This paper proposes a recognition algorithm for the degree of rehabilitation of walking individuals using a Doppler radar sensor. First, the time-series macro velocities of the whole body were estimated by applying a moving average filter and differentiating the phase of the complex signal for the walking individual. Next, the time-series micro velocities of both legs were obtained by subtracting the time-series macro velocities of the entire body from the time-series velocities extracted by applying the conventional method and Kalman filter to the spectrogram. The step time and maximum velocities corresponding to each leg can be decomposed by estimating the local maximum values in the time-series microvelocities of both legs, resulting in new values to show the real-time degree of rehabilitation for walking individuals. In the experiments, we observed that the proposed method was capable of successfully recognizing both normal and abnormal walking movements.