Extracting the activity of a particular neural fiber from the extracellular recording of a peripheral nerve is quite important from different clinical perspectives. While traditional neural recording methods are unable to provide such granularity, new signal recording and processing techniques have offered promising solutions recently. A multi-electrode cuff in conjunction with a delay and sum beamforming structure has been used to detect the activity of different fibers in a nerve based on the propagation velocity of action potentials. However, as it is shown in this paper, simple delay and sum beamforming method encounters severe selectivity problem and signal distortion especially at higher velocities. In addition to scrutinizing the performance of the delay and sum beamformer and its inherent problems, here we propose a new beamforming method, based on broadband sensor array signal processing techniques, which exhibits much better selectivity with uniform frequency-velocity response. Our simulation results show how the new method can faithfully extract individual neural fiber activities and explore potential applications which could emerge from using such technique.