In recent years, recognizing human athletic actions has become an increasingly significant part of the analysis. FPGA-based recognition has been developed to precisely predict human actions for future training. To ensure the viability of such long-term athletics training monitoring systems, it is necessary to have intelligent surroundings in which human athletics actions may be automatically identified. This Artificial Neural Network (ANN) based Field-Programmable Gate Array (FPGA) is implemented in the system described here. The data acquired from a smart environment may be utilized for its machine learning algorithms to infer human behavior; however, these methods must first be trained on annotated datasets. Due to the required time and labor, the cost of creating data sets such as records and annotations is high. In addition, the difficulty of human athletics enables these athletes to precisely imitate more demanding forms of the sport. Hierarchical models have the potential to reflect relief more precisely, but determining the appropriate amount of complexity may be a challenging endeavour. Last but not least, to roll out an automated human behaviour monitoring system all over the globe, we need a model behaviour that can be applied to each new house and then fine-tuned so that it properly matches the actions of the people who live in that specific home. Instead, the technique is most beneficial for decreasing the need for annotations and maximizing the utility of annotations to acquire annotations when applied to a dataset that comprises weeks of data. This is because the approach can maximize the usefulness of annotations.