To develop a gait rehabilitation robot for hemiplegic patients, quantitative evaluations of patient ability is needed. Patient's walk phase, which includes time balance of stance and swing legs, is one of the most useful indexes. However, conventional methods measuring walk phase require laborious preparations. In this paper, a novel algorithm estimating walk phase on a treadmill by observing DC motor current is proposed. In comparison of this algorithm and conventional methods, it was verified that the proposed algorithm had as the same accuracy as foot switch. Moreover, the proposed algorithm could estimate stance phase in 0.2 (s) errors between measurements of force plate mostly (4 out of 5 healthy subjects). However, result from the 5th subject showed that the proposed algorithm had erroneously identified stance phase as swing phase when little body weight loaded on leg. This characteristic is often observed in hemiplegic gait. Therefore, the proposed algorithm might need improvement of motor current threshold. However, this algorithm had capable of estimating the time of loading body weight on leg, and thus could be useful as a quantitative evaluation tool.