This paper suggests the novel algorithm for the estimating gait parameters of the hemiplegic patients using a 3-axis accelerometer. The signal processing for algorithm consists of a bandpass filter and a least square acceleration filter. To evaluate the performance of the algorithm, the correlation coefficients of the stride and the step time between the 3-axis accelerometer and the Vicon motion analysis system are compared. In consequence, correlation coefficient ranged from 0.90 to 0.99 for patients and ranged from 0.92 to 0.99 for normal subjects. The results showed that the novel algorithm is very useful for estimating not only hemiplegic gait but also normal gait.