<p>Filtered backstepping control (FBC) is widely adopted in flexible joint robots (FJRs) due to its ability to reduce computational complexity and avoid direct differentiation of signals with measurement noises. However, a fundamental limitation of conventional FBC approach lies in the introduction of filtered errors, which typically restrict system stability to a semi-global level. To enable global tracking, this paper proposes a backstepping control framework using barrier function-based filters. These filters eliminate the need for signal differentiation under measurement noises and integrate barrier functions to confine the filtered errors within prescribed bounds. This design prevents the accumulation of filtered errors that could otherwise undermine global stability. A Lyapunov-based analysis rigorously proves the global uniform boundedness of all closed-loop signals. Simulation results demonstrate that the proposed method significantly enhances tracking accuracy while preserving computational efficiency, offering a robust and effective solution for high-performance control of FJRs.</p>