Herein, a novel methodology MVCNN for real-time human fall event detection in the compressed domain of videos using motion vectors (MV) and revamped YOLOv3 using CNN. The videos in MPEG-4 and H.264compressionare considered for the present study. Any video source without any prior setup could be considered by adapting the proposed method to various video codecs and camera settings. Existing algorithms for human fall event detection in a compressed domain video have some limitations in this regard such as (i) requirement of keyframes at a fixed interval, (ii) usage of P frames only, and (iii) support for a single codec only. These limitations are overcome in the proposed method by using arbitrary keyframe intervals, using both P and B frames, and supporting more than one codec, e.g.MPEG-4 and H.264 codecs. The experimentation is carried out using the benchmark datasets, namely, Le2i, UR, and Multiple Cameras Fall Event Datasets. The fall event detection accuracy of the proposed method in the compressed domain is found to be comparable to that observed in raw video data by using other recent methods. The proposed method outperforms other methods in the literature on the compressed domain by significantly improving the fall event detection inference in the compressed domain.