Event based cameras are biological retinas that consist of micro second temporal resolution, sparse stream of asynchronous brightness-change events These features allow for efficient scene understanding, particularly in fast-moving or difficult lighting conditions. This survey is important progress in event-based vision, video-to-event conversion, Human Activity Recognition (HAR), and detecting abnormal or misbehavior patterns. We show the methods, datasets, representations, and learning models, emphasizing the growing trend of generating synthetic events from RGB videos to encourage wider use. Through the use of the information provided in this survey, we hope to provide suggestions for improving delinquency action recognition algorithms and to increase the speed of efficiency of said algorithms in real-time applications.