Video action recognition is a significant research domain in the field of computer vision, which has various practical applications, such as surveillance systems, sports analysis, health monitoring, intelligent video analysis, etc. However, the videos obtained from the environment are noisy, meaning that the videos may be blurred, the lighting conditions may be poor, the objects may be occluded, the background may be cluttered, etc. As a result, the accuracy of the video action recognition system is compromised. In this paper, a video action recognition system is proposed, which is more efficient in handling noisy videos. To achieve this, the video is processed, and the frames are obtained. Then, the obtained frames are enhanced using the Super Resolution Generative Adversarial Network, after which the InceptionV3 convolutional neural network is employed for the action classification. By enhancing the video frames, the accuracy of the video action recognition system is improved. Experiment results prove that the proposed system is more efficient compared to the other systems.