The identification of unusual human behaviour has become increasingly important in diverse areas such as security, healthcare, and entertainment, leading to a surge of interest in abnormal human activity recognition in recent times. Modern anomaly detection systems for video surveillance are adequate, but they are expensive to compute and need specialized gear. Existing methods for detecting suspicious activity in video surveillance rely on hand-crafted features and statistical methods, which may not be capable of capturing complex spatiotemporal relationships in video data. These methods require prior knowledge of the activity to be detected and may not generalize well to different environments. To overcome these limitations, an approach for recognizing abnormal behaviour based on CNN and Long Short-Term Memory (LSTM) has been proposed in this paper. A specialized LSTM model has been developed to capture extended patterns of spatial and temporal data, which is particularly useful for recognizing abnormal human activity. This technology has garnered significant interest due to its potential applications in various fields such as security, healthcare, and entertainment. Our proposed system performs exceptionally well in the recognition of anomalous behaviour on the DCSASS (Distributed Camera System for Activity and Situation Sensing) dataset of which it achieved a recognition rate of 94%.