In this work, we presented how vision-based systems of artificial intelligence can be designed in order to gain insight into human posture and precisely classify postural abnormalities. For this, OpenPose and machine learning algorithms to analyze walking patterns of participants in order to observe anomalies during the gait cycle are investigated. The results indicate that the best re-sult classification of normal or abnormal gait can be achieved by combining the lateral and frontal perspective, by using angles features and by classifying the data with a linear SVM. Fur-thermore, our work is independent of the OpenPose framework, which means that our results can be used by any approach that can provide 2D coordinate data. In summary, this study pro-vides valuable insight into the human posture data and provides a reliable system for detecting abnormal gait.