Measurement with a single inertial motion measurement unit (IMU) can be useful for monitoring gait information in rehabilitation at hospital or healthcare at home. This study focused on stride length estimation with an IMU because our preliminary study showed different estimation errors between some hemiplegic subjects. This was considered to be caused by differences of foot movements during walking between healthy subjects and paralyzed subjects. Therefore, this study aimed at evaluating estimation error of stride lengths under different calculation methods of rotation matrix that is needed to obtain foot orientation. Experimental results showed that the rotation matrix calculated from Euler angles could be effective for gaits of healthy subjects, but not for some of paralyzed subjects. The rotation matrix calculated from quaternion was shown to be effective for both healthy and paralyzed subjects. Mean absolute error of stride length estimation with the quaternion-based method was smaller than about 5% for healthy subjects and the non-paralyzed side of hemiplegic subjects and smaller than 6% for paralyzed side of the hemiplegic subjects. Average values of estimation error were less than -4% for 3 hemiplegic subjects. Stride length estimation method and correction method of integration error would be required to be evaluated with many motor disabled subjects.