Gait recognition is a promising biometric technique that recognizes individuals based on their walking style captured from a distance. Most of the cutting-edge techniques for gait recognition, especially those based on deep learning, require intense computational power and large amounts of training data, which might not be practical in situations with limited resources or in real time applications. This paper proposes a new hybrid method, GaitReason, that combines appearance-based matching with an extra reasoning module to make gait recognition more accurate in the tough conditions mentioned above. The extended reasoning technique is a combination of several similarity measures such as Euclidean distance, structural similarity index, skewness, and kurtosis differences through a weighted fusion scheme. These statistical features will capture both global and local gait characteristics. The proposed approach enhances discrimination between individuals and improves robustness against occlusion, noise, clothing variation, and viewing conditions. Experimental results on the CASIA-B and OU_MVLP datasets show that the model accurately recognizes people under different noise levels. With competitive processing time, the GaitReason method is suitable for real-time or near real-time applications. A detailed ablation study is done on the CASIA-B dataset to show the importance of each feature.