There are many techniques and methods for gait analysis, but the data are not easy for the clinician to read and to correlate clinically in patients with neurological deficits such as stroke. In this study, we tried to simplify the kinetic gait data of 102 functionally ambulant stroke patients in a more simple way through pattern recognition in addition to conventional gait parameters and compared them with 50 healthy subjects. All of the stroke patients had received rehabilitation programs and were functionally ambulant. The computer DynoGraphy (CDG) system was used for gait analysis. The foot contact patterns based on pattern recognition from the gaitline and bipedal cyclogram, were correlated to motor function and other gait parameters. The gaitline patterns of the stroke patients were classified into 3 patterns according to the length of the gaitline by comparison to the unaffected side (pattern 1=normal, pattern 2= mild abnormal and pattern 3=remarkable abnormal). The cyclogram patterns for the stroke patients were also classified as 4 patterns according to the symmetry of affected to unaffected side ( pattern1=normal, pattern 2=mild abnormal, pattern 3=moderate abnormal and pattern 4=severe abnormal). All the fifty healthy subjects were in pattern 1(normal pattern) of gaitline and cyclogram. The motor conditions of the stroke patients by Brunnstrom's classification were found significantly correlated to the patterns of gaitline and cyclogram (p＜0.05). In this study, we also applied the calculation of symmetry index and consistency index as parameters to evaluate the gait of stroke patients. We proposed that the symmetry index is a good indicator for hemiplegic gait, while the consistency index is good for ataxic gait in monitoring the motor improvement. The results of this study suggest that the change of walking ability through pattern recognition might be comprehensive for gait evaluation of stroke patients.