For massive multiple input multiple output (MIMO) system, the minimum mean square error (MMSE) linear detection algorithm can achieve nearoptimal performance. However, the MMSE linear detection involves complicated matrix inverse, which limits the application of the MMSE detection algorithm. To reduce the operation complication, a modified MMSE algorithm is proposed which uses a structuresimple BarzilaiBorwein (BB) iterative algorithm to avoid the matrix inverse. The BB iterative technology is introduced in massive MIMO signal detection, and the initial iteration value is also optimized based on the characteristic of channel hardening to further quicken the convergence of the iterative process. The theoretical and simulated results show that the proposed BB iterative detection algorithm performs better than the recently proposed Neumann series approximation algorithm, while the computational complexity is reduced by about one order compared to the Neumann series approximation algorithm whose truncated order i equals 3. Moreover, the convergence rate of the proposed algorithm is fast. Under a given initial value, it can achieve performance very close to the MMSE algorithm with just several iterations.