior on convergence rate, complexity and stability, the proposed scheme is capable of performing an adaptive receiver with its low cost in terms of training bits. Theoretical analysis and simulation results show that, without loss of the BER performance, the proposed scheme can achieve a simpler structure compared with the traditional coherent Rake receiver and a higher throughput compared with the conventional adaptive receiver scheme, and therefore can be considered as a promising candidate for low cost, low power consumption and low complexity indoor UWB application scenarios.